🫱 Introducing 𝐍𝐞𝐮𝐫𝐚𝐥 𝐂𝐨𝐦𝐩𝐮𝐭𝐞𝐫s: 𝐰𝐡𝐚𝐭 𝐢𝐟 𝐀𝐈 𝐝𝐨𝐞𝐬 𝐧𝐨𝐭 𝐣𝐮𝐬𝐭 𝐮𝐬𝐞 𝐜𝐨𝐦𝐩𝐮𝐭𝐞𝐫𝐬 𝐛𝐞𝐭𝐭𝐞𝐫, 𝐛𝐮𝐭 𝐛𝐞𝐠𝐢𝐧𝐬 𝐭𝐨 𝐛𝐞𝐜𝐨𝐦𝐞 𝐭𝐡𝐞 𝐫𝐮𝐧𝐧𝐢𝐧𝐠 𝐜𝐨𝐦𝐩𝐮𝐭𝐞𝐫 𝐢𝐭𝐬𝐞𝐥𝐟? Beyond today's conventional computers, agents, and world models, Neural Computers (NCs) are new frontiers where computation, memory, and I/O move into a learned runtime state. We ask: whether parts of runtime can move inward into the learning system itself. This is our first step toward the Completely Neural Computer (CNC): a general-purpose neural computer with stable execution, explicit reprogramming, and durable capability reuse. Work done with Mingchen Zhuge (@MingchenZhuge), Changsheng Zhao, Haozhe Liu (@HaoZhe65347 ), Zijian Zhou (@ZijianZhou524 ), Shuming Liu (@shuming96 ), Wenyi Wang (@Wenyi_AI_Wang ), Ernie Chang (@erniecyc ), Gael Le Lan, Junjie Fei, Wenxuan Zhang, Zhipeng Cai (@cai_zhipeng ), Zechun Liu (@zechunliu ), Yunyang Xiong (@YoungXiong1 ), Yining Yang, Yuandong Tian (@tydsh ), Yangyang Shi, Vikas Chandra (@vikasc), Juergen Schmidhuber (@SchmidhuberAI)

🔔 new 𝗔𝗴𝗲𝗻𝘁-𝗮𝘀-𝗮-𝗝𝘂𝗱𝗴𝗲 paper: 𝗖𝗮𝗻 𝗔𝗜 𝗮𝗴𝗲𝗻𝘁𝘀 𝗲𝘃𝗮𝗹𝘂𝗮𝘁𝗲 𝗔𝗜 𝗮𝗴𝗲𝗻𝘁𝘀 𝗮𝘀 𝗲𝗳𝗳𝗲𝗰𝘁𝗶𝘃𝗲𝗹𝘆 𝗮𝘀 𝗵𝘂𝗺𝗮𝗻𝘀? 𝗬𝗲𝘀, 𝘁𝗵𝗲𝘆 𝗰𝗮𝗻! 📄 arxiv.org/abs/2410.10934v1 👨‍💻 github.com/metauto-ai/agent-… Introducing 𝗔𝗴𝗲𝗻𝘁-𝗮𝘀-𝗮-𝗝𝘂𝗱𝗴𝗲, a groundbreaking proof-of-concept that reduces costs and time by 97%, while providing rich, intermediate feedback. It precisely captures the natural step-by-step processes of agentic systems. We also developed 𝗗𝗲𝘃𝗔𝗜, a new benchmark featuring 55 automated AI development tasks and 365 requirements. Agent-as-a-Judge not only outperforms LLM-as-a-Judge but also closely mirrors human evaluations with greater efficiency and precision. The real game-changer? It provides reliable reward signals, paving the way for scalable, self-improving agentic systems. Thanks my Meta/KAUST mentors/peers/collaborators @SchmidhuberAI @tydsh @zechunliu @vikasc @YoungXiong1 @vikasc @Obs01ete @erniecyc @oneDylanAshley ... – at Menlo Park, CA