Nangona umgca phakathi kwe-AI automation kunye ne-agency ye-autonomous isibopho, i-technical frontier yokufumaneka, i-agents ebonakalayo, ezinokufumaneka, ezinokufumaneka ngokuhambelana kwi-ecosystems. Thina siye kunye noMathieu Jung, umlawuli we-Marketing ye-Product eYurophu.UkucingaUkubuyekeza into elandelayo kwi-intelligent infrastructure yeyure, iingxaki ezivula, kunye neengcali kunye neenkampani kufuneka zibonise ngokufanayo malunga ne-agent-based computing.
Ishan Pandey:Ukuqala, unokufumana ukuba "I-Agents ye-AI ye-Trusted" yintoni kwi-framework yakho, kunye neendlela ezahlukileyo kwi-AI ye-agents ezizodwa ezibonakalayo kwi-automation okanye iimveliso ze-LLM?
Matthieu Jung:Zonke siyazi ukuba i-AI agents ziquka i-interface entsha. Zibonisa indlela yokwenza, ukwakha, kunye nokuxhumana. Kodwa ukwandisa, i-agents kufuneka into eyodwa: ukhuseleko. I-Privacy ayikho kwimibuzo. I-Agents isebenza kunye neengxaki ezincinane: amaxwebhu ezizodwa, i-crypto assets, i-wallet. Ngaphandle kwe-privacy, akukho ukhuseleko, kwaye ngaphandle kwe-agents ze-trust ayikwazanga. Yeyona yoko i-Agents ze-Trusted AI zithembisa ngokupheleleyo kwi-enclaves ezigqongileyo ezifana ne-Intel TDX. I-Confidential Computing Stack yethu ibhalisele i-logic, i-data kunye neengx
Ishan Pandey:Ukusebenzisana phakathi kwe-iExec kunye ne-Eliza OS kubonakala ukuxhaswa kwe-infrastructure ye-privacy kunye ne-computing ye-autonomous. Indlela yokufumana ngokwenene kunye ne-verifiability kunye ne-confidentiality?
Matthieu Jung:Kuyinto ingxaki enkulu, kwaye kubalulekile ukuba i-Trusted AI Agents iyahlukileyo. I-tension phakathi kwe-confidentiality kunye ne-verifiability ibonakala kwi-architectures ezininzi, kodwa ayikho kwimeko yethu.
Ukusetyenziswa kwe-ElizaOS x iExec stack, sisebenzisa i-Trusted Execution Environments (TEEs), ngokukodwa kwi-Intel TDX, ukunceda kuphela i-agent logic eyenziwe nge-ElizaOS, kodwa i-AI model ngokwayo, eyayisetyenziswa ukuqeqesha okanye ukucacisa i-agent. Oku kubalulekileyo. Oku kuthetha ukuba yonke indawo yokusebenza, kuquka i-logic kunye neemparamitha ze-model sensitive, ibekwe ukusuka ukufikelela ngaphandle kwe-cloud hosts okanye abasebenzisi be-system. Kwakhona, le enclaves zenza izitifiketi ze-cryptographic ezibonisa ukufikelela kwe-execution. Ngoko ke siyenza ukufumana i-verifiability: wonke umntu unokufuneka into ef
Ishan Pandey:Kwihlabathi apho iinkampani ezininzi ze-AI zihlanganisa kakhulu kwi-API ze-centralized kunye neemodeli ze-closed, ngoko kufuneka ukwakha i-agents ezihlangeneyo ezihlangeneyo kwi-design?
Matthieu Jung:I-iExec yenza lula ukuhlola zonke iingxaki kwi-chain. iExec iye yenza i-iApp Generator, isixhobo ye-developer evumela abathengi ukwakha kunye nokuvelisa i-apps ze-confidential kwimiveki. I-MCP Model Context Protocol iye yenza i-server eyenziwe kwi-Agents ye-Trusted ukuze ziyafumaneka ngexesha elide.
Ishan Pandey:Yintoni i-infrastructure ye-iExec inikeza i-agent ye-help kumazwe, i-state, okanye i-logic yayo kumazwe omnye?
Matthieu Jung:I-iExec isebenza i-agents ngaphakathi kwe-Intel TDX enclaves ezisebenze i-proofs ezihlangeneyo ye-code, i-inputs, kunye ne-outputs. Ezi i-proofs zingathunyelwa kwi-chain ukuze wonke umntu unokufuneka i-identity ye-enclave kunye nokubonisa ukuba i-agent yenza ngokufanelekileyo okuyithiweyo, ngaphandle kokubonisa idatha ye-private okanye i-code.
Ishan Pandey:Yintoni iingxaki zokusetyenziswa ukuba iingxaki ze-AI ziye zithunyelwe ukuba zithunyelwe kwiinkwenkwezi okanye kwiinkqubo? Ngaba siqhelekileyo ukuba i-intelligent ye-Intelligent ye-Intelligent ye-Intelligent?
Matthieu Jung:Ukuvumela i-agents ukuba zithembele kwiinkwenkwezi okanye i-apps, kubaluleke iziphumo ezintsha zeengxaki kunye neengxaki ezidlulileyo. I-bug kwi-protocol eyodwa ingangena kwi-protocol eyahlukileyo. Kwakhona, ngaphandle kwezinga ezininzi, ukuvavanya i-responsibility okanye ukuguqulwa kwimiphumo embalwa kunokwenzeka. Yoko ke, ukhuseleko kunye nokulawulwa kubaluleke kakhulu xa kuxhomekeka kwi-agents ezahlukileyo kunye ne-automating iingxaki kwi-blockchain.
Ishan Pandey:Ngaba unokufundisa kwimeko ye-hypothetic okanye i-use case apho i-Agent ye-AI ye-Trusted isetyenziswe kwi-wild, into enze, indlela yokubonisa ukuqhuba kwayo, yaye ngoko i-imeyile engcono kunokuba iimodeli eziqhelekileyo?
Matthieu Jung:Izixhobo ze-AI zibonisa i-trade yakho, ukulawula i-deFi, ukufumana iingcebiso, ukufikelela i-trade, kunye nokuguqula i-portfolios. Kufuneka ukuxhaswa kunye needatha ezincinane kunye neengcebiso ezifana ne-private key yakho okanye i-wallet yakho. Ngei-iExec, i-logic ye-agent isebenza ngokufanelekileyo kwi-TEE. Abathengisi bangakwazi ukuvikela iinkcukacha zabo, ukuxhaswa kwe-access ngokufanelekileyo kunye ne-model, kwaye kufumana iingcebiso ze-confidential. Zonke iingcebiso kunye nenkqubo zokusebenza ngokufanelekileyo kunye ne-verified.
Ishan Pandey:Okugqibela, ukuba abathengi kunye nabathengisi abalandeli i-Trusted AI Agent, iingcebiso ezidlulileyo ezidlulileyo ezidlulileyo kwiintsuku ezidlulileyo ze-12?
Matthieu Jung:Kwakhona sinazi ukuba i-agents ziyafumaneka kwi-interface entsha. Kwakhona, oku kuthetha ukuba siza kuqala ukufumana i-agent prompts kunye ne-data exposures, kwaye oku kuthetha ukuba i-private execution iyiphi na ingxaki. I-Agents ye-innovation eyayithanda kwinqanaba elandelayo i-MCP okanye i-Model Context Protocol. Yenza i-agents ukunikezela i-snapshots encrypted yayo zokusebenza njengoko zithembisa kwaye zithembisa kwi-apps ezininzi. iExec iye yenza i-MCP server. Lezi zithembiso zibonisa ukuba izakhiwo ezininzi ze-decentralized, i-privacy-first-agents kunokuba zithembise kwi-
Ngaba ushiye ukuthetha kwaye ushiye iinkcukacha!
