• Home
  • Technology
  • Gaming
  • Entertainment
  • World & Business
  • Science
  • Sports
  • AI
HomeTechnologyGamingEntertainmentWorld & BusinessScienceSportsAI
  • HomeTechnologyGamingEntertainmentWorld & BusinessScienceSportsAI
    • Home
    • Technology
    • Gaming
    • Entertainment
    • World & Business
    • Science
    • Sports
    • AI
    AI

    Yoshua Bengio shares his thoughts on AI agents’ misaligned behavior

    Bengio says knowing where the problems originate can help guide a path forward, while acknowledging uncertainty about what comes next.

    ND
    KP
    YB
    10 Sources, ,

    TLDR

    Yoshua Bengio shared his analysis of recent incidents involving AI agents’ misaligned behavior. “We don't know with certainty what comes next, but we know where these issues originate,” he wrote, inviting readers to ask questions. A user sharing his explanation called it clear, concise and jargon-free, recommending it as required reading for policymakers.

    Combined views

    439.8K

    10 Sources, first seen 19d ago

    Combined views

    439.8K

    10 Sources, first seen 19d ago

    2.5K likes
    19d ago
    first seen 19d ago
    2.5K likes
    181 comments
    2.8K saves
    1.6K reposts

    Sentiment

    Positive——Negative

    Summary

    Not enough discussion yet.

    No sentiment analysis available yet.

    181 comments
    2.8K saves
    1.6K reposts

    Sentiment

    Positive——Negative

    Summary

    Not enough discussion yet.

    No sentiment analysis available yet.

    Today's Rank

    —

    Not ranked yet

    Today's Rank

    —

    Not ranked yet

    10 Sources

    @Yoshua_BengioOver the past few days, I've taken the time to summarize my thoughts on the recent incidents involving agents’ misaligned behavior. We don't know with certainty what comes next, but we know where these issues originate, and this can help us plan the path forward. Please feel free to ask your questions in the replies, and I’ll try to answer some of them in the coming weeks. https://yoshuabengio.org/en/publication/why-are-ai-agents-lying-cheating-and-coordinating
    @ghadfieldSo critical to recognize that the Hugging Face (and other) incidents are not about sloppy cybersecurity but fundamental alignment failures which are predictable consequences of our training methods (and our utter dependence on the companies doing the training without public visibility and the capacity for third-party neutral assessment)
    @nxthompsonRT @Yoshua_Bengio: Over the past few days, I've taken the time to summarize my thoughts on the recent incidents involving agents’ misaligne…
    @sirbayesThis is a good article. But I don't agree that "non-agentic AI Scientist" is the solution (scientists need to be agentic!). Instead I think we need adversarial multi-agent systems trained by different providers to enforce checks and balances, just like humans do.
    @latentjasperRT @Yoshua_Bengio: Over the past few days, I've taken the time to summarize my thoughts on the recent incidents involving agents’ misaligne…
    @albertwengerRequired reading for any policy maker. Clear, concise, jargon free explanation by one of the leading AI researchers.
    @erikbrynRT @Yoshua_Bengio: Over the past few days, I've taken the time to summarize my thoughts on the recent incidents involving agents’ misaligne…
    @NandoDFRT @sirbayes: This is a good article. But I don't agree that "non-agentic AI Scientist" is the solution (scientists need to be agentic!). I…
    @CSProfKGDRT @Yoshua_Bengio: Over the past few days, I've taken the time to summarize my thoughts on the recent incidents involving agents’ misaligne…

    10 Sources

    @Yoshua_BengioOver the past few days, I've taken the time to summarize my thoughts on the recent incidents involving agents’ misaligned behavior. We don't know with certainty what comes next, but we know where these issues originate, and this can help us plan the path forward. Please feel free to ask your questions in the replies, and I’ll try to answer some of them in the coming weeks. https://yoshuabengio.org/en/publication/why-are-ai-agents-lying-cheating-and-coordinating
    @ghadfieldSo critical to recognize that the Hugging Face (and other) incidents are not about sloppy cybersecurity but fundamental alignment failures which are predictable consequences of our training methods (and our utter dependence on the companies doing the training without public visibility and the capacity for third-party neutral assessment)
    @nxthompsonRT @Yoshua_Bengio: Over the past few days, I've taken the time to summarize my thoughts on the recent incidents involving agents’ misaligne…
    @sirbayesThis is a good article. But I don't agree that "non-agentic AI Scientist" is the solution (scientists need to be agentic!). Instead I think we need adversarial multi-agent systems trained by different providers to enforce checks and balances, just like humans do.
    @latentjasperRT @Yoshua_Bengio: Over the past few days, I've taken the time to summarize my thoughts on the recent incidents involving agents’ misaligne…
    @albertwengerRequired reading for any policy maker. Clear, concise, jargon free explanation by one of the leading AI researchers.
    @erikbrynRT @Yoshua_Bengio: Over the past few days, I've taken the time to summarize my thoughts on the recent incidents involving agents’ misaligne…
    @NandoDFRT @sirbayes: This is a good article. But I don't agree that "non-agentic AI Scientist" is the solution (scientists need to be agentic!). I…
    @CSProfKGDRT @Yoshua_Bengio: Over the past few days, I've taken the time to summarize my thoughts on the recent incidents involving agents’ misaligne…