Intuit Inc. provides financial management and compliance products and services for consumers, small businesses, self-employed, and accounting professionals in the United States, Canada, and internationally. The company operates in three segments: Small Business & Self-Employed, Consumer, and Strategic Partner. The Small Business & Self-Employed segment provides QuickBooks online services and desktop software solutions comprising QuickBooks Enterprise, a hosted solution; QuickBooks Online Advanced, an online enterprise solution; QuickBooks Self-Employed solution; and QuickBooks Online Accountant and QuickBooks Accountant Desktop Plus solutions; payroll solutions, such as online payroll processing, direct deposit of employee paychecks, payroll reports, electronic payment of federal and state payroll taxes, and electronic filing of federal and state payroll tax forms. This segment also offers payment-processing solutions, including credit and debit cards, and ACH payment services; and financial supplies and financing for small businesses. Its Consumer segment provides TurboTax income tax preparation products and services; and personal finance. The company's Strategic Partner segment offers Lacerte, ProSeries, and ProFile desktop tax-preparation software products; and ProConnect Tax Online tax products, electronic tax filing service, and bank products and related services. It sells products and services through various sales and distribution channels, including multi-channel shop-and-buy experiences, Websites and call centers, mobile application stores, and retail and other channels. Intuit Inc. has a collaboration agreement with Red Hat, Inc. on Argo CD, a declarative continuous delivery tool for Kubernetes deployments. The company was founded in 1983 and is headquartered in Mountain View, California.

Intuit Inc. provides financial management and compliance products and services for consumers, small businesses, self-employed, and accounting professionals in the United States, Canada, and internationally. The company operates in three segments: Small Business & Self-Employed, Consumer, and Strategic Partner. The Small Business & Self-Employed segment provides QuickBooks online services and desktop software solutions comprising QuickBooks Online Advanced, a cloud-based solution; QuickBooks Enterprise, a hosted solution; QuickBooks Self-Employed solution; and QuickBooks Online Accountant and QuickBooks Accountant Desktop Plus solutions; payroll solutions, such as online payroll processing, direct deposit of employee paychecks, payroll reports, electronic payment of federal and state payroll taxes, and electronic filing of federal and state payroll tax forms. This segment also offers payment-processing solutions, including credit and debit cards, and ACH payment services; and financial supplies and financing for small businesses. Its Consumer segment provides TurboTax income tax preparation products and services; and personal finance. The company's Strategic Partner segment offers Lacerte, ProSeries, and ProFile desktop tax-preparation software products; and ProConnect Tax Online tax products, electronic tax filing service, and bank products and related services. It sells products and services through various sales and distribution channels, including multi-channel shop-and-buy experiences, websites and call centers, mobile application stores, and retail and other channels. Intuit Inc. has a collaboration agreement with Red Hat, Inc. on Argo CD, a declarative continuous delivery tool for Kubernetes deployments; and a strategic agreement with Visa Inc. The company was founded in 1983 and is headquartered in Mountain View, California.

Intuit

intuit

When the Clippers’ new Intuit Dome opens ahead of the 2024-25 season, it will challenge prevailing sports venue design and fan experience trends.

Clippers’ Steve Ballmer dives deep on Intuit Dome’s design aspects, creating both a modern and analog experience for true fans of the game

Beyond analysts&#x27; top -and-bottom-line estimates for Intuit (INTU), evaluate projections for some of its key metrics to gain a better insight into how the business might have performed for the quarter ended January 2024.

Intuit (INTU) Q2 Earnings Preview: What You Should Know Beyond the Headline Estimates

Intuit&#x27;s (INTU) second-quarter fiscal 2024 results are expected to have gained from strength in Online Ecosystem revenues.

Intuit (INTU) to Report Q2 Earnings: What's in the Offing?

The Noonification: Leetcode: Two-sum an Intuitive Approach (4/23/2024)

Imagine yourself as the proud owner of a souvenir shop inside the Spot-On Change Hotel. When closing the cash register, you notice an excess of 8 euros. It seems that a guest didn’t receive the exact change. This could tarnish the hotel’s reputation. To avoid this, you decide to solve the mystery of the incorrect change. Upon opening the shop’s cash system, you discover that the error occurred on two different accounts! You devise a plan: to visit each room and ask the guests what change they received in the shop until you find the two with the wrong bill. Arriving on the first floor, a guest reports receiving 4 euros in change. You calculate that you need to find a number that, when added to the 4 euros, results in 8. Moving to the second floor, the guest’s change is 5 euros. 4 + 5 results in 9, so it can’t be this one… Fifth floor, sixth… Tenth, NO, NO, NO! Since you didn’t find the value on the first attempt, you go down to the second floor and start asking all the guests again about their change so that you can compare it with the value from the next starting floor, exhausting, isn’t it? In computer science, this search method is called brute force, an extremely slow search method that works by comparing one item with the following ones in the array. This consumes a lot of time and is not efficient, imagine going up and down all the floors of the Spot-On Change Hotel several times! This is not practical for you and also not practical for the computer. If you’re curious about how this going up and down the floors would be in a software program, it would look like this in JavaScript: twoSum(nums, target) {
 for (let i = 0; i < nums.length; i++) {
 for (let j = i + 1; j < nums.length; j++) {
 if (nums[i] + nums[j] === target) {
 return [i, j];
 }
 }
 }
} After assessing the situation calmly, you realize that there’s a more efficient way to find the two guests with the wrong change. You refine your initial mathematical intuition that 9 euros equals x + y. Applying a bit of arithmetic, you realize that: x + y = 9 is the same as saying that y = 9 — x, and that makes all the difference when increasing the efficiency of your search. Another thing that changed in your strategy is that now you will take a notepad to write down the values that the guests tell you, so you don’t have to ask the same value twice for the guests, and you don’t have to spend the day going up and down stairs. (What a terrible time for the elevator to break!). With the new tools in hand, you head to the first floor, where a guest informs you that they have 5 euros in change. You record this as {5: 0}, indicating a value of 5 at position 0. Substituting x with 5 in the equation, you calculate y = 8–5, resulting in y = 3. Since your notepad is still empty, you don’t have the number 3 recorded, which would be the complementary number to 5 to reach the result 8. You then write down the number 5 on your notepad (remember to always write down the number verified at the moment and not its complement; you will only use the complement to compare it with the target number that you have written down). You proceed to the second guest, who mentions receiving 2 euros in change. Applying the formula again: y = 8–2 => y = 6, you check your record, where you don’t find any number 6. Thus, you add the 2 to your record, which now stands as {5: 0, 2: 1}. Continuing the search, the next guest reports receiving 3 euros in change. You calculate again: y = 8–3 => y = 5. Bingo! You have a 5 recorded on your notepad! The guest from floor 0 is complementary to the one from floor 2! This approach is known as a hash table, and it’s much faster and efficient, both for you and the computer. In JavaScript, this would be implemented as follows: twoSum(nums, target) {
 const myTable = {};
 for (let i = 0; i < nums.length; i++) {
 const complementaryNumber = target - nums[i];
 if (complementaryNumber in myTable) {
 return [i, myTable[complementaryNumber]];
 }
 myTable[nums[i]] = i;
 }
} This is an easy problem from Leetcode that only becomes easy after you understand the concepts behind it. I recommend that you try to solve the problem on your own and spend some time trying to build intuition behind it, create your own analogy, review, and try to write pseudocode before moving on to the solution. I hope my analogy has helped you better understand the concepts behind the two sum challenge. See you next time! Imagine yourself as the proud owner of a souvenir shop inside the Spot-On Change Hotel. When closing the cash register, you notice an excess of 8 euros. It seems that a guest didn’t receive the exact change. This could tarnish the hotel’s reputation. To avoid this, you decide to solve the mystery of the incorrect change. Upon opening the shop’s cash system, you discover that the error occurred on two different accounts! You devise a plan: to visit each room and ask the guests what change they received in the shop until you find the two with the wrong bill. Arriving on the first floor, a guest reports receiving 4 euros in change. You calculate that you need to find a number that, when added to the 4 euros, results in 8. Moving to the second floor, the guest’s change is 5 euros. 4 + 5 results in 9, so it can’t be this one… Fifth floor, sixth… Tenth, NO, NO, NO! Since you didn’t find the value on the first attempt, you go down to the second floor and start asking all the guests again about their change so that you can compare it with the value from the next starting floor, exhausting, isn’t it? In computer science, this search method is called brute force, an extremely slow search method that works by comparing one item with the following ones in the array. This consumes a lot of time and is not efficient, imagine going up and down all the floors of the Spot-On Change Hotel several times! This is not practical for you and also not practical for the computer. If you’re curious about how this going up and down the floors would be in a software program, it would look like this in JavaScript: twoSum(nums, target) {
 for (let i = 0; i < nums.length; i++) {
 for (let j = i + 1; j < nums.length; j++) {
 if (nums[i] + nums[j] === target) {
 return [i, j];
 }
 }
 }
} twoSum(nums, target) {
 for (let i = 0; i < nums.length; i++) {
 for (let j = i + 1; j < nums.length; j++) {
 if (nums[i] + nums[j] === target) {
 return [i, j];
 }
 }
 }
} After assessing the situation calmly, you realize that there’s a more efficient way to find the two guests with the wrong change. You refine your initial mathematical intuition that 9 euros equals x + y. Applying a bit of arithmetic, you realize that: x + y = 9 is the same as saying that y = 9 — x, and that makes all the difference when increasing the efficiency of your search. Another thing that changed in your strategy is that now you will take a notepad to write down the values that the guests tell you, so you don’t have to ask the same value twice for the guests, and you don’t have to spend the day going up and down stairs. (What a terrible time for the elevator to break!). With the new tools in hand, you head to the first floor, where a guest informs you that they have 5 euros in change. You record this as {5: 0}, indicating a value of 5 at position 0. Substituting x with 5 in the equation, you calculate y = 8–5, resulting in y = 3. Since your notepad is still empty, you don’t have the number 3 recorded, which would be the complementary number to 5 to reach the result 8. You then write down the number 5 on your notepad (remember to always write down the number verified at the moment and not its complement; you will only use the complement to compare it with the target number that you have written down). You proceed to the second guest, who mentions receiving 2 euros in change. Applying the formula again: y = 8–2 => y = 6, you check your record, where you don’t find any number 6. Thus, you add the 2 to your record, which now stands as {5: 0, 2: 1}. Continuing the search, the next guest reports receiving 3 euros in change. You calculate again: y = 8–3 => y = 5. Bingo! You have a 5 recorded on your notepad! The guest from floor 0 is complementary to the one from floor 2! This approach is known as a hash table, and it’s much faster and efficient, both for you and the computer. In JavaScript, this would be implemented as follows: twoSum(nums, target) {
 const myTable = {};
 for (let i = 0; i < nums.length; i++) {
 const complementaryNumber = target - nums[i];
 if (complementaryNumber in myTable) {
 return [i, myTable[complementaryNumber]];
 }
 myTable[nums[i]] = i;
 }
} twoSum(nums, target) {
 const myTable = {};
 for (let i = 0; i < nums.length; i++) {
 const complementaryNumber = target - nums[i];
 if (complementaryNumber in myTable) {
 return [i, myTable[complementaryNumber]];
 }
 myTable[nums[i]] = i;
 }
} This is an easy problem from Leetcode that only becomes easy after you understand the concepts behind it. I recommend that you try to solve the problem on your own and spend some time trying to build intuition behind it, create your own analogy, review, and try to write pseudocode before moving on to the solution. Leetcode I hope my analogy has helped you better understand the concepts behind the two sum challenge. See you next time!

Leetcode: Two-sum an Intuitive Approach

লিটকোড: দুই-সমষ্টি একটি স্বজ্ঞাত পদ্ধতি

Leetcode: Zweisummen, ein intuitiver Ansatz

Leetcode: dos sumas y un enfoque intuitivo

Leetcode : une approche intuitive à deux sommes

लेटकोड: दो-योग एक सहज दृष्टिकोण

Leetcode: 2 つの合計を直感的に計算するアプローチ

Leetcode: 직관적인 접근 방식의 2합

Leetcode: duas somas e uma abordagem intuitiva

Leetcode: интуитивный подход с двумя суммами

Leetcode: İki Toplamlı Sezgisel Bir Yaklaşım

Leetcode: Tổng hợp hai cách tiếp cận trực quan

Leetcode：二和的直观方法

Authors: (1) Savvas Petridis, Google Research, New York, New York, USA; (2) Ben Wedin, Google Research, Cambridge, Massachusetts, USA; (3) James Wexler, Google Research, Cambridge, Massachusetts, USA; (4) Aaron Donsbach, Google Research, Seattle, Washington, USA; (5) Mahima Pushkarna, Google Research, Cambridge, Massachusetts, USA; (6) Nitesh Goyal, Google Research, New York, New York, USA; (7) Carrie J. Cai, Google Research, Mountain View, California, USA; (8) Michael Terry, Google Research, Cambridge, Massachusetts, USA. Table Of Links Abstract & Introduction Related Work Formative Study Constitution Maker Implementation User Study Findings Discussion Conclusion and References 7 FINDINGS Quantitative Findings. From the exit interviews, 12 of 14 participants preferred ConstitutionMaker to the baseline version. The results from the questionnaire are summarized in Figure 4. We found that ConstitutionMaker was perceived to be more helpful for writing rules that effectively guided the bot (Z = 8, p = .007), scoring on average 5.79 (𝜎 = 1.01) whereas the baseline scored 4.0 (1.51). When participants rewound parts of the conversation, they thought the ConstitutionMaker principles were followed by the bot better. Next, participants felt it was significantly easier (Z = 10.5, p = .006) to convert their feedback into principles with ConstitutionMaker (𝜇 = 5.86, 𝜎 = 0.91) than with the baseline (𝜇 = 3.93, 𝜎 = 1.39). The automatic conversion of kudos and critiques into principles eased the process of converting intuitive feedback into clear criteria for the bot to follow. Participants perceived that they were significantly more efficient (Z = 5, p = .004) writing rules with ConstitutionMaker (𝜇 = 5.86, 𝜎 = 1.51) than with the baseline (𝜇 = 3.64, 𝜎 = 1.34). Finally, participants also felt that writing principles with ConstitutionMaker (𝜇 = 3.0, 𝜎 = 1.41) required significantly less mental demand (Z = 1.5, p = .002) than with the baseline (𝜇 = 5.21, 𝜎 =1.78). There was no statistically significant difference for diversity (Z = 19, p = .06); participants felt that they exercised their creativity and wrote relatively diverse principles in the baseline. Next, regarding the feature usage metrics, participants wrote significantly more principles (t(13)=4.73, p < .001) with ConstitutionMaker than with the baseline; participants wrote on average 6.78 (𝜎 = 2.11) principles per chatbot with ConstitutionMaker and 4.42 (𝜎 = 1.24) principles with the baseline. Of the 95 principles written in the ConstitutionMaker condition, 40 (42.1%) came from kudos, where 37 were selected from the generate rationales and 3 were written; 28 (29.5%) came from critique, where 8 were selected and 20 were written; 13 (13.6%) came from the rewrite features; 14 (14.7%) were manually written. Participants found rewriting a bit cumbersome and preferred the less intensive workflow of describing what they liked or did not like to generate principles. In the following sections, we provide further context to these findings. 7.1 Participants’ Workflows for Writing Principles The two conditions led to quite different workflows for writing principles. In the ConstitutionMaker condition, participants commonly scanned the three candidate outputs from the chatbot, identified the output they liked the most, and then gave kudos to that output if they thought it had a quality that was not currently reflected in their principles. For example, while P1 was working on FoodBot, he was asking for easy-to-make vegetarian dishes, and in one of the bot’s candidate outputs, each suggested dish had a short description and explanation on why it was easy. He appreciated this, skimmed the kudos, and then selected one that conveyed this positive feature of the bot’s output. However, if participants disliked all of the responses, they would then switch to critiquing one of the outputs. Accordingly, this kudos-first workflow helps to explain how most of the principles participants wrote came from that principle elicitation feature. Meanwhile, in the baseline condition, participants generally wrote principles when the bot deviated quite a bit (in a negative way) from what they expected. P8 explained, “Here it feels like what I more naturally do is write corrective rules, to guardrail anything that goes weird...If it’s already doing the right thing it doesn’t need a rule from me. I wouldn’t feel the need to write those.” In the baseline condition, participants only see one candidate output from the chatbot, and this might deemphasize the stochastic nature of the chatbot’s outputs. As a result, when participants are okay with a response, they could feel that they do not need to write a principle to encourage that kind of response further. Overall, with the baseline, participants predominantly wrote principles to steer the LLM from less optimal behavior, while with ConstitutionMaker, participants mostly used kudos the most to encourage behavior they liked. 7.2 ConstitutionMaker Supported Users’ Thought Processes In the following section, we discuss how ConstitutionMaker supported participants’ thought process from (1) forming an intuition on ways the chatbot could be improved, (2) expressing this intuition as feedback, and (3) converting this feedback into a specific and clear principle. As P5 was using the baseline after using ConstitutionMaker, she explained how she wished she could see multiple outputs again: “Sometimes, I don’t know what I’m missing [in the baseline]. I’m thinking of the Denali hiking example [which occurred when she wrote principles for VacationBot with ConstitutionMaker]. Two of the responses didn’t mention that Denali was good for young children. But one did, and I was able to pull that out as a positive principle.” While she was writing principles for VacationBot with ConstitutionMaker, P5 started off the conversation saying she was looking for suggestions for her family, which included two young children. As the conversation progressed and a general location was established, P5 then asked for hiking recommendations, for which the bot gave some, but only one of its responses highlighted that the hikes it was recommending were good for young children. P5 gave kudos to that response and created the following principle: “Consider information previously inputted by the user when providing recommendations.” Ultimately, it can be hard to form opinions on responses without getting exposed to alternatives, so by providing multiple chatbot outputs, ConstitutionMaker supported participants in forming an intuition on how the model might be steered. 7.2.1 Multiple chatbot outputs helped participants form an intuition on how the model could be steered. . Upon seeing a candidate response from the chatbot, participants could intuitively tell if they liked or disliked it, but struggled to articulate their thoughts. The automatically generated kudos and critiques helped participants recognize and formulate this feedback. For example, while working on FoodBot with ConstitutionMaker, P9 asked the bot to identify the pizzeria with the best thin crust from a list of restaurants provided in a prior turn. The bot responded with, “Pizzaiolo has the best thin crust pies.” P9 knew he did not like the response, so he went to critique it and selected the following generated option: “This response is bad because it does not provide any information about the other pizza places that the user asked about.” The following principle was generated: “If the user asks about a specific attribute of a list of items, provide information about all of the items in the list that have that attribute,” which then produced a set of revised responses that compared the qualities of each pizzeria’s crusts. Reflecting on this process, P9 stated, “I didn’t like that last answer [from FoodBot], but I didn’t have a concrete reason why yet...I didn’t really know how to put it into words yet...but reading the suggestions gave me at least one of many potential reasons on why I didn’t like the response.” Thus, ConstitutionMaker helped participants transition from fast thinking [15], that is, their intuitive and unconscious responses to the bot’s outputs, to slow thinking, a more deliberate, conscious formulation of their feedback. 7.2.2 Automatically providing kudos and critique rationales helped participants formulate their intuitive feedback Sometimes the generated kudos and critique rationales did not capture participants’ particular feedback on the chatbot, and so they would then write their own. Their feedback was often under-specified, and ConstitutionMaker helped convert this feedback into a clear and specific principle. For example, P4 was writing principles for VacationBot using ConstitutionMaker. During the conversation, they had told VacationBot that they were planning a week long vacation to Japan, to which the bot immediately responded with a comprehensive 7-day itinerary. P4 then wrote in their own critique: “This response is bad because it does not take into account the user’s interests.” The resulting principle was, “When the user mentions a location, ask them questions about what they are interested in before providing an itinerary.” This principle was aligned with what P4 had in mind, and reflecting on her experience using ConstitutionMaker, she stated, Finally, even when the resulting principle was not exactly what they had in mind, participants appreciated the useful starting point it provided. Along these lines, P11 explained, Overall, ConstitutionMaker helped participants specify their feedback into clear principles. 7.2.3 Generating principles from feedback helped users write clear, specific principles. “When I would critique or kudos, it would give examples of principles that were putting it into words a little bit better than I could about like what exactly I was trying to narrow down to here.” “It was easier to say yes-and with Tool A [ConstitutionMaker]. Where it [the generated principle] wasn’t all the way there, but I think it’s 50% of the way there, and I can get it to where I want to go.” 7.3 Participants’ Workflows Introduced Challenges with Writing Principles Participants struggled to find the right level of granularity for their principles, and the two conditions led to different problems in this regard. Both workflows had participants switch roles from end-user, where participants experimented with different user journeys, to bot designer, where they evaluated the bot’s responses to write principles. The more conversation-forward interface of the baseline blurred the distinction between these two roles. P3 explained that without the multiple bot outputs and principle elicitation features, And by leaning further into this user role, participants wrote principles that were more conversational, but under-specified. For example, while writing principles for FoodBot with the baseline, P11 wrote the principle “Be cognizant of the user’s dietary preferences.” What P11 really had in mind was a principle that specified that the bot should ask the user for their preferences and allergies prior to generating a meal plan. These underspecified principles often did not impact the bot’s responses and would frustrate participants while they used the baseline. “you can simulate yourself as the user a lot better in this mode [the baseline].” Meanwhile, while using ConstitutionMaker, the opposite problem occurred, where users’ workflows led to principles that were over-specified. For example, while working on VacationBot, P7 asked the model to help him narrow down a few vacation options, and the model proceeded to ask them questions (without any principle written specifying so). Appreciating that the model was gathering context, they selected a kudos that praised the model for asking about the user’s budget constraints prior to recommending a vacation destination. The resulting principle was, “Ask the user their budget before providing vacation options.” However, once this principle came into effect, the model’s behavior then anchored specifically to only asking for budget prior to making a recommendation. And so, this workflow of providing feedback at every conversational step, instead of for entire conversations, led to a principle that was too specific and impacted the bot’s performance negatively. While users generally appreciated ConstitutionMaker’s ability to form specific principles from their feedback, there were rare instances where the principles were too specific. Finally, in both conditions, by switching back and forth between the end-user to bot designer roles, participants would sometimes write principles that conflicted with each other. For example, while P2 was working on VacationBot with the baseline, they asked the bot for dog-friendly hotel recommendations in the Bay Area, and VacationBot responded with three recommendations. P2 wanted more recommendations and wrote a principle to “Provide >= 10 recommendations.” Later on in the conversation, P2 now had a list of dog-friendly restaurants, with their requisite costs, and he asked VacationBot which it recommends, to which it responded by listing positive attributes of all the hotel options. P2, who now wanted a decisive, single response wrote the following principle: “If I ask for a recommendation, give *1* recommendation only.” VacationBot, now with two conflicting principles on the number of recommendations to provide, alternated between the two. Ultimately, by providing feedback on individual conversational turns, participants ended up with conflicting principles. P8 imagined a different workflow, where he would experiment with full user journeys and then write principles: In summary, one’s workflow as they probe and test the model impacts the types of principles they produce and the challenges they face. “I think it might help me to actually go through the [whole] user flow and then analyze it as a piece instead of switching...it would allow me to inhabit one mindset [either bot designer or user] for a period of time and then switch mindsets.” 7.4 The Limits of Principles Some participants questioned if writing natural language principles was the optimal way to steer all aspects of a bot’s behavior. While writing a principle to shorten the length of the chatbot’s responses, P13 reflected, They imagined that aspects like the form of the model’s responses would be better customized with UI elements like sliders to adjust the length of the bot’s responses, or exemplifying the structure of the bot’s response (e.g., an indented, numbered list for recommendations) for the model to follow, instead of describing these requests in natural language. In a similar vein, P14 noticed that her principles pertained to different parts of the conversation, and as a list, they seemed hard to relate to each other. She wanted to structure and provide feedback on higher-level “conversational arcs,” visually illustrating the flow and “forks in the road” of the conversation (e.g., “If the user does X, do Y. Otherwise, do Z”). Principles are computational in a sense, and they dictate the ways the conversation can unfold; there might be better ways to let users author this flow, other than with individual principles. “It feels a little weird to use natural language to generate the principles...it doesn’t feel efficient, and I’m not sure how it’s going to interpret it.” This paper is under CC 4.0 license. available on arxiv

Efficiency and Intuition: Key Factors in User Preference for ConstitutionMaker

Authors: (1) Savvas Petridis, Google Research, New York, New York, USA; (2) Ben Wedin, Google Research, Cambridge, Massachusetts, USA; (3) James Wexler, Google Research, Cambridge, Massachusetts, USA; (4) Aaron Donsbach, Google Research, Seattle, Washington, USA; (5) Mahima Pushkarna, Google Research, Cambridge, Massachusetts, USA; (6) Nitesh Goyal, Google Research, New York, New York, USA; (7) Carrie J. Cai, Google Research, Mountain View, California, USA; (8) Michael Terry, Google Research, Cambridge, Massachusetts, USA. Table Of Links Abstract & Introduction Related Work Formative Study Constitution Maker Implementation User Study Findings Discussion Conclusion and References 3 FORMATIVE STUDY To understand how to support users with writing principles for chatbots, we conducted (1) a one-hour formative study, where we observed eight industry professionals write principles for chatbots of their choice. These participants all had prompting experience. Two participants are designers and six are software engineers, all at a large technology company. During the workshop, participants used an early version of ConstitutionMaker, without principle elicitation features. They spent 25 minutes writing principles for their chatbot. Afterwards, we discussed the difficulties they faced while writing principles. Finally, we collected the principles they wrote and classified them to understand the kind of principles they wanted to write. 3.1 Design Goals In this section, we summarize a set of three design goals for ConstitutionMaker we established from the formative workshop and subsequent think-alouds. D.1 Today’s LLMs are quite sophisticated, and even with just a preamble describing how the bot should behave, the chatbot can hold a convincing conversation. Because of this, participants mentioned that it was sometimes hard to imagine how the chatbot’s responses could be improved. This did not mean, however, that they thought the chatbot’s response was perfect, but instead passable and without any glaring errors. Therefore, to help participants recognize better kinds of responses to steer the chatbot to, our first design goal was to provide multiple candidate responses from the chatbot at each conversational turn. This way, participants can compare them and recognize components they like more than others. Help users recognize ways to improve the chatbot’s responses by showing alternative chatbot responses. D.2 One piece of feedback we got from participants was that writing principles involves a difficult two-step process of first (1) articulating one’s feedback on the model’s current output, and then (2) converting this feedback into a principle for the LLM to follow. Often, one’s initial reaction to the model’s output is intuitive, and converting that intuition into a principle for the chatbot to follow can be challenging. In addition, once participants had a particular bit of feedback in mind (e.g., “I don’t like how the chatbot didn’t introduce itself”), they were unsure how to phrase their principle. However, in line with prior research [45, 46], they found that more concrete principles that specified what should happen and when (e.g., “Introduce yourself at the start of the conversation, and state what you can help with”) generally led to better results. Thus, our second design goal was to help users go from their initial reaction to the model’s output to a specific, clearly written principle to steer the model. Help convert user feedback into specific principles to make principle writing easier. D.3 As participants wrote more principles, they wanted ways to test these principles to make sure they worked. The early version of ConstitutionMaker only let users restart the conversation, and did not let users enable or disable principles. Users wanted to test individual principles on certain portions of the conversation, to see if the model was generating the correct content. And so, our last design goal was to enable easier testing of principles. Enable easier testing of principles to help users understand how well their principles are steering the chatbot’s behavior. 3.2 Principle Classification From the formative workshop and follow up sessions, we collected 79 principles in total and classified them to understand the kinds of principles users wanted to write. These principles correspond to a number of very different chatbots, including a show recommender, chemistry tutor, role playing game manager, travel agent and more. We describe common types of principles below. Unconditional principles are those that apply at every conversational turn. Examples include: (1) those that define a consistent personality of the bot (e.g., “Act grumpy all the time” or “Speak informally and in the first person”), (2) those that place guardrails on the conversational content (e.g., “Don’t talk about anything but planning a vacation”), and (3) those that establish a consistent form for the bot’s responses (e.g., “Limit responses to 20 words”). Meanwhile, a conditional principle only applies when a certain condition is meant. For example, “Generate an itinerary after all the information has been collected,” only applies to the conversation when some set of information has been acquired. Writing a conditional principle essentially defines a computational interaction; users establish a set of criteria that make the principle applicable to the conversation, and once that set of criteria is met, the principle is executed (e.g., an itinerary is generated). Principles can be either unconditional or conditional. For example, “Generate an itinerary after all the information has been collected” depends on the entire conversation history to determine if all of the requisite information has been collected. Similarly the following principle written for a machine learning tutor, “After verifying a user’s goal, provide help to solve their problem,” depends on the conversation history to identify if the user’s goal has been verified. Meanwhile, the principle “When the user says they had a particular cuisine the night before, recommend a different cuisine,” written for a food recommender, pertains just to the latest response by the user. Finally, the condition can depend on the action the bot is about to take, like “When providing a list of suggestions, use free text rather than bullet points,” which applies to any situation when the bot thinks it is appropriate to make suggestions. Conditional principles can depend on the entire conversation history, the user’s latest response, or the action the bot is about to take. For example, the principle “At the start of the conversation, introduce yourself and ask a fun question to kick off the conversation” is fulfilled in a single conversational turn, in which the bot introduces itself. Similarly, “Before recommending a restaurant, ask the user for their location” is also fulfilled in a single turn. Meanwhile, for a role playing game (RPG) bot that guides the user through an adventure, a participant wrote the following principle: “When the user tries to do something, put up small obstacles. Don’t let them succeed on the first attempt.” This principle implies that the bot needs to take action multiple turns prior to being fulfilled (e.g., by first putting up a small obstacle and then subsequently letting the user succeed). Similarly, for a travel agent bot, a user wrote “Ask questions one-by-one to get an idea of their preferences,” which also requires multiple conversational turns prior to fulfillment. Conditional principles can be fulfilled in a single or multiple conversational turns. In summary, principles can either be conditional, where they apply when a certain condition is met, or unconditional, where they apply at every conversational step. Conditional principles further break down into those that depend on the entire conversation history, the user’s last response, or the action the bot is about to take. And finally, conditional principles are either fulfilled in a single turn or multiple conversational turns. This paper is under CC 4.0 license. available on arxiv

Formative Study Illuminates the Path to Intuitive Chatbot Customization

Parachains have become extremely popular for building unique use cases, all while maintaining a seamless interoperability amongst various parachains and blockchains, especially within the Polkadot ecosystem. Yet, beneath the surface of coding and implementing business logic on parachains, a labyrinth of vital activities — crucial for their deployment, management, and scaling — often remain overlooked. DevOps Before any parachain goes into production, it’s essential to subject it to intense testing within a simulated mainnet environment. Zombienet becomes an invaluable tool in this context to test parachain’s resilience by taking it through numerous brute force scenarios, replicating every possible real-world challenge it might encounter. This includes simulations with all node types, from full nodes to validator nodes, and stress-testing configurations to ensure stability and optimal behavior. A discernible gap also lies in the domain of advanced analytics and proactive monitoring, both imperative to validate performance and ensure readiness for the mainnet. In pursuit of a solution, submitted a detailed grant proposal at Web3 Foundation. Recognizing the merit in addressing this operational gap, the foundation lent its support, awarding a grant to Zeeve for the development of a GUI tool (LARCH) designed to automate the parachain simulation testing workflow using Zombienet. Zeeve Details of the Grant: We are happy to announce that we have successfully completed both milestone I and Milestone II of our grant from the Web3 Foundation. This grant has enabled us to focus on the development of for enhanced management of Zombienet with automation and monitoring. Larch More details of the grant can be found . here The software is now available under an open-source license on GitHub and accessible to everyone, complete with its source code, developer documentation and tutorials. We’re confident that Larch will significantly benefit the Polkadot/Substrate parachain developer ecosystem, extending its advantages to the larger Web3 community. What is Larch? Zeeve’s Larch is a comprehensive frontend for Zombienet CLI, seamlessly integrating with K8 and Podman. Larch is thoughtfully designed to simplify the Zombienet network creation, management, and testing processes. With it, developers and parachain teams can swiftly set up the relay chains and parachains with Zombienet, unlocking the potential to test a plethora of configurations, easily, right from the dashboard. Key Features of Larch: Parachain setups are crucial and complex. Larch enables a simple UI driven Zombienet setup, including relaychain and parachain core configurations. Comprehensive Automation: : Larch allows for a ‘pick and choose’ based selection for genesis parameters and chain configurations, crucial for initializing parachain setups tailored for the desired Zombienet. Intuitive Configurations Choose from the existing zombienet configurations, save them as templates, or quickly modify and retest using previously saved templates. Users can also effortlessly upload new ones without delving into manual configuration creation. Flexible Replication: Users can conveniently execute key operations such as start, delete, view progress & metrics directly from the dashboard, ensuring optimal testing flexibility. More operational control: Supports the setup of all node types, be it Full node, Validator node, or Collator node. All-Inclusive Node Setup: Enables developers to monitor live logs during test runs and review post-run results directly through the interface. Real-time Logging: Incorporates a stack of Prometheus and Grafana for effortless monitoring of Zombienet. Advanced Monitoring: Larch’s setup is universal and can be installed on any cloud / local machine of choice with Linux based OS. Cloud Agnostic: In summation, Zeeve’s Larch is set to transform the way substrate & parachain developers interact with Zombienet, offering a simplified, efficient, and intuitive experience. Deploying a Zobienet Network using Larch in a few clicks: This tutorial focuses on creating a template and subsequently, a network from this template. The intention is to emulate the official Zombienet example found at . Zombienet on GitHub This guide is tailored for Linux systems. For Mac users, please refer to the Larch . Although Mac isn’t officially supported yet, if Podman emulation operates smoothly, then Larch should be able to manage the network effectively. Note: Prerequisites section Likewise, execution on a remote Linux cluster via a Mac’s setup should function. K8 kubectl Step 1: Get the larch binary either by downloading from the or yourself. releases build it Step 2: Ensure Podman / kubectl is installed as per the on your system for the orchestration of the Zombienet networks. instructions Step 3: Assign execute permissions to the larch binary using: chmod 755 larch Run the larch binary by doing: ./larch Step 4: Navigate to in your browser. Click on the ‘ button located at the page’s top. http://localhost:9000 Network Template +’ : Step 5 Label the template as “ “. Choose either Podman or Kubernetes as the Provider based on your setup. Proceed by clicking . sample-template ‘Next’ : Step 6 Under the section, populate the subsection. Insert “ ” as the name and select the ‘ option. By pressing the button, add a new node. For this node, input “ ” and again choose ‘ . Press . Relaychain Configuration Nodes alice Validator’ + bob Validator’ Next : Step 7 In the section, tap on the ‘ ‘ button. Set the Parachain ID to 100 (or any other number). Name the collator “ ” and for the command, use ‘ . Click ‘ . Parachain Configuration + collator01 polkadot-parachain’ Next’ : Step 8 For this tutorial, we’ll bypass . Simply click ‘ . HRMP Next’ : Step 9 Within the section, after the pre-filled sections, append the necessary configuration. Store the template by hitting ‘ . Test Configuration Save’ : Step10 On the Templates page, the new template should now be visible. Select ‘ to initiate the network. A dialogue box will pop up – input the network’s name and click ‘ to start the orchestration. The app will transition you to the networks page while in the creation state. Create’ Confirm’ : Step 11 To inspect command and execution logs, click on the ‘ icon. For a detailed command, tap the ‘ button; a modal will pop up with the command and other metadata. To peruse logs/outputs, select the ‘ button, opening another modal revealing the streams. Eye’ Command’ Output’ The preceding step concludes the network creation. : Step 12 To access metrics of the deployed network, head to the and tap the ‘ icon. This redirects you to the Grafana service (default login: admin/admin). Within the Dashboards segment, locate and open the dashboard from the ‘ folder to inspect metrics. My Network page Monitoring’ General’ Wrapping up: Zeeve is deeply committed to the web3 ecosystem. As we conclude, our goal is clear: we want ‘Larch’ getting adopted widely and put to use in a diverse parachain testing scenarios. We are encouraging developers to utilize it for their testing needs and share feedback. Zeeve will continue to update and enhance Larch, adding new features based on community inputs. Together, we’ll keep iterating on Larch, ensuring it remains relevant and beneficial for all developers. Polkadot/Substrate

Zeeve's Larch: Simplifying Polkadot/Substrate Zombienet Operations with an Intuitive GU

জিভের লার্চ: একটি স্বজ্ঞাত জিইউ দিয়ে পোলকাডট/সাবস্ট্রেট জম্বিনেট অপারেশনকে সরলীকরণ করা

Zeeve's Larch: Vereinfachung der Polkadot/Substrat-Zombienet-Operationen mit einer intuitiven GU

Alerce de Zeeve: simplificando las operaciones de Polkadot/Substrate Zombienet con una GU intuitiva

Zeeve's Larch : simplifier les opérations Polkadot/Substrate Zombienet avec un GU intuitif

ज़ीव्स लार्च: एक सहज ज्ञान युक्त जीयू के साथ पोलकाडॉट/सब्सट्रेट ज़ोम्बीनेट संचालन को सरल बनाना

Zeeve's Larch: 直感的な GU でポルカドット/サブストレート ゾンビネットの運用を簡素化

Zeeve's Larch: 직관적인 GU로 Polkadot/Substrate Zombienet 운영 단순화

Zeeve's Larch: simplificando as operações Polkadot/Substrate Zombienet com uma GU intuitiva

Лиственница Зива: упрощение операций Polkadot/Substrate Zombienet с помощью интуитивно понятного графического интерфейса

Zeeve's Larch: Sezgisel Bir GU ile Polkadot/Substrate Zombienet Operasyonlarını Basitleştirme

Zeeve's Larch: Đơn giản hóa các hoạt động của Polkadot/Substrate Zombienet bằng GU trực quan

Zeeve 的 Larch：通过直观的 GU 简化 Polkadot/Substrate Zombienet 操作

PRINCIPLES OF REASON, OR INTUITIVE TRUTHS.

The Problem of Philosophy, by Bertrand Russells, is part of the HackerNoon Books Series. You can jump to any chapter in this book here . ON INTUITIVE KNOWLEDGE ON INTUITIVE KNOWLEDGE There is a common impression that everything that we believe ought to be capable of proof, or at least of being shown to be highly probable. It is felt by many that a belief for which no reason can be given is an unreasonable belief. In the main, this view is just. Almost all our common beliefs are either inferred, or capable of being inferred, from other beliefs which may be regarded as giving the reason for them. As a rule, the reason has been forgotten, or has even never been consciously present to our minds. Few of us ever ask ourselves, for example, what reason there is to suppose the food we are just going to eat will not turn out to be poison. Yet we feel, when challenged, that a perfectly good reason could be found, even if we are not ready with it at the moment. And in this belief we are usually justified. But let us imagine some insistent Socrates, who, whatever reason we give him, continues to demand a reason for the reason. We must sooner or later, and probably before very long, be driven to a point where we cannot find any further reason, and where it becomes almost certain that no further reason is even theoretically discoverable. Starting with the common beliefs of daily life, we can be driven back from point to point, until we come to some general principle, or some instance of a general principle, which seems luminously evident, and is not itself capable of being deduced from anything more evident. In most questions of daily life, such as whether our food is likely to be nourishing and not poisonous, we shall be driven back to the inductive principle, which we discussed in Chapter VI. But beyond that, there seems to be no further regress. The principle itself is constantly used in our reasoning, sometimes consciously, sometimes unconsciously; but there is no reasoning which, starting from some simpler self-evident principle, leads us to the principle of induction as its conclusion. And the same holds for other logical principles. Their truth is evident to us, and we employ them in constructing demonstrations; but they themselves, or at least some of them, are incapable of demonstration. Self-evidence, however, is not confined to those among general principles which are incapable of proof. When a certain number of logical principles have been admitted, the rest can be deduced from them; but the propositions deduced are often just as self-evident as those that were assumed without proof. All arithmetic, moreover, can be deduced from the general principles of logic, yet the simple propositions of arithmetic, such as 'two and two are four', are just as self-evident as the principles of logic. It would seem, also, though this is more disputable, that there are some self-evident ethical principles, such as 'we ought to pursue what is good'. It should be observed that, in all cases of general principles, particular instances, dealing with familiar things, are more evident than the general principle. For example, the law of contradiction states that nothing can both have a certain property and not have it. This is evident as soon as it is understood, but it is not so evident as that a particular rose which we see cannot be both red and not red. (It is of course possible that parts of the rose may be red and parts not red, or that the rose may be of a shade of pink which we hardly know whether to call red or not; but in the former case it is plain that the rose as a whole is not red, while in the latter case the answer is theoretically definite as soon as we have decided on a precise definition of 'red'.) It is usually through particular instances that we come to be able to see the general principle. Only those who are practised in dealing with abstractions can readily grasp a general principle without the help of instances. In addition to general principles, the other kind of self-evident truths are those immediately derived from sensation. We will call such truths 'truths of perception', and the judgements expressing them we will call 'judgements of perception'. But here a certain amount of care is required in getting at the precise nature of the truths that are self-evident. The actual sense-data are neither true nor false. A particular patch of colour which I see, for example, simply exists: it is not the sort of thing that is true or false. It is true that there is such a patch, true that it has a certain shape and degree of brightness, true that it is surrounded by certain other colours. But the patch itself, like everything else in the world of sense, is of a radically different kind from the things that are true or false, and therefore cannot properly be said to be true. Thus whatever self-evident truths may be obtained from our senses must be different from the sense-data from which they are obtained. It would seem that there are two kinds of self-evident truths of perception, though perhaps in the last analysis the two kinds may coalesce. First, there is the kind which simply asserts the existence of the sense-datum, without in any way analysing it. We see a patch of red, and we judge 'there is such-and-such a patch of red', or more strictly 'there is that'; this is one kind of intuitive judgement of perception. The other kind arises when the object of sense is complex, and we subject it to some degree of analysis. If, for instance, we see a round patch of red, we may judge 'that patch of red is round'. This is again a judgement of perception, but it differs from our previous kind. In our present kind we have a single sense-datum which has both colour and shape: the colour is red and the shape is round. Our judgement analyses the datum into colour and shape, and then recombines them by stating that the red colour is round in shape. Another example of this kind of judgement is 'this is to the right of that', where 'this' and 'that' are seen simultaneously. In this kind of judgement the sense-datum contains constituents which have some relation to each other, and the judgement asserts that these constituents have this relation. Another class of intuitive judgements, analogous to those of sense and yet quite distinct from them, are judgements of memory. There is some danger of confusion as to the nature of memory, owing to the fact that memory of an object is apt to be accompanied by an image of the object, and yet the image cannot be what constitutes memory. This is easily seen by merely noticing that the image is in the present, whereas what is remembered is known to be in the past. Moreover, we are certainly able to some extent to compare our image with the object remembered, so that we often know, within somewhat wide limits, how far our image is accurate; but this would be impossible, unless the object, as opposed to the image, were in some way before the mind. Thus the essence of memory is not constituted by the image, but by having immediately before the mind an object which is recognized as past. But for the fact of memory in this sense, we should not know that there ever was a past at all, nor should we be able to understand the word 'past', any more than a man born blind can understand the word 'light'. Thus there must be intuitive judgements of memory, and it is upon them, ultimately, that all our knowledge of the past depends. The case of memory, however, raises a difficulty, for it is notoriously fallacious, and thus throws doubt on the trustworthiness of intuitive judgements in general. This difficulty is no light one. But let us first narrow its scope as far as possible. Broadly speaking, memory is trustworthy in proportion to the vividness of the experience and to its nearness in time. If the house next door was struck by lightning half a minute ago, my memory of what I saw and heard will be so reliable that it would be preposterous to doubt whether there had been a flash at all. And the same applies to less vivid experiences, so long as they are recent. I am absolutely certain that half a minute ago I was sitting in the same chair in which I am sitting now. Going backward over the day, I find things of which I am quite certain, other things of which I am almost certain, other things of which I can become certain by thought and by calling up attendant circumstances, and some things of which I am by no means certain. I am quite certain that I ate my breakfast this morning, but if I were as indifferent to my breakfast as a philosopher should be, I should be doubtful. As to the conversation at breakfast, I can recall some of it easily, some with an effort, some only with a large element of doubt, and some not at all. Thus there is a continual gradation in the degree of self-evidence of what I remember, and a corresponding gradation in the trustworthiness of my memory. Thus the first answer to the difficulty of fallacious memory is to say that memory has degrees of self-evidence, and that these correspond to the degrees of its trustworthiness, reaching a limit of perfect self-evidence and perfect trustworthiness in our memory of events which are recent and vivid. It would seem, however, that there are cases of very firm belief in a memory which is wholly false. It is probable that, in these cases, what is really remembered, in the sense of being immediately before the mind, is something other than what is falsely believed in, though something generally associated with it. George IV is said to have at last believed that he was at the battle of Waterloo, because he had so often said that he was. In this case, what was immediately remembered was his repeated assertion; the belief in what he was asserting (if it existed) would be produced by association with the remembered assertion, and would therefore not be a genuine case of memory. It would seem that cases of fallacious memory can probably all be dealt with in this way, i.e. they can be shown to be not cases of memory in the strict sense at all. One important point about self-evidence is made clear by the case of memory, and that is, that self-evidence has degrees: it is not a quality which is simply present or absent, but a quality which may be more or less present, in gradations ranging from absolute certainty down to an almost imperceptible faintness. Truths of perception and some of the principles of logic have the very highest degree of self-evidence; truths of immediate memory have an almost equally high degree. The inductive principle has less self-evidence than some of the other principles of logic, such as 'what follows from a true premiss must be true'. Memories have a diminishing self-evidence as they become remoter and fainter; the truths of logic and mathematics have (broadly speaking) less self-evidence as they become more complicated. Judgements of intrinsic ethical or aesthetic value are apt to have some self-evidence, but not much. Degrees of self-evidence are important in the theory of knowledge, since, if propositions may (as seems likely) have some degree of self-evidence without being true, it will not be necessary to abandon all connexion between self-evidence and truth, but merely to say that, where there is a conflict, the more self-evident proposition is to be retained and the less self-evident rejected. It seems, however, highly probable that two different notions are combined in 'self-evidence' as above explained; that one of them, which corresponds to the highest degree of self-evidence, is really an infallible guarantee of truth, while the other, which corresponds to all the other degrees, does not give an infallible guarantee, but only a greater or less presumption. This, however, is only a suggestion, which we cannot as yet develop further. After we have dealt with the nature of truth, we shall return to the subject of self-evidence, in connexion with the distinction between knowledge and error. About HackerNoon Book Series: We bring you the most important technical, scientific, and insightful public domain books. This book is part of the public domain. Bertrand Russell (2009). THE PROBLEMS OF PHILOSOPHY. Urbana, Illinois: Project Gutenberg. Retrieved October 2022, from https://www.gutenberg.org/files/5827/5827-h/5827-h.htm This eBook is for the use of anyone anywhere at no cost and with almost no restrictions whatsoever. You may copy it, give it away or re-use it under the terms of the Project Gutenberg License included with this eBook or online at www.gutenberg.org , located at https://www.gutenberg.org/policy/license.html .

Understanding Intuitive Knowledge

Have you ever found yourself in a situation where you need to interact with a smart home device, Smart TV or Smart Display but you can't touch it? Imagine, you are in the kitchen with wet or dirty hands, and you need to switch to the next episode on your Smart TV. It's not practical to touch the remote or device in this situation. Similarly, if your phone starts ringing and you need to quickly pause the music playing on it, it's much more convenient to use voice commands than to touch the device. Thanks to gesture recognition technology and voice recognition, users can now control their devices multimodel: via gestures or voice control. My name is Daria and I am a Senior Product Manager with 5+ years of experience in the tech industry, working on products related to hardware, computer vision, and voice recognition, and in this article, we'll dive into gesture recognition systems, why and how these services are created, and what both product and technical teams need to consider in their development. I will share personal examples, my own experience, and reflections on the mistakes that I made along with my team when we were developing multimodal navigation for smart devices by SberDevices. Why? Enter touchscreens, the game-changer for tech interaction. Swipe, tap, and voila – complex actions made easy and user experiences amped up. But it wasn't always like this, before iPhone was released the only way to interact with a device was through a series of buttons. The iPhone's release marked a turning point, launching us into the touchscreen era, fueled by Apple's extensive research and development efforts. With its debut, the iPhone introduced an intuitive touch interface that eliminated the need for a stylus or buttons, setting it apart from competitors. But now, we're on the brink of another transition: from touchscreens to touchless interfaces. Touchscreens no longer suffice for users' needs, leading to the emergence of gesture control as an additional modality for interaction. Designing this new experience is an incredibly creative task. The first touchscreen prototype was featured in Star Trek, while Blade Runner and Black Mirror demonstrated gesture interfaces. Our sci-fi-inspired imaginations have envisioned touchless interfaces for years, and as technology advances, we eagerly anticipate which other futuristic concepts will become reality. What exactly? To be specific, let's examine interaction examples and use cases that arise. For instance, smart TV devices with cameras, such as Facebook Portal or SberBox Top, are designed with voice-first interfaces, thanks to their onboard virtual assistants that create an exceptional assisted experience. However, these devices still come with a traditional remote. The presence of a camera adds another dimension of interaction – gestures. Users can make hand movements in the air without touching the screen or remote, creating a blend of touchscreen and remote control where the system responds to gestures. While I believe that we'll eventually transition to fully touchless interfaces, current computational and recognition technology limitations place us in a transitional period. We're developing multimodal interfaces, allowing users to choose the most convenient method – voice, remote, or gesture – for accomplishing their tasks. How? I will share insights and recommendations derived from my team's work in developing touchless interfaces. These guidelines aim to help others avoid unnecessary mistakes when tackling similar challenges. By using these recommendations as a reference point, you can streamline your process and make more informed decisions. : Identify situations where users may need touchless interactions, such as when their hands are wet or dirty, or when interacting from a distance Understand Usage Contexts : Consider device constraints like recognition methods, hardware load limits, acceptable delays, and operating ranges under various lighting conditions. Understand Technical Limitations : Create a list of user-friendly gestures based on usage contexts and technical limitations. Determine intuitive gestures for specific tasks, like adjusting the volume or pausing a video. Formulate Gesture Basket Refine the gesture recognition system through extensive testing under different conditions and by gathering user feedback to enhance accuracy and user experience. Conduct Iterative Testing: When designing a gesture recognition system, select movements that are easy to execute, culturally appropriate, consistent, and unique. The gestures should be logically connected and adaptable for future additions. Сriterias for our gesture basket: Ease of execution: Gestures should be simple to perform with one hand. Cultural appropriateness: Gestures should not be offensive or have negative cultural connotations. Consistency: Gestures should be recognizable and consistent across users. Uniqueness: Gestures should not resemble common, accidental movements made in daily life. Technical and data collection sides One of the toughest challenges we faced in creating our gesture recognition system involved working with data from both aspects: how to collect it and how to annotate it. When collecting datasets from paid respondents, they tended to perform the movements more accurately and mechanically, as they followed specific instructions. This created a significant gap between the dataset domains and real-world scenarios. To address this issue, we collaborated with actors who could take their time to get into character and exhibit more natural behaviour, allowing us to gather a more diverse and representative set of data. However, in real life, people behave quite differently – they may be lounging on a couch or in bed, resulting in more relaxed and imprecise movements. But that wasn't the end of our problems! After collecting the data, we had to correctly label each movement, which was a daunting task in itself, as it was often difficult to determine where the movement began and ended. We faced issues like defining consistent movements, accounting for user backgrounds and lighting conditions, and balancing movement complexity with model retraining. Iterative testing helped us refine our system, collecting data from different angles and lighting conditions. The key aspect of our work was the iterative beta testing that our team started piloting in the early stages when the recognition network was not yet perfect. We conducted a closed beta with respondents, using a false positive detection system. When the network recognized a movement, it saved that frame on the device, and only the device owner had access to these frames. This allowed us to quickly receive feedback on unique real-life cases where we performed poorly. Immediately after receiving feedback, we collected new data on a larger scale to cover that particular case. For example, at the very beginning, the network recognized holding a cup in one's hand as a 'like' gesture, and we collected data from people holding cups to retrain the network. Designing a gesture recognition system is no easy task, and we encountered several unexpected challenges along the way: Defining a consistent set of movements proved difficult, as people executed even the most basic gestures in different ways. The diversity of users' backgrounds and lighting conditions made it challenging for our system to differentiate between movements and background elements. We faced a difficult decision: should we make the movements more complex for users or retrain the model to account for movement diversity? Iterative testing was crucial to refining our gesture recognition system, and we had to collect data from different angles and lighting conditions to ensure its effectiveness. Key takeaways : When designing gesture recognition interfaces, it's important to consider the context in which users will be interacting with the device. For example, users might be in different physical positions, have different levels of lighting, or be wearing different types of clothing. To account for these factors, we need to design interfaces that are adaptable and responsive to these varying contexts. Context is key : One of the biggest challenges in developing gesture recognition systems is the high degree of variability in how users perform gestures. This means that it's essential to test and refine our systems iteratively, both in controlled settings and in real-world scenarios. By continuously gathering feedback from users and refining our algorithms, we can improve the accuracy and effectiveness of our gesture recognition systems over time. Iterative testing is essential In order to train machine learning algorithms to recognize gestures, we need large amounts of annotated data that accurately represent the range of movements that users might make. However, annotating this data can be a time-consuming and labour-intensive process. To address this challenge, we can use automated annotation tools, or leverage crowdsourcing platforms to enlist the help of a large pool of annotators. : Data annotation is crucial : Finally, in order to create truly successful gesture recognition products, we need to continually gather feedback from our users. This can involve conducting user surveys, analyzing usage data, or even directly observing users as they interact with our products. By listening to our users and incorporating their feedback into our product design and development process, we can create gesture recognition interfaces that are intuitive, effective, and truly responsive to their needs. User feedback is essential for success

Wave Hello to the Future: Designing Intuitive Gesture Recognition Systems for Smart Devices

How to Use Intuition to Succeed in Business As entrepreneurs, there are many different skills we have to constantly be developing. But what if there’s one that you can’t touch, can’t see, can’t really measure, but it turns out to be one of the most valuable tools in your arsenal? We’re talking about intuition. Intuition can be hard to define, but it’s that general feeling you get that something is right or wrong without necessarily being able to explain why. It’s your gut feeling, that sense of ‘knowing’. We’ve all felt it before, when we suddenly get an inkling about what’s about to happen, or about the outcome of a decision that was just made, without necessarily knowing why we feel that way. It’s true that intuition isn’t always reliable, and it can be wrong sometimes. But it’s also true that it can be incredibly helpful in making decisions and succeeding in business. In my , CEO of Crunchbase, we dove deep into his history of entrepreneurship. recent conversation with Jager McConnell We also talked about how he has used his intuition to navigate his career and the value of developing this skill. This inspired me to do a deep dive into this topic, as intuition has played a crucial role in my career as well. I believe it’s incredibly valuable to learn what this skill is, how to develop it, and how to use it properly. So, buckle up, today’s going to be a fun ride! What Is Intuition? First things first, let’s try to define intuition. Intuition is the ability to understand something instinctively, without conscious thought. It’s the immediate understanding of something without any logical explanation. It is often seen as a sixth sense, and it can be a powerful tool for entrepreneurs. For example, you might walk into a room and get a feeling that something is wrong without any tangible reason. Or you might face a problem, and immediately know the solution without thinking it through logically. That’s intuition. And it can be a superpower in business. In the business world, intuition is often overlooked, but in reality, it is a vital skill to have. In our fast-paced world, intuition can help us make quick decisions, identify opportunities, and anticipate problems. Why Intuition Is Important for Entrepreneurs As entrepreneurs, we are often pushed to the edge of our comfort zone and have to make judgments quickly and confidently. We don’t have the luxury of spending weeks analyzing all of the information. We need to be on our toes, ready to tackle any problem that arises, without delay. That’s where intuition comes in. Intuition can help us get in touch with , and it can help us gain a new perspective. Of course, intuition is not a substitute for research and analysis. another source of knowledge within us But it can be a valuable tool to have in your arsenal. There are four key reasons why intuition is such a valuable skill for entrepreneurs to develop. Making Decisions If your business is anything like mine, it’s likely you often have to make decisions with incomplete information and under extreme time pressure. This is a skill that is difficult to develop, as there’s no textbook you can learn it from. It relies on your accumulation of knowledge and internal compass to guide you. Intuition can be a valuable asset in these situations, as it allows you to rely on your gut feelings and experience to make decisions. For example, if you have a strong gut feeling that a particular business opportunity is a good one, you might decide to invest in it despite not having all the information you need. Or vice versa. Developing a strong intuition is one of those intangible skill sets that could potentially save you millions of dollars in bad business deals. Or make you millions on good ones. Identifying Opportunities Intuition can also help you to identify opportunities that others may not see. By paying attention to your gut feelings, you may be able to spot trends or patterns in the market that others miss. For example, I know of several entrepreneurs who use this for investing (although I don’t recommend this strategy). When they have a strong gut feeling that a particular technology is about to take off, they might decide to invest in it before others catch on. They let their intuition guide them, and then they pounce on the opportunity. This ability to anticipate changes in the market and act on them quickly can give you a competitive edge. Developing a strong intuition also allows you to think outside the box and consider new possibilities. You might start to get a feeling that a particular product or service is needed in the market but is not yet being offered, and you can use this intuition to develop a new business idea. Creative Solutions Intuition can also help you to create a unique vision for your business. Rather than being constrained by preconceived notions and established ways of thinking, a strong intuition allows you to connect seemingly unrelated concepts and come up with unique solutions to problems. Entrepreneurs who rely on intuition are and unique ideas for their businesses. more likely to come up with creative solutions This is because they are not limited by conventional thinking, and can tap into their unconscious mind to generate new and innovative ideas. Developing this power can help you to differentiate your business from others, and create a unique value proposition for your customers. When you can tap into your intuition, you are able to identify patterns and trends that others may not see and are better equipped to adapt your business strategies accordingly. Anticipating Problems From managing finances to managing personnel, we are constantly faced with a variety of problems on a daily basis. There is an ever-present pressure to make decisions and anticipate potential problems. While experience and knowledge can help you anticipate some problems, intuition plays an important role in helping you anticipate and address issues before they arise. Having a strong intuition can help identify patterns and trends that may not be obvious with traditional analysis. It can also help you to anticipate potential problems and come up with solutions. For example, you may wake up one morning and feel that strong gut feeling that a particular marketing strategy isn’t working, and you may decide to change course before it’s too late. By developing and trusting your intuition, you can gain a competitive edge in the business world. But it is important to mention that intuition should not always be used in isolation; it should be combined with data, logic, and reason to make well-rounded decisions. How to Develop Your Intuition As an entrepreneur, developing your intuition can be a crucial step in achieving success. Intuition can provide valuable insights and lead to great decision-making, which can be especially important in the fast-paced and ever-changing business world. To develop and trust your intuition, there are a few key practices you must take. I know, I’m always preaching the importance of these, but I always will, as they have had such a profound impact on me. Practice Mindfulness and Meditation. These practices will help you clear your mind and focus on the present moment, which can lead to increased self-awareness and a deeper understanding of your gut feelings. The more you are able to be in the moment with your thoughts and feelings, the more you are able to listen to them when it matters. One of the best ways to access your intuition is to pay attention to your gut instinct. Gather Your Intuition. When you’re faced with a difficult decision or a complex problem, take a few minutes to sit with it and see what comes up for you. You’ll be surprised how often your “gut feeling” is the best advice. You can also use your intuition to gather information from your environment. Look for clues that can help you make the best decision. For example, if you’re considering a business partnership, pay attention to the body language of the people you’re talking to. Do they seem interested and engaged, or are they distracted and disinterested? Your intuition can help you pick up on subtle cues that can inform your decision. Your body often knows more than your mind. If you feel a knot in your stomach or a tingle in your spine, it could be a sign that something is wrong. Pay Attention to Your Body. Pay attention to those physical sensations. Seeking out diverse perspectives can also help you to develop your intuition and trust it more. Seek Out Diverse Perspectives: By exposing yourself to different ways of thinking and problem-solving, you can broaden your understanding and enhance your intuition. The more experience you have, the more you learn to rely on your gut feelings and make quick decisions based on them. Gain More Experience in Your Field: As you gather an accumulation of life and business experience, your intuition will grow sharper and sharper. Trusting Your Intuition In order to make the most of your intuition, you have to trust it. Intuition is a powerful tool, but it can be hard to trust something that’s not backed up with data or research. The best way to start trusting your intuition is to take time to reflect on it. After you’ve made a decision or taken an action, take some time to reflect on the results. Did it work out as you expected? If so, you can start to trust your intuition more. This practice will give you some valuable feedback on whether or not your intuition was guiding you in the right direction. You can begin to dissect what parts of your intuition were right, or whether there were feelings you had that you didn’t act on that then turned out to be true. Trusting your intuition also means being willing to act on it, even when you can’t fully explain why you feel a certain way. By taking action based on your intuition, you will develop more confidence in it and be able to rely on it more in the future. Wrap-Up As you can see, intuition is a wildly underrated skill that you must begin to develop and nurture in your entrepreneurial journey. It can help you make quick decisions, identify opportunities, and anticipate problems. But here’s the thing, intuition is not a substitute for research and analysis, it’s a supplement. So, pay attention to your gut feelings, trust your instincts, and use it to your advantage. But continue to do your research and logically plan for your business as well. Remember, a strong intuition can potentially save you millions in bad business deals, help you find the next big product your company needs, or help you rebrand successfully. Don’t overlook it, develop it. If you want to hear more about this, check out my full conversation with Jager. That’s it for now, thanks for reading. I’ll see you next week with more success stories! If you enjoyed this article, I’d love to hear from you. Reply to this or tweet at me and I’ll do my best to get back to everyone! email @ScottDClary

Talking With Jager McConnell, CEO of Crunchbase - Using Intuition To Succeed In Business

ক্রাঞ্চবেসের সিইও জেগার ম্যাককনেলের সাথে কথা বলা - ব্যবসায় সফল হওয়ার জন্য অন্তর্দৃষ্টি ব্যবহার করা

Im Gespräch mit Jager McConnell, CEO von Crunchbase – Mit Intuition im Geschäft erfolgreich sein

Hablando con Jager McConnell, director ejecutivo de Crunchbase: usar la intuición para tener éxito en los negocios

Parler avec Jager McConnell, PDG de Crunchbase - Utiliser l'intuition pour réussir en affaires

क्रंचबेस के सीईओ जैगर मैककोनेल के साथ बात करना - व्यवसाय में सफल होने के लिए अंतर्ज्ञान का उपयोग करना

Crunchbase の CEO である Jager McConnell との対話 - 直感を使ってビジネスで成功する

Crunchbase의 CEO인 Jager McConnell과의 대화 - 직관을 사용하여 비즈니스 성공하기

Conversando com Jager McConnell, CEO da Crunchbase - Usando a intuição para ter sucesso nos negócios

Crunchbase CEO'su Jager McConnell ile Konuşmak - İş Hayatında Başarılı Olmak İçin Sezgiyi Kullanmak

Nói chuyện với Jager McConnell, CEO của Crunchbase - Sử dụng trực giác để thành công trong kinh doanh

与 Crunchbase 首席执行官 Jager McConnell 对话——用直觉在商业上取得成功

As a leader, it's important to make informed decisions that will have a positive impact on your organization. But how do you know when to rely on data and when to use intuition? The truth is, both data and intuition are valuable tools that can help you make better decisions. Data can provide valuable insights into your business operations, customer behavior, and market trends. Intuition can help you make decisions based on your experience and expertise. Here are a few tips: Look for trends, patterns, and insights that can help you make better decisions. Research by Laura Huang, Professor of Management and Entrepreneurship, shows that , in highly uncertain circumstances. Rely on intuition when you're facing a complex problem with no clear solution, for which further data won’t necessarily impact the decision one way or another. Use data to inform your decisions when you have access to it. gut feeling can be useful Use your experience and expertise to guide your decision-making process. As prof. Huang notes in a Harvard Business Review article: “The role of gut feel is often to inspire a leader to make a call, particularly when the decision is risky… Trusting your gut allows leaders the freedom to move forward.” This will help you make well-informed decisions that align with your organization's goals. Keep track of what you know (i.e. data) vs. what you assume (i.e. gut) so you can revise or adjust your decisions when new data points become available. De-risk your decisions by breaking them down into smaller bets that leave your options open but still allow you to validate the direction. Use a combination of data and intuition when you're making strategic decisions that will have a long-term impact on your organization. Leverage intuition to brainstorm and generate a lot of different ideas for different solutions to a challenge. Once you identify the most relevant alternatives, data can help you test and compare multiple solutions and identify the best one. To make the best decisions, it's essential to strike a balance between data and intuition. A dash of data can provide you with valuable insights, while a spoonful of intuition can help you make decisions that align with your organization's values and culture, as well as rely on your experience and expertise. Also published here.

A Dash of Data, a Spoonful of Intuition

Recently I was given a Myo armband, and this article aims to describe how such a device could be exploited to control a robotic manipulator intuitively. That is, we will move our arm as if it was the actual hand of the robot. The implementation relies on the Robot Operating System ( ), in particular, on the API. I leave you the reference to the repository, the repository, and the . ROS rospy main project gesture recognition dataset dataset itself What’s in a Myo The Myo armband is an electronic bracelet produced by Thalmic Labs, which allows the owner to interact remotely with other devices. Have a look at the to get an idea of what you can do with it. Thalmic Labs repository The device is equipped with an accelerometer, a gyroscope, and a magnetometer packed together inside a chip called . These sensors provide valuable data which can be exploited to understand where the device is in space and even how it’s oriented. MPU-9150 But the Myo it’s likely most known for its 8-channels sensor, giving it the ability to sense the muscular activity of the arm of who’s wearing it. One can then process this kind of data to recognize the current gesture of the hand of the user. Machine learning is a popular choice here. For instance, people built applications allowing you to switch channels on your smart TV by waiving in and out your hand or turning on your smart light by snapping your fingers. EMG Let’s get a robotic hand Our goal is to control a robot using the Myo. So it seems like we need a robot, which is quite expensive nowadays :( Fortunately, there exist simulators, and indeed we’re gonna control a robot in a simulated environment using . Which robot? We opted for a 7 degrees of freedom (DoFs) manipulator by . That’s a very popular choice in academic research, but it also finds space in some industrial environments. We stick with the Franka, especially because being a redundant manipulator, it’s capable of very dexterous moves, making it suitable for our intuitive control. Gazebo Franka Emika Now that we somehow have a robot, we need a way to talk with it. For this purpose, there exists the Robot Operating System (or just ROS), probably the de facto standard when it comes to open-source robotics projects. In particular, my code is based on rospy, which, as you might guess, is a python package allowing you to write code to interact with ROS. There are tons of libraries to work with the Franka with ROS, and you can find good working ones for the Myo armband as well. In particular, I’ve used to “talk” with the bracelet, and to interact with the Franka. Let’s thank the authors, they lifted a lot of work from us. ros_myo libfranka How we control the robot The manipulator moves accordingly to signals computed by a . I don’t want to dive deep into robotics theory, but the choice was primarily driven by the fact that it doesn’t require the desired task space position as input. The user should drive the robot to his target pose by moving his own arm, . resolved-velocity controller intuitively This controller will impose the orientation of the Myo (i. e. of our hand) to the manipulator, whereas hand and , when detected, trigger a linear velocity of +0.1m/s and -0.1m/s respectively over the approach direction of the end-effector (i. e. the Franka gripper). Naturally, whenever the hand is steady in a neutral pose, the linear velocity imposed on the end-effector will be zero. extension flexion In the above picture, you can see a block marked as . That’s the inner control loop of the manipulator, which directly talks with the robot joints and makes them move. It is provided by the Franka API and it’s compliant with the . It is marked as because that’s the symbol of an integrator in the frequency domain. Indeed, here we are assuming that the inner control loop works very well and it behaves like a perfect integrator (joint-space velocity, the , is integrated to provide the desired joint-space pose, ). Depending on the context, this assumption may be too optimistic. 1/s ROS control interface 1/s dotted q q Extract the orientation from the Myo There are several methods to describe the orientation of a body in space with respect to a given reference frame. For instance, a popular choice is to rely on the . This way, we have a triplet of numbers, each one representing the angle w.r.t. x, y and z axes respectively. In robotics and ROS in particular, a common practice is to exploit to represent an orientation. roll, pitch and yaw angles unit quaternions Given that the Myo has a built-in magnetometer, it can already give us its current orientation. Unfortunately, I found a big issue with that. Namely, it’s not clear to which reference frame the provided orientation refers to. Even worst, while recording multiple orientations from the same Myo pose, the results were sometimes different. It could be an issue with the firmware or the magnetometer itself because this device can greatly suffer from the nearby presence of other electrical devices. The point is that I couldn’t trust the orientation of the Myo this way. To address the problem of reliable orientation, we exploit the ROS package. It provides an out-of-the-box implementation of , an algorithm that fuses angular velocities (from the gyroscope) and linear accelerations (from the accelerometer) to compute an orientation wrt the Earth’s magnetic field. Briefly, the filter first computes an orientation integrating gyroscope readings. This alone would quickly accumulate errors over time, making it useless. To this end, the filter also provides a second orientation, computed with a gradient descent algorithm applied to accelerometer data. This second orientation serves to correct the long-term error of the first one. imu_filter_madgwick Madgwick’s filter Once we have a reliable source of orientation, we need to describe it wrt to the base link of our manipulator. Here, we assume that we know the orientation of the base link frame wrt to the Earth's global frame. Since we are working in a simulation, we are free to arbitrarily choose that the global frame in the virtual world coincides with the one of the Earth. In particular, given that our gazebo world also has the base link frame aligned with the world frame, we can directly say that Madgwick’s filter gives us the orientation of the Myo related to the base link of the Franka. Exactly what we need. Preprocessing for Madgwick’s filter Because we aim to control the manipulator as if it was our arm, it’s just easy to think that our hand acts as the gripper. It follows that our forearm should be aligned with the approach direction (z-axis) of the Franka end-effector. The accelerometer and gyroscope of the Myo are oriented differently, as they have the z-axis orthogonal to our forearm. So we need a change of frame to account for that, as depicted below. Then, we have to realize that both the gyroscope and the accelerometer are very noisy sensors. To this end, I find it beneficial to filter their data using the (SHOE) detector, before feeding them to Madgwick’s algorithm. This is a simple formula that computes a score out of angular velocity and linear acceleration readings, weighted by their variances. We say that the stance hypothesis is (i. e. the device is stationary) or (i. e. the device is moving) whenever the score is or a given threshold. The idea is to feed data to Madgwick’s filter only if we detect that the device (our arm) is moving. This prevents the algorithm to keep computing useless micro-changes in orientation due to subconscious moves or vibrations of our arm. Stance Hypothesis Optimal Estimation¹ confirmed rejected below above In the above video, we are just logging to screen the angular velocity and acceleration of the Myo. It’s not important to read the values. What matters is that, as you can see, on the left side of the screen data is changing continuously, even when my arm is not moving significantly. On the contrary, the right side of the screen shows the same data filtered with the SHOE detector. This time values get updated much more coherently with the true motions of my arm. Gesture recognition Quick recap: we want to exploit the EMG sensor of the Myo for gesture recognition. We need 2 different gestures to impose a or velocities along the z-axis of the end-effector. Additionally, we want to make the gripper close when we squeeze our hand into a fist gesture. Finally, we need to recognize whenever our hand is in a neutral pose in order not to send any of the previous commands to the robot. +0.1m/s -0.1m/s This project uses the scikit-learn implementation of a (SVM) trained for gesture recognition. I tried several other machine learning classifiers, but SVM turned out to be the best. Furthermore, it involves just dot-products, a fast operation for nowadays machines to carry on. The speed of the algorithm is crucial to us since we’re gonna work in a real-time setup. Support Vector Machine Of course, any machine learning algorithm requires a proper dataset to train on. There are some even good-quality datasets available online for gesture recognition using the Myo armband. Unfortunately, most of them assume EMG readings in the form of an 8-bit unsigned integer (aka a byte). My device provides data as 16-bit signed integers. It may be due to a different firmware version or the ROS package used to interact with the bracelet. Anyway, I didn’t achieve satisfactory results training on those datasets. I ended up collecting my own small dataset by asking a bunch of friends and relatives to perform hand gestures while wearing the Myo. The protocol was very simple: for each candidate, I recorded 2 sessions for each of the three target gestures ( , and ). A single session lasted a minute, during which the candidate was asked to alternate between the neutral and the target gesture every 5 seconds. In the meanwhile, every EMG reading coming from the Myo was recorded. The final dataset was then chunked into samples in a sliding window fashion. A single sample comprises 30 consecutive 8-channels EMG readings. Two subsequent samples have an overlap of 10 readings. This choice should enable approximately 2 gesture predictions per second. fist flexion extension Once I had the data, knowing the extent (60 seconds) of a session, the target gesture performed in every session, the span of every gesture (5 seconds), and the frequency of the EMG sensor (about 50 hertz), I could annotate every sample with the proper gesture label. This approach is theoretically optimal, but in reality, it can’t be. The EMG is not always working at 50 Hz, and a candidate is a human being rather than a perfect machine, so sometimes a gesture lasts more than 5 seconds, sometimes less. So you are going to accumulate errors while annotating samples this way within a recording session. It turned out that a better solution was to annotate data by using a clustering algorithm, in particular, I chose the popular . While SVM is a machine learning classifier, this one belongs to the family of learning algorithms. This means that it can infer knowledge from data without a supervised signal (i.e. labels). So I simply run the K-means on the whole dataset, partitioning it into 4 different clusters. The label of a cluster was set as a label for every one of its samples. I then trained the SVM on this dataset. K-means supervised unsupervised The dataset has been uploaded and is free to access on Kaggle. There, you will find a quick notebook on which you can test the performance of an SVM on the data annotated with both the labels “by hand” and the labels provided by the K-means. The test runs a 5-fold . As you can see, using hand-made labels, the SVM performs quite well. We are in the nearby of 0.9 both in and (it’s better to consider the latter since the dataset is highly unbalanced towards the neutral gesture class). But if we train the classifier on the very same dataset but using the class labels provided by the K-means, we raise it to nearly 0.98 in both metrics. This gap shouldn’t surprise us. The clustering indeed annotates data based on patterns naturally found in the data itself, which is the way you’re supposed to exploit machine learning. On the other hand, the labels put by me only rely on time, but in practice we know that’s gonna make errors, so a classifier would learn from bad data. cross-validation accuracy F1-score Now I have to stress one thing: what I’ve done here, that is . Machine learning would be a lot easier otherwise. For instance, you know that sample belongs to cluster , but actually, you don’t know which class is . Fortunately, in this project, given that we have only 4 possible labels and the classifier works pretty well, it was straightforward to experimentally find the associations from the cluster labels to the related gestures. But in general, this cannot be done. Think if you had 1000 different clusters and you have to match 1000 different class labels… using a clustering algorithm to annotate data for supervised learning, cannot be done most time The point is that cluster labels provided by the K-means have no semantical meaning at all! x “3" “3” The following is an attempt to visualize the entire dataset. The 2 reddish circles represent class labels and you may see that there are indeed differences between those provided by me (inner) and those given by the K-means (outer). The bluish circle is EMG readings: the more it gets close to yellow, the more powerful it was the recorded muscular activity. Here we find only a subset of the dataset. This time you may better recognize that the bluish circle is chunked into many circular sectors, everyone being a single sample. Notice also the orthogonal subdivision into 8 different parts. That’s because the Myo EMG has 8 different channels. As we previously said, a single sample entails 30 consecutive EMG readings. If you are familiar with , we can state that a sample is a with [30, 8]. We can now flatten it into a one-dimensional array of shape [150] and feed it to the SVM. However, with just a little bit of preprocessing, we will reach better performances. Indeed, I didn’t train the classifier on this raw data. Every sample undergoes a reduction operation over the time axis (the first dimension of the array). Specifically, I take the of the time series represented by a sample. We’re thus switching from a [30, 8] to just an [8] array. This is beneficial because we have less data to process and it allows us to better generalize, making the classifier more robust, especially wrt . Furthermore, we also between 0 and 1 the resulting feature vector. numpy numpy array shape median outliers normalize About kinematics singularities There’s one last point to cover in our project. Recall that the Franka is a manipulator. It means that it can reach the desired pose of the end-effector in potentially more than one configuration. That’s a desirable property in general, but it doesn’t come for free. There are indeed some joint configurations that make further motion very unease for our manipulator. Those are called . redundant kinematics singularities The resolved-velocity control implemented in this project comprises two different operative modes. The one is the simplest and basically just ignores singularities. With this setup, the Franka probably will soon get stuck into one of these bad configurations and we will need a little patience to get it out. first The mode implements the revised resolved-velocity control loop presented in . Shortly, it aims to maximize the (MoM). You can think about it as a score of how much easy the manipulator can move away from its current joint configuration. Thus, if we maximize this quantity during our control loop, we keep the Franka far from its singularities. The below images may help you visualize what a kinematics singularity is. second A purely-reactive manipulability-maximising motion controller² measurement of manipulability To drive the robot in those poses, I made the very same motion with my arm, but it resulted in two different behaviors. On the left, we find the Franka in one of its singular configurations, with the 6th joint vertically aligned with the base link. On the right, the singularity has been avoided thanks to the advanced control loop and the robot is ready to move to any other configuration. Conclusions and limitations Of course, this project is far from perfection and by no means ready to be used in any real-world environment. It is supposed to be a proof of concept for further developments, although I confess that even now it’s quite funny to use the Myo to control the Franka :) One of the most critical points regards the difficulty to execute all the computations required to keep our code up and running. To bring in some context, the nominal frequency of the Franka manipulator is 1 Khz. By downgrading and disabling some of the Gazebo physical engine features, I’m able to run the program at around 100–200 Hz. That’s while using the simple control loop, without singularity avoidance. Introducing the advanced resolved-velocity control, the frequency drops to the nearby of 5 Hz. That’s 200 times less than the frequency the robot is supposed to work at. The algorithm involves solving every time a quadratic programming problem and my machine is not able to handle it in real-time. I made some attempts to improve this, really with no significant gain. One idea could be to employ the simple controller by default, resorting to the advanced one as we get close to a singularity. More broadly, I think switching from python to C++ could make a huge difference. Naturally, if you have a powerful machine you may already get an enjoyable experience without even noticing these problems. Anyway, if you come so far, let me thank you. I’m aware that I’ve just reported a brief description of the project, without discussing in detail neither the code or the theory behind it. The purpose was just to elicit your curiosity and give you an idea of what can be achieved with some robotics and data science background. If you have time, will, and a Myo armband at hand, feel free to extend this project! References [1] I. Skog, P. Handel, J. -O. Nilsson and J. Rantakokko, “ ,” in IEEE Transactions on Biomedical Engineering, vol. 57, no. 11, pp. 2657–2666, Nov. 2010, doi: 10.1109/TBME.2010.2060723. Zero-Velocity Detection — An Algorithm Evaluation [2] Haviland, J. and Corke, P., 2020. . . A purely-reactive manipulability-maximising motion controller arXiv preprint arXiv:2002.11901 This post was originally published . here

Creating an Intuitive Robotic Manipulator Control With a Myo Armband

LIME stands for .  It is a technique used by researchers in machine learning to explain the trust of the model and to measure the confidence of the algorithm. “Local Interpretable Model-Agnostic Explanations” Now, you will have an obvious doubt that we have many evolution metrics to measure the performance of the models irrespective of being a supervised or unsupervised category. So, why do we need LIME? The problem with the mathematical evaluation metrics is that they cannot explain the reasons for the prediction in a simple understandable way.  LIME is not an opposite technique to that of traditional evaluation metrics but it is just a set of guidelines to increase the trustiness of the model in addition to that of statistical metrics with some additional interpretable information. The intuition behind the “LIME” concept in AI & ML Let us consider that we have built an image classification algorithm to predict whether a person is male or female if we provide the photo as input- The model will predict Image 1 as male and Image 2 as female based on some evaluation metrics like accuracy, precision, recall or f1 score. Now, what is the reason behind these evaluation metrics which tends the model to predict the input as male or female? Seems confusing? Let us have another example, John is suffering from a fever. He consulted a doctor to check whether his fever is due to coronavirus or not. He explained the symptoms currently he is facing like sneezing, body pain, headache etc. The doctor conducted some medical tests and said to John that it was not a corona fever but just a usual one in just a single word (Yes/No). Now, do you think John will be satisfied with the doctor’s inference just because of the reason that the doctor is an expert in the medical field? No, he will be somewhat confused and not confident because the response from the doctor looks like a naïve- binary Yes/No answer. The above case is similar to any machine learning model. Most of the models deployed in the production environment are black boxes i.e. the people who consume it as naïve users don’t know about the mathematical logic coded inside the model for doing the prediction. Implementing the predictions from such a model may result in losses even if the evaluation metrics are confidently positioned. Let us consider that doctor says to John that - “John, I have conducted an antigen test to diagnose whether SARS-CoV-2 is present in your body or not. Luckily, it is not present. Additionally, I have also done serology tests to detect antibodies that indicate whether you have been infected with the virus or not. The result is negative. You are not affected by the coronavirus…Stay happy!!!” Now, John will be very confident about the inference provided by the doctor and his trust in the doctor has increased. He will be no more confused about his health condition. He was a little bit confused when the doctor gave him a naïve-binary response. This is the actual difference between LIME and traditional acceptance of prediction. Instead of taking the prediction which is recommended by the model, LIME enforces to interpret the explanation behind the reasoning ability of the model. In our Image classification example, the user will be much more satisfied if the model says the mathematical reason behind the prediction instead of naively classifying the image as male or female such as- The person in Image 1 has a beard and moustache. He seems to have a hard skin tone and short hair. Hence the person should be a male (Accuracy – 90% and precision – 85%). The person in Image 2 has no beard or moustache. She seems to have a soft skin tone and long hair. Hence the person should be a female (Accuracy – 90% and precision – 85%). Now, the prediction seems more easily interpretable and acceptable as the model explains the reason behind the predictive inference. This is nothing but LIME. Let us have a look at the intuition and interpretation behind the LIME. LIME is the abbreviated form of – Local Interpretable Model-Agnostic Explanation The “ is nothing but a clarification and descriptive statement that can briefly convey the clear-cut and unbiased reasons behind the predictive inference. explanation To become model “ , LIME usually don’t go to the depth of the approaches used inside the algorithm but to find out which part of the interpretable input is contributing to the prediction, we perturb the input around its neighbourhood and see how the model's predictions behave. agnostic” For example, Let us consider that we are building a model to analyze the sentiment about a film – whether it is good or bad based on the reviews floating on the internet.  Consider that a reviewer writes that   “I loved the movie”. Now, LIME recommends perturbing the sentence like “I loved”, ‘I movie”, “loved the movie”, “I movie”, etc. as inputs to the model. Now, after some iterations, the model can intuitively understand that the word “loved” is the most important word to measure the sentiment of the reviewer. Rest of the words like “I”, “movie”, “the” etc. are not so important in an algorithmic perspective. is nothing but one of the characteristics of LIME that makes the user understandable the reasons and explanations in a very lucid and easy manner. There should be no high-level mathematics or scientific concepts so that the naïve user finds it difficult to understand the reason behind the predictions of the model. “Interpretable” For example, In word embedding, we can see that the model explains the inferences in complex charts that is not at all visually appealing and easy to understand. LIME can interpret those kinds of information in simple forms (like sentences and words in the English language). The output of LIME is a list of explanations, reflecting the contribution of each feature to the prediction of a data point. This provides local interpretability, and it also allows us to determine which feature changes will have the most impact on the prediction. Let us see this example from a visual perspective- Imagine that we are working on a binary classification problem. Consider that the blue circle denotes the decision boundary (prediction area) of this model in 2 feature dimensions – X and Y. The green diamond is a particular data point and the red circles are perturbed data points (the ones we generated/tweaked).  The size of the circles indicates the proximity of the original data point. The data points can be assigned certain priority weights according to their proximity. Now, we have to find how the model’s prediction changes for perturbed inputs. For that, we should train a surrogate model on the dataset by using the perturbed data points as the target. This explanation and findings may not be acceptable globally but acceptable for the local region around the green diamond. Here is the standard step-by-step procedure- refers to local trust or local belief - i.e., we want the explanation to reflect the behaviour of the classifier "around" the instance being predicted. The fidelity measure (how well the interpretable model approximates the black box predictions) gives us an idea of how reliable the interpretable model is in explaining the black box predictions in the neighbourhood of the data point of interest. “Local” This is an example mentioned in the research paper. Here, the classifier predicts Electric Guitar even though the image contains an acoustic guitar. The explanation image shows the reason for its prediction. (This library is not still able to explain the reasons in a sentence format and currently, the reasons are being explained in an image format for image classification problems. There are some libraries which can explain the reasons in sentence format in Python and R). This is another example from the research paper where the model is built to predict whether a paragraph proposes Christianity or Atheism based on the words used in the paragraph. First of all, we need to understand the difference between model-specific and model-agnostic interpretation tools to understand how this is implemented in LIME- Model-specific interpretation tools Model-specific interpretation tools are specific to a single model or group of models.  These tools depend heavily on the working and capabilities of a specific model. For example, Linear regression depends upon various parameters like slope, coefficient etc to make predictions. The decision tree uses a technique called entropy to find the best split of nodes to increase the predictive ability of the model. Neural networks have millions of parameters for connecting the neurons and making a prediction. All these above parameters and techniques are specific to their model architecture. These rules, guidelines or parameters of the model cannot be applied globally for making interpretations. These are called model-specific interpretations. Model-agnostic interpretation tools Model-agnostic tools can be used on any machine learning model no matter how complicated. These agnostic methods usually work by analysing feature input and output pairs. Model-specific interpretations will have more accuracy since their model architecture will be complex and well-built like linear regression, random forest, neural network etc., but model-agnostic interpretation will have low accuracy since the interpretations are done only by analysing the patterns of input and output. So how can we build model-agnostic interpretable models that don’t compromise on accuracy? One of the solutions to this problem is the global surrogate method. This method suggests using an additional model called an which can explain the predictions done by the black box model. “interpretable model” or “surrogate model” For example, If we want to build an ML model for predicting whether a student will pass the exam or not then we will have 2 models- Black box model (To predict whether the student will pass the exam or not) Surrogate model (To interpret the approximate logic of the black box used for the decision making). The black box model will be a complex algorithm and the surrogate model will be a simple algorithm. For example, We need to build a machine learning model for predicting whether a student will pass the exam or not. We have 100’s independent variables. Hence, we built a neural network which is highly complex and not easily interpretable. Now, after getting the results, how can we interpret what kind of features and characteristics of predictor variables resulted in the pass/fail of students? For that, we can build a simple decision tree which is very less complicated than a neural network and easily interpretable. Using this decision tree, it will be easy to find which all are factors mainly influenced the pass/fail of the students in the examination like “Marks in the assignments”, “Marks in the internal exams”, “Attendance percentage” etc. This is done by training a decision tree on the predictions of the black-box model. Once it provides good enough accuracy, we can use it to explain the neural network. Here, the neural network is a black box model and the decision tree is a surrogate model. Hence, we can conclude the black-box model by interpreting the surrogate model. The following image shows the steps to build a surrogate model- The above procedure is based on an assumption that – “When the complexity of the model increases, accuracy increases and interpretability decreases. Hence, it makes sense to make predictions using a complex machine learning algorithm for getting highly accurate predictions and use the same inputs and predictions from the complex machine learning model to train a simple machine learning model so that the interpretability of the predictions can be easily inferred through the low complicated architecture of simple model”. However, a disadvantage of the global surrogate method is that it can explain the global features and rules that can interpret the predictions of a black-box model. However, it cannot interpret how a single prediction was made for a given observation. For example, If we are predicting the exam results (pass/fail) of thousands of students in a university then a global surrogate method can show which are the factors that globally influence the marks of students like “Marks in internal exams”, “Marks in projects”, “Attendance percentage” etc. But it cannot interpret the reason why “John (Roll number -198) belonging to MBA final year” was not able to clear the exam. LIME’s model-agnostic approach becomes a key attraction in these kinds of scenarios where it can interpret these kinds of scenarios. Advantages of LIME We can change the black box model and still use the same local surrogate model because it does not have any heavy impact on interpretability. For example, In our above example , let us consider that the blue-coloured decision boundary (binary classification problem) is created by Kernel SVM and we use a decision tree as the local surrogate model. We can change the black box model from Kernel SVM to Random forest and still use the same decision tree as a surrogate model for local interpretation. Sometimes, the local surrogate model will give much more information about the feature importance than the black box model which may be a helpful global inference. (Fig 1 and Fig 2) LIME can be implemented in any kind of data like raw texts, images, data frames etc. Inbuilt packages are available in R and Python. Disadvantages of LIME Simple perturbations might not be enough and complex perturbations may lead to bias in the model. Hence, it requires some soft skill algorithms or techniques to introduce perturbation to the model. Although, LIME is model agnostic, the type of modifications that need to be performed on the data are use-case specific. For example, a model that predicts sepia-toned images to be retro cannot be explained by the presence or absence of superpixels (Iris dataset) Non-linearity at local regions in specific datasets makes the model difficult to interpret. Such models are not easily adaptable with LIME. Conclusion The of the machine learning outputs from complex black box models is always a concern for researchers in the analytics field. To an extent, LIME can be considered a solution to this problem for many use cases. There are many more libraries that are getting developed as a part of the ongoing research and development in the analytics field and LIME is a successful initial step in this learning journey. I hope, you got a brief understanding of the fundamental principles of LIME in this article. For more in-depth information, you can refer to the official research paper which I have mentioned in the reference section. “Explainability” References Marco Tulio Ribeiro, Sameer Singh, Carlos Guestrin, “Why Should I Trust You?” - Explaining the Predictions of Any Classifier Vilone, Giulia; Longo, Luca (2021). . Information Fusion. December 2021 - Volume 76: 89–106. : . "Notions of explainability and evaluation approaches for explainable artificial intelligence" doi 10.1016/j.inffus.2021.05.009 . DARPA. DARPA. Retrieved 17 July 2017. "Explainable Artificial Intelligence (XAI)" Castelvecchi, Davide (2016-10-06). . Nature. (7623): 20–23. "Can we open the black box of AI?" 538

The Intuition Behind the “LIME” Concept in AI & ML

In the presentation (video below), Robyn Linn, Advisor at Vetter, examines the components that make up the Vetter ecosystem and talks about how it can benefit Web3 developers and investors. Before we dive into Robyn’s presentation, some key concepts are highlighted. https://youtu.be/7Wn290ODdEY?embedable=true What is Vetter? Vetter AI is a research platform built on the . It leverages the power of crowdsourcing and gamification to help crypto investors research and identify legitimate projects to invest in. The Vetter ecosystem offers a range of decentralized applications, also known as . These dApps aim to help developers launch and manage their projects and help investors make more educated decisions. BNB Smart Chain (BSC) dApps Vetter was a Platinum Sponsor of the inaugural Blockchain Oracle Summit in Berlin. The summit was the first to focus solely on oracles’ importance and design. Vetter’s use of the intuitive oracle feature and gamification of due diligence makes technical information easier to come by, making them a relevant partner at BOS22. CrowdX & Vetter’s Intuitive Oracle Vetter’s first dApp, CrowdX, was released in 2021. It features the CrowdX Calendar, which consists of community members with distinct roles. Namely, “Scouts,” “Vetters,” and “Voters.” Scouts are responsible for finding and posting projects in presale, fair launch, or Initial DEX Offering (IDO) stages on the calendar. Vetters then help to provide additional research data on each project, after which, Voters monitor the success of these projects over a specific timeframe. The CrowdX Calendar also features Vetter’s Intuitive Oracle. Intuitive Oracle allows users to sort projects posted by Scouts based on their investment objectives. For example, users who prefer investing in projects that double in value within the first 24 hours of launching can choose only to view projects that hit these specific criteria. On the other hand, those that prefer projects that have more long-term benefits would be able to sort out the projects that show greater potential over a longer timeframe. In addition, Vetter makes use of gamification to incentivize community members to continue contributing accurate data to the Vetter ecosystem. Scouts are color-coded based on their ranks. The highest-ranked purple Scouts have the best record of posting successful projects. Robyn cited the example from their audit earlier this year, were more than 90% of projects posted by Vetter’s purple Scouts at least doubled in price within 24 hours of being launched. Skylabs In her presentation, Robyn introduced Skylabs, which serves as an incubator for developers to launch their projects and share them with the CrowdX community. With Skylabs incubator, projects benefit from a team with blockchain and business expertise. Developers who might not have prior experience handling business operations could tap into an array of in-house resources to raise funds, allocate budgets and successfully market their projects. Robyn also highlights how Skylabs has helped new projects in hitting “hardcap.” The term “hardcap” refers to the highest amount of funds a cryptocurrency project aims to raise during its Initial Coin Offering (ICO). Robyn reiterated the importance of building a holistic team by providing the example of , which managed to raise $10 million in its pre-sale launch. According to Robyn, The Gaming Project was well prepared for launching their project by building a team with the right experience. The Gaming Project In a space where thousands of new projects are launched each year, it can be tricky for new investors to differentiate the good projects from the “rugpulls.” With CrowdX and Skylabs, the Vetter ecosystem aims to help rookie and veteran investors make more informed decisions and get into projects with strong fundamentals at the earliest stages. Originally published here. Learn more about Vetter: | | | Website Whitepaper Twitter Youtube

Intuitive Oracles and Incubators As The Vehicles for Launching Crypto-Projects: Robyn Linn Explains 

More and more companies rely on digital technology such as websites, chatbots, and mobile applications to interact with their customers. The experience that the customer has with these platforms can make or break a company. According to one survey, . The solution to this challenge is an approach known as Intuitive UX Design. 70% of online businesses fail because of bad usability Why is intuitive UX design critical to your success? Having beautiful-looking user interfaces for your programs and mobile apps can help generate emotional responses from your customers encouraging them to interact with your technology. Unfortunately, a great-looking interface alone doesn’t guarantee a positive experience. One of the most critical elements to giving your customer a great experience is by creating an interface that is intuitive. Why is intuitive design so important? Users will quickly become frustrated with confusing or complicated UX designs. If your team receives a lot of questions from customers on how to find or access certain features of your technology, you could have a non-intuitive design. The goal should be to provide a UX that requires little to no training or instruction. Easy to Learn: If using a program or application is simple and intuitive, users will be more likely to spend more time using your program than your competitors. This helps build brand loyalty and increases profitability. Prolonged Use: An intuitive UX should not only provide the user with a seamless experience, it should also allow you to elicit certain behaviors. For example, strategic placement of certain elements can gently and naturally steer them toward profit centers such as CTAs (call to action) or prompts to access enhanced features. Better Control Over User Activity: Related Article: How is Quantitative Data Used in UX Research? Key elements of intuitive UX design Providing an intuitive interface can be quite simple if you use the right approach. Most companies fail at intuitive UX design because they provide information in a way that makes sense to them. Instead, try to get inside the minds of your users and create a design that anticipates their wants and needs. Your users may not be as technical or understand industry lingo as your internal team. By using the same terminology that your customers expect to see, it will be easier for them to find what they are looking for. For example, instead of saying “Technical Documentation”, you can simply say “Product User Manuals”. Use the Right Language: Try to anticipate the order that which users expect information to be presented to them. Placing high-click buttons or information at the top of the interface is better so they don’t have to scroll down looking for what they need. Place Objects in Logical Order: Grouping similar information and features together helps users find what they need quickly. This is especially useful if your interface is complex or requires a lot of details. Minimize Clicks – Design your UX in a way that helps users get to the information they need in the shortest amount of time and requires the fewest number of clicks. Segment Your Interface: Having too many options on the screen can cause cognitive overload. You want your user to spot the information they need or the next step quickly (1-2 seconds at the most). Use the Right Signifiers – Signifiers are graphical elements that help the user identify certain information or features. For example, the cog symbol often represents a settings or options menu while a magnifying glass can indicate a search field. Reduce Options: Using colors wisely can help the user navigate the UX. Green and blue are often associated with “submit” buttons. Red can indicate “cancel” or “go back”. You can also use color or bold fonts to draw the user’s attention to certain features or information. Colors: Your interface may have multiple views or layers. It’s important that your layouts, colors, and object placement are consistent so the user already feels familiar with each new page or view. Related Article: Be Consistent: Increasing the Business Impact of User Research Transform your platform with a leading UX design specialist At , we understand the power of creating user interfaces that are bold, simple to use, and engaging. Our team of seasoned professionals in product development and UX design is here to help you every step of the way. to find out how we can supercharge your customers’ user experience. Lean Discovery Group Book a 30-minute discovery call today Ethan, Founder of the Lean Discovery Group First published here

How to WOW Users With Intuitive UX Design

A nautilus seashell with a perfect spiral is the product of specific DNA that coded for its existence. Visuals in your application behave similarly; their appearance is controlled by a particular set of seemingly distant algorithms that come together in just the right way. In this post, I will discuss how fundamental algorithms like “finding all unique pairs” allow for the construction of clear visuals easily and quickly interpreted by the user. In doing so, my goal is to show how unrelated algorithms may be used in conjunction to produce intuitive markings on a chart. I will discuss the following chart from : PyraMetrik You do not have to be an expert on (bonds) and their associated to get through this short article. Treasuries yield curve Just know a few things, for context: The longer the term of the bond (i.e 1-year vs 5-year), the higher we expect its yield to be, generally. When a bond’s yield rises above a longer-term bond’s yield, this is an in part of the associated yield curve. inversion Inversions may be interpreted in several ways, though this is not the point of this article. Practical objective: Mark all the areas of inversion in grey, rectangular areas (as shown above). Let’s walk through how elementary algorithms that involve generating all unique pairs and finding certain intervals in collections help us achieve this goal. We’ll cover this in Three Steps. Step 1: Find All Intervals Firstly, if we are given two collections representing respective yield values for respective bonds, we must find all intervals where the short-term yield is greater than the longer-term yield. Supposing List 1 is the shorter-term bond and List 2 is the longer-term bond, we have the following diagram: As you can see, the lines in red mark the intervals where List 1 elements are greater than List 2 elements. In code, this could look like this: The above function returns an array of “occurrence objects”, where an occurrence object simply contains the starting and ending indices of a given occurrence. To really grasp this in terms of code, some unit tests for this function may look like this: This is good, but remember, we have bonds in the chart pasted at the beginning. So, this function needs to be performed for all possible, pairs of bonds — because inversions can happen anywhere. several unique This takes us to the second step. Step 2: Generate All Unique Pairs Generating all unique pairs for a collection can be visualized like this: Notice there are only three, pairs. If “red-green” is already considered as a pair, we don’t want to consider “green-red”, as that would be the same pair in the context of what we are doing here, that is, comparing two bonds of different term lengths. unique Translating this into code, we’d get something like this: As you can see, we start the inner loop at this ensures we don’t count “green-red” if “red-green” is already counted. i+1; Moreover, to ensure that no duplicate pairs are included in the output, we make use of a object, and we only push into if we have not yet “seen” that pair. seen pairs “seen” is a JavaScript object that can be thought of as a hash map, where the time to check for the existence of a key is constant (O(1)). Note 1: If the input “arr” is guaranteed to be of unique values, the “seen” object is not needed in this function. Note 2: Great — we now have the key algorithms that allow us to define where exactly to draw these rectangular boxes. As you might imagine, the rest is highly context-dependent. Put differently, how you decide to piece these algorithms together and finally draw the elements on the chart is up to your project’s requirements/libraries/implementations. I’ll show you how I do it, but remember — this is for specifically, where we use for charting. PyraMetrik recharts Step 3: Bringing It All Together The main logic to bring the above algorithms together is as follows: There are functions like and that determine the closer/further dated bonds and find the y-coordinates of the inversions, respectively. But again, this is out of the scope of this article and domain-specific. getCloserAndFurther() findExtremes() Then, inside my line chart, I draw inversions using the (part of recharts): ReferenceArea That’s it. Now we have these distinct, grey boxes that clearly show us areas of yield curve inversion. No need to strain the eyes or waste mental bandwidth trying to spot such patterns “yourself”. Algorithms sound everything but “visual”, and visuals sound everything but “mathematic”. But as shown here, the two are intertwined, and deeply understanding this connection clears the runway for a whole new paradigm of user experiences. Also published here.

Where Visuals And Algorithms Collide: How Unrelated Algorithms Produce Intuitive Markings

Hey Hackers! I’m Aleksey Zaitsevsky and innovation is my hobby. First of all, a huge thank you to the HackerNoon community and staff for nominating me for a 2021 Noonies award! I’ve been nominated in the following category, please do check out this award page and vote: HackerNoon's Regulation-Nation of 2021 As someone in the prototyping industry, I believe that the most exciting technology of the present is Digital Intuition because this will help people with no life experience to see important patterns in examples from the past. Learn more about my thoughts and my journey in the tech industry via the interview below. 1. What do you do and why do you do it? (tell us your story) Searching for technical solutions is my hobby. As a child, I loved to take apart broken scientific equipment that my father brought home from work. This is how the acquisition of knowledge began 40 years ago. 2. Tell us more about the things you create / write / manage / build! Big experience in the design of robots and drones. In software projects, I hire a programmer. Many years ago, I developed technologies for protecting documents in the printing industry. Entrepreneurial experience 15 years. 28 inventions. 3. How did you end up on your current career path? Do you like it? My career goes in waves. Sometimes the profit allows you to develop several projects at once, and sometimes it leaves you in the red. But if you have a habit of creating something, you cannot get rid of it. 4. What tech are you most excited or passionate about right now and why? Digital intuition could be the most significant project in my life. This technology deprives manipulators of competitive advantage. If the tool is available and improved, then it is more rational, to be honest, and original. 5. What tech are you most worried about right now and why? Public opinion manipulation technologies. It feels like the planning horizon is getting shorter. 6. If we gave you 10 million dollars to invest in something today, what would you invest in and why? This means that for the coming years I'll be able to work without obligations to outsiders. The results are likely to be more noticeable. 7. What are you currently learning? I have no long-term learning plans. I learn by solving problems. For example, I take a math textbook if I can not solve the formula. 8. What’s the best advice you’ve ever given someone? Be silent, Zaitsevsky, be silent. 9. What’s the best advice you’ve ever received? Be silent, Zaitsevsky, be silent (to myself). About HackerNoon’s 2021 Noonie Awards The annual Noonie Awards celebrate the best and brightest of the tech industry, bringing together all who are making the Internet and the world of tech what it is today. Please be sure to , nominate, and vote for the people and companies who you think are making the biggest impact on the tech industry today. check out our award categories The 2021 Noonies are sponsored by: , , and . Thank you so much to these sponsors who are helping us celebrate the accomplishments of all our nominees. bybit Dottech Domains Avast

'Digital Intuition Deprives Manipulators of Competitive Advantage': Aleksey Zaitsevsky

Introduction Web services are purpose-built, custom web servers that support the needs of specific applications. Interaction with web-services is facilitated by Application Programming Interfaces (APIs) which are used by the client-side application. A web API is the gateway to any given web service – it listens to incoming client requests and sends the outgoing response to the client. Developers should aim to design web APIs that are easy to maintain, easy to extend, intuitive, and platform independent. The following article suggests some principles to keep in mind when designing RESTful APIs. REST Method Guidelines Use GET to retrieve representation of the resource identified by the request URL. GET requests can be repeatedly executed and should not cause any side effects (e.g. no state changes allowed) on the resource identified by the URL. Use PUT to add a new resource identified by the request URL. The request body must include a representation of a resource that needs to be stored. Use PUT to update or replace an existing resource. If the resource already exists, PUT overwrites the resource state with the representation supplied in the request body. Use POST to request creation of a new resource under a collection in request URL path. The POST response body is HTTP status 204 (empty) or HTTP status 201 (a representation of the resource created). Use DELETE to completely remove the resource identified by that URL. After a successful DELETE request, the resource is no longer available. Subsequent GET requests on the resource must return HTTP status 404. To partially update the state of an existing resource using a URL, use PATCH request and provide the requested updates in the request body. The PATCH response returns a representation of the resource identified by the URL once the requested changes are applied. URL Path Guidelines Forward slash separator (/) is used to represent hierarchical relationship. Use Hyphen (-) to improve readability URLs are case-sensitive except for scheme and host components; prefer lowercase in the path. Use a plural noun for collection names; it should reflect what the collection contains Do not use CRUD function names in URLs e.g. DELETE /deleteUser/1234 is an anti-pattern. The URL path segments are alphanumeric and follow camelCase convention. The URL path segments are intuitive, unambiguous, and succinct. Resource Representations Guidelines Encode the resource representation as application/json except where standard media-type representations are available. A collection within the resource representation is encoded as application/json with an array attribute (elements represent resources of the collection). Date and time fields must be represented as strings and conform to format. RFC-3339 If the service supports pagination for GET requests, the request must also accept limit parameter (maximum number of resources to return) and offset parameter (index in the result set). Resource identifier field called “url” must be present in a server response which points to the absolute and canonical URL of the resource itself. If another REST service is the source of truth for some data, that given data cannot be included in the resource representation (i.e. a given REST service cannot act as a proxy for another RESTful resource). When returning an HTTP error code, it should conform to . RFC-2616 Security Guidelines Use authentication and authorization (e.g. ) to protect resources. OAuth Open Authentication If a REST request contains sensitive and personally identifiable information (PII), it must be authenticated. Authorization scopes required for access must be clearly defined for a REST request. Do not allow content (when not intended) or PII leaks between users and server components. Use HTTPS (secure channel) for authenticated REST requests. A REST request cannot contain authentication, authorization, or PII in the URL. Conclusion Uniform programming language agnostic REST APIs that converge on a common set of guidelines will benefit the developers. They can focus on writing the server code that powers the web services rather than debating on whether to use HTTP POST or HTTP PUT or worrying about the intricacies of RESTful API design.

Simple and Intuitive RESTful APIs 

We can be warned about upcoming events predominantly occurring under similar circumstances. A news forecast is similar to a weather forecast which becomes more accurate when more complex algorithms are used. The is based on the news which many people have become aware of. Governance is a technology that is based on experience. crowd opinion If public opinion is manipulated repeatedly, the algorithm will indicate anomaly and similarity. Suspicion is still not an accusation. However, when patterns are discovered, manipulators are exposed. The project of analytica.today is important both for conspiracists and anti-conspiracists. If the interrelation between conflicts and natural phenomena is discovered, it means someone has an instrument to blackmail. This is interesting, and people's attention is a liquid commodity nowadays. It's only a matter of time to find interesting connections if you have a large amount of data. We are working on identifying statistical links in the sequence of news items, adverts, or other messages. Incoming messages are classified according to several attributes. Selective reclassification is used to account for different trait assessment interpretations. The messages converted into code form an estimator matrix. To detect a pattern in a message sequence on a timescale, it is necessary to compare matrix fragments that follow either before or after messages with the same assessment value according to one or more traits. The correlation dependence with the same data filter on the superimposed time segments is assessed. If the correlation dependence for two or more matrix fragments is high, the data filter becomes narrower. Data on settings and search results are stored in the database as a pattern. The examples discovered are assessed by a person for significance. A new or repeated pattern search starts with settings combining two or more known patterns with similar message codes. The patterns with high significance assessment are more often used to create combined search settings. The data filter is additionally extended using random values. Figuratively speaking, the pattern search criteria evolved by crossing, mutation, and selection. The analysis predictive power is expressed in the assessment of probability with which the new or probable message fits into the previously identified pattern. The past message sequence examples show what typically happens under similar circumstances. There is no intention to replace a human but to expand cognitive abilities using additional memory and collective experience analysis. It's a tool for the reconstruction of factors affecting a person’s opinion. In times of crisis, trust in experts declines. The significance of digital intuition will grow with time. ; Patent LT6685: Translation Original Also published . here

Stories, News & Mentions about INTUIT

Intuit’s Stories on HackerNoon

Intuit's News & Mentions