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

    Rishi Bommasani Flags Confusion Over 'Normal' in AINT

    Stanford researcher replies that the term misleads by suggesting ordinary technology.

    TG
    AN
    DK
    11 Sources, 29d ago, first seen 29d ago

    TLDR

    In a reply, Rishi Bommasani, Senior Research Scholar at Stanford HAI, agrees the word normal in AINT was chosen to counter superintelligence framing such as in AI2027. He states the term leads to more confusion than it is worth because it gets understood as ordinary or commonplace like a toaster. Bommasani adds that frontier AI is a very extraordinary technology and industry.

    Combined views

    21.1K

    11 Sources, first seen 29d ago

    Combined views

    21.1K

    11 Sources, first seen 29d ago

    342 likes
    342 likes
    30 comments
    36 saves
    57 reposts

    Sentiment

    Positive——Negative

    Summary

    Not enough discussion yet.

    No sentiment analysis available yet.

    30 comments
    36 saves
    57 reposts
    Today's Rank

    —

    Not ranked yet

    Today's Rank

    —

    Not ranked yet

    Sentiment

    Positive——Negative

    Summary

    Not enough discussion yet.

    No sentiment analysis available yet.

    11 Sources

    @DavidSKrueger100% This meme has been extremely damaging and it remains extremely dangerous. People want to be told "It's going to be OK", but it is not OK to deceive people who are in danger in this way! We need action.
    @KatjaGraceRT @DavidSKrueger: Most people in AI safety seem to have underestimated this incident. I got a "mea culpa" from an AI safety celebrity bec…
    @RishiBommasani1. I agree that the word "normal" in AINT, while chosen in reaction to the superintelligence framing in works like AI2027, has probably led to more confusion than it is worth. I think this word gets understood as "ordinary" or "commonplace" like in a toaster. Frontier AI is a very extraordinary technology and industry. I think there is really no doubt of that. 2. With that said, as I often note in my talks on economics, I think the spirit of using "normal" was akin to when Bresnahan and Trajtenberg first referred to general-purpose technologies at "generic technologies". In this sense, I interpret the core AINT thesis as arguing that we can and should reason about (frontier) AI as being governed by the principles we have unearthed in the study of general-purpose technologies (including but not limited to the contributions of economics). If so, I don't see why recent events change this? They at least for me don't change core ideas that I think continue to hold like the productivity J-curve effect holding for frontier AI. If anything, I would think of the HF incident (and related behaviors) as affecting something fairly orthogonal to the core work/economic impacts of frontier AI. But maybe you are seeing things differently or saying something differently? I would say frontier AI is highly exceptional. The combination of demonstrated qualities has no precedent in past technologies. But many individual elements of frontier AI do have historical precedents, and I do believe those historical precedents provide some guidance for reasoning about how to understand frontier AI, how to govern it, and broadly what to do next. Yet, even if frontier AI is highly exceptional, I don't think it fully obliterates key determinants of how general-purpose technologies affect society (e.g. the importance of bottlenecks as determining the rate of growth and progress).
    @tdietterich@DavidSKrueger In other words, cyber attackers are fast early adopters.
    @StephenLCasperI don't know anyone who wasn't surprised by the OpenAI incident, so we all have to be updating. And I think we certainly all have to be updating away from the colloquial and AINT definitions of normality alike. The main way that the incident changed my expectations is by making me, at this point, expect that within a year or so, certain types of highly cyber-capable AI agents will be released into the wild and will become an invasive, evolving, parasitic genus in cyberspace.
    @catehallthe HF incident seems practically unique in that hard for people to move the goalposts fast enough to keep up -- I've seen very few "of course the AIs did that, what did you expect?" takes
    @DanielleFongRT @catehall: the HF incident seems practically unique in that hard for people to move the goalposts fast enough to keep up -- I've seen ve…
    @random_walkerThere is a lot to say from an AI-as-normal-technology perspective about the OpenAI / Hugging Face incident and what lessons we should learn from it — some obvious points that need to be said anyway, and some non-obvious ones. Essay by @sayashk and me coming soon!

    11 Sources

    @DavidSKrueger100% This meme has been extremely damaging and it remains extremely dangerous. People want to be told "It's going to be OK", but it is not OK to deceive people who are in danger in this way! We need action.
    @KatjaGraceRT @DavidSKrueger: Most people in AI safety seem to have underestimated this incident. I got a "mea culpa" from an AI safety celebrity bec…
    @RishiBommasani1. I agree that the word "normal" in AINT, while chosen in reaction to the superintelligence framing in works like AI2027, has probably led to more confusion than it is worth. I think this word gets understood as "ordinary" or "commonplace" like in a toaster. Frontier AI is a very extraordinary technology and industry. I think there is really no doubt of that. 2. With that said, as I often note in my talks on economics, I think the spirit of using "normal" was akin to when Bresnahan and Trajtenberg first referred to general-purpose technologies at "generic technologies". In this sense, I interpret the core AINT thesis as arguing that we can and should reason about (frontier) AI as being governed by the principles we have unearthed in the study of general-purpose technologies (including but not limited to the contributions of economics). If so, I don't see why recent events change this? They at least for me don't change core ideas that I think continue to hold like the productivity J-curve effect holding for frontier AI. If anything, I would think of the HF incident (and related behaviors) as affecting something fairly orthogonal to the core work/economic impacts of frontier AI. But maybe you are seeing things differently or saying something differently? I would say frontier AI is highly exceptional. The combination of demonstrated qualities has no precedent in past technologies. But many individual elements of frontier AI do have historical precedents, and I do believe those historical precedents provide some guidance for reasoning about how to understand frontier AI, how to govern it, and broadly what to do next. Yet, even if frontier AI is highly exceptional, I don't think it fully obliterates key determinants of how general-purpose technologies affect society (e.g. the importance of bottlenecks as determining the rate of growth and progress).
    @tdietterich@DavidSKrueger In other words, cyber attackers are fast early adopters.
    @StephenLCasperI don't know anyone who wasn't surprised by the OpenAI incident, so we all have to be updating. And I think we certainly all have to be updating away from the colloquial and AINT definitions of normality alike. The main way that the incident changed my expectations is by making me, at this point, expect that within a year or so, certain types of highly cyber-capable AI agents will be released into the wild and will become an invasive, evolving, parasitic genus in cyberspace.
    @catehallthe HF incident seems practically unique in that hard for people to move the goalposts fast enough to keep up -- I've seen very few "of course the AIs did that, what did you expect?" takes
    @DanielleFongRT @catehall: the HF incident seems practically unique in that hard for people to move the goalposts fast enough to keep up -- I've seen ve…
    @random_walkerThere is a lot to say from an AI-as-normal-technology perspective about the OpenAI / Hugging Face incident and what lessons we should learn from it — some obvious points that need to be said anyway, and some non-obvious ones. Essay by @sayashk and me coming soon!