In the epic poem The Inferno, Dante journeys through the layers of hell—resembling an upside-down nine-layer cake as pictured —guided by the poet Virgil who is the height of human reason and understanding. Virgil's insight is so complete that he can read Dante's mind, understanding his thoughts and emotions better than Dante does.
The following is loosely paraphrased from The Inferno:
[Virgil]: "Here take this watch, it is the Amazon Halo Band and it grants you divine passage through this hellish notch."
*Dante puts on the watch*
[Dante]: "Thank you wise one, but my trembling hands advise me to run from this place far from the sun."
*Virgil subtly checks his iPhone*
[Virgil]: "Dante I can see your mind, it is not fear you feel, but anticipation for our journey as-yet undefined."
[Dante]: "Wise is a word too weak to describe your perception, you have described my feelings better than my own introspection."
Most scholars are unaware that Virgil read Dante's mind using the Amazon Halo Band (shown above), rather than with learned insight. Instead of devoting his entire life to understanding the human mind through reasoning, all he had to do was plunk down $69.99 (the pre-sale price, $99.99 normally) and voila!
The Amazon Halo Band is unprecedented, not just because of its price, there are plenty of smart watches out there for under $70, like Fitbit and Wyze, but because of its ambition:
"[Amazon Halo Band] uses machine learning to analyze energy and positivity in a customer's voice so they can better understand how they may sound to others, helping improve their communication and relationships." [source]
Amazon designed the Halo Band to buttress four pillars of user health—activity, sleep, body data, and relationships. The first two are standard fare for wearables. Low cost activity tracking has been around since the early 2000s, and night time tracking moved onto wrists in 2013 with the launch of Beddit’s crowdfunded sleep tracker.
While body data is still cutting edge, relationship building sounds like downright science fiction. Amazon Halo is the mobile application behind the Halo Band, and it runs machine learning on stripped-down bathroom selfies to determine users’ body fat percentages. The type of machine learning Amazon uses, known as deep neural networks (DNNs), is intensive, and it requires uploading your underwear pics to Amazon’s cloud.
According to Amazon: “Your body scan images are automatically deleted from the cloud after processing. After that, the images and resulting personalized 3D body model are only stored locally on your phone...You can also opt-into storing your Body scan images in the secure Amazon cloud.”
[source]
However, client-to-server encryption leaves your images open to harvesting off of Amazon’s server by a third party—a rogue employee, or hacker. Amazon’s promise to delete images certainly reduces the attack surface, but the history of big tech data policies point to the asymmetry between promised data hygiene and layers of dirt behind the scenes. Check out this post for more on the vulnerabilities of client-to-server encryption.
How does Amazon’s relationships feature work? For now, Halo analyzes your voice, and lets you know when conversations went well, and when they didn’t. The voice dataset Amazon has amassed from Alexa devices is unfathomably vast, but as anyone who uses sarcasm knows, inflection can reverse the meaning of words and others’ reactions to them. Just extending the vowels in “this meatloaf is so great Mom.” to “this meatloaf is sooo greeaat Mom.” is the difference between a smile and a thank you and getting wacked over the head.
In other words, raw speech-to-text doesn’t get us very far in understanding the nuances of human communication, but tone of voice does. Apparently Amazon employees have been using this feature to improve the quality of their internal meetings and anecdotal reports say the advice Halo Band offers is quite helpful. But would you be okay with being Dante to Amazon’s Virgil and letting them inside the most sacred of sanctuaries—the mind itself?
Amazon anticipated these privacy concerns, and according to them: “your Tone speech samples are processed locally on your phone and deleted automatically after processing.” [source] The less intensive processing required for audio files allows the algorithm to run locally, on users devices, as opposed to body images which require the computational horsepower stabled in Amazon’s cloud.
But it’s unclear how much metadata, or data about data, is stored on Amazon servers. In the case of tone of voice analysis, metadata includes the insights themselves, your actual emotional state, which is far more sensitive than raw audio files which Amazon has already harvested ad infinitum from Alexa devices.
The slow creep of Amazon and big tech into our lives is worth a comprehensive review. Amazon already sits on our front porch, the Ring doorbell camera, and in our living room, Alexa devices, and now it wants to attach itself to our bodies, and inject itself into our brains. These products have a great deal to offer us, so it’s important to weigh the tradeoffs rather than defaulting to dystopian hysteria.
Fighting Technology with Technology
But what if there was a way to use these incredible technologies to improve your life without having to accept privacy tradeoffs? Two technologies, blockchain and confidential computing, provide a way for independent smart device manufacturers to sell devices whose data and machine learning insights are handled trustlessly. A decentralized blockchain can issue device identifiers, and user login credentials, which are decoupled from the manufacturer. These identifiers can then be used to send encrypted data to neutral confidential computing environments where analyses on that data can be run trustlessly. No more underwear shots uploaded to a company’s private cloud.
In other words, the sensitive data sloshing through smart devices, things like passwords and personal data analysis, need not be processed on private servers, like Amazon’s cloud. These services can be provided as a kind of public utility, private in their execution, public in their availability. Blockchain and confidential computing, are a potent combination which eliminate many of the most troublesome privacy tradeoffs you face when purchasing a product or using a service from a technology company.
The rate of technological advancement has outpaced our intuitions about what exactly we are giving up when we introduce new products into our lives. The Amazon Halo Band is a leap forward in functionality that many consumers will hail without introspection. But if we stop and examine what it means to have a private company analyze the emotional lives of millions of people without oversight, regulation, or proof of how that data gets used, we start to understand the importance of investing in alternatives that are transparent and that prioritize user safety and security.
Spoiler alert, Dante gets out of hell and ditches his Amazon Halo Band. But we may not all be so lucky. Fortunately, blockchain and confidential computing offer an escape hatch.