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

    Slowing OpenAI and Anthropic's frontier AI development without antitrust exemptions

    The critic argues that potential cyberattack liability and customer demand for predictable models give OpenAI and Anthropic business reasons—not just altruistic ones—to favor reliability over raw power.

    VK
    JA
    GM
    15 Sources, ,

    TLDR

    A post supports OpenAI and Anthropic slowing frontier AI development if they believe their unreleased models are dangerous enough to warrant it. But it argues that neither antitrust exemptions nor a regulatory approval process that supersedes product liability is needed. Calling the companies a frontier-AI “duopoly,” the critic also objects to evaluators policing competitors it says are not at the frontier.

    Combined views

    8.2M

    15 Sources, first seen 18d ago

    Combined views

    8.2M

    15 Sources, first seen 18d ago

    66.3K likes
    18d ago
    first seen 18d ago
    66.3K likes
    3.3K comments
    13K saves
    10.8K reposts
    3.3K comments
    13K saves
    10.8K reposts

    Sentiment

    Positive34.7%65.3%Negative

    Summary

    Accounts criticized AI regulation stances involving China and superintelligence as regulatory capture or hypocritical, while positive replies praised antitrust critiques and AI's help with writing clarity.

    Based on 185 sentiment-bearing replies from 173 accounts across 10 conversations.

    Today's Rank

    —

    Not ranked yet

    Today's Rank

    —

    Not ranked yet

    Sentiment

    Positive34.7%65.3%Negative

    Summary

    Accounts criticized AI regulation stances involving China and superintelligence as regulatory capture or hypocritical, while positive replies praised antitrust critiques and AI's help with writing clarity.

    Based on 185 sentiment-bearing replies from 173 accounts across 10 conversations.

    15 Sources

    @DavidSacksDario has written that we need to “pace the frontier,” and Sam has agreed. People may be surprised by my response: go ahead. You guys are the frontier. By any reasonable metric — market share, revenue growth, model capability — the two of you have a duopoly on frontier intelligence. You’ve also claimed the lead is widening because of recursive self-improvement. I don’t see what you see in the lab. If the unreleased models are scary enough that you think you should slow down, I support your decision to be responsible. But stop pretending you need anyone else’s permission. Stop pretending antitrust law has to be suspended so you can form a cartel. Stop pretending you need a regulatory approval process that supersedes product liability. Stop pretending METR is independent when it is intertwined with Anthropic’s investors and staff. Stop pretending you need those same evaluators to police competitors who aren’t even at the frontier. Most of all, stop pretending the motivation to slow down is purely altruistic. You face massive product-liability exposure if your products enable a truly damaging cyberattack. The market already punishes models that behave in unpredictable or unauthorized ways. After the Hugging Face episode, it is simply good business for OpenAI and Anthropic to trade some raw power for reliability and predictability. Call it alignment if you want. It is also just giving customers what they want. Pacing the frontier would also create breathing room for a more intelligent conversation about regulation than Bernie Sanders’ “shut it all down.” China is very unlikely to join a global agreement, as you know, and that has to be taken into account as well. So go ahead and pace the frontier. You are the ones setting it. The easiest way not to build superintelligence is for you to agree not to build it. Demanding your preferred regulatory framework as the price of that will look like blackmail of the public and the political system. So just do it. If you do, you’ll buy goodwill for the next conversation. If you don’t, we’ll know this was just another bid for regulatory capture — or an election-season psyop.
    @KelseyTuoc@CharlieBull0ck imo you can do this if and only if you're willing to share what you prompted it with. (no one is, though)
    @knowerofmarketsVery well said, emotionally intelligent response and pretty much sums up how I’ve felt about all of this, which is - stop complaining to us and just do something yourselves. It’s clear that any lab employee knows more about RSI than a congressman chosen at random. Why lean on the latter to dictate how decisions at YOUR company should be made?
    @BrianRoemmeleA brilliant take on the Dario AI Theater. Thank you for clarity and not rubber stamps.
    @ScobleizerI totally disagree. Using AI can help you get your points across better. This was awesome. And I knew AI was involved because of the em dashes. I can’t stand this attitude.
    @jachiam0If you owe the bank $100, it's your problem. If you owe the bank $100 billion, it's the bank's problem. Similarly: if you sell the government one gun, it has the monopoly on violence. If you sell the government all of the guns, you have the monopoly on violence. This is roughly where I'm worried things are going with AI. It is hard to see any future where every facet of military strength doesn't eventually depend on (and maybe get fully executed by) AI. Force projection depends on technological progress in many fields (e.g. drones, sensors, guidance systems, chemicals, etc). Well, the relevant fields are going to eventually be automated with most or all R&D conducted by AI. Strategy development and situational awareness depend on the ability to ingest a ton of data, think ahead, and reason over many variables. Well, that is going to all be done by AI at a certain point. Building stockpiles of materiel and replenishing supplies as they get used up depends on the efficacy of an industrial base. Well, eventually all of the supply chains are going to get instrumented and loaded with robots and increasingly automated and orchestrated by AI. The AI will continuously redesign the supply chains to eke out efficiency gains until they're quasi-alien and no human has sufficient expertise to understand the whole thing end-to-end. Intelligence analysis, image processing, surveillance... It goes on and on. It's the Alison Burke "Square Hole" video, everything goes in the AI bucket eventually. And when these things get done by AI, it won't be AI designed and run by the government. It will be AI supplied by probably four or five frontier labs. Even if the government manages some in-house expertise for using open source near-frontier models, the government is still ultimately going to be forced, in order to remain competitive, to engage with the frontier labs. Frontier models will not come from the government because it has no idea how to make this technology. Government has largely hollowed out its capacity for technological excellence, ceding more and more ground to the private sector over the years. There is a veritable Gordian knot of reasons why it can't build, and it won't be fixed anytime soon (if it can be fixed at all). This puts the labs in a powerful position to influence government policy through choices in model design and usage policies. This is not entirely a bad thing: many of the people staffing these labs are citizens who are entitled to participate in the democratic process through vocalizing their opinions and building what they want to build. They should not have zero power. But the power that stands to concentrate onto them is quite immense, moreso, I believe, than any historical comparison. I am not saying all of these things are going to happen fast, nor am I saying they will happen overnight. (Robots are still a ten year problem and you furthermore have a long tail of integration problems beyond that.) But it feels like the writing is on the wall that one of the fundamental mandates of government - to provide for the national defense and own the monopoly on violence - is on track to devolve from it very quickly (on historical timescales), maybe only over 1-2 decades at this point. Certainly on cyber offense and defense, it will fully devolve from government in a matter of months. A couple of months ago I signed the Pacing the Frontier letter. I have quibbles with the wording: I don't think RSI is exactly the quantity of interest here, or at the very least, it's not the only one. But thinking through all of the above issues, for me the inescapable conclusion is that the government has to have some levers of choice here. They need to be able to modulate how fast the frontier goes, and choose when and how different steps happen. They need to have staff in the inner loop of these decisions so that government has continuous situational awareness and remains deeply informed. They need to be able to choose which advances are gated on national and global readiness. They need - and I say this despite feeling it is remarkably unpleasant - to be able to choose which advances belong exclusively to government and which should be widely accessible. This is not something I greatly like. I don't expect great things from government when it comes to technology. But, I don't want to live in a society where a small group of AI companies is a layer through which all decisions about the use of force have to be filtered. Those powers belong to the duly legitimated institutions of the people, by the people, and for the people.
    @BlackHCSo it's frontier models trying to persuade people now that we shouldn't pace the frontier? 🤯😅
    @GaryMarcus“stop pretending you need anyone else’s permission. Stop pretending antitrust law has to be suspended so you can form a cartel. Stop pretending you need a regulatory approval process that supersedes product liability. Stop pretending METR is independent when it is intertwined with Anthropic’s investors and staff. Stop pretending you need those same evaluators to police competitors who aren’t even at the frontier. Most of all, stop pretending the motivation to slow down is purely altruistic.”
    @w01feI agree with all of this emphatically, except I would qualify the superintelligence ban. Currently, to my knowledge nobody has any solid idea how to align or control superintelligemce, so I think something that amounts to a *temporary* ban until we do is a pretty good idea to the extent it could be reasonably and practically enforced worldwide. But of course that’s the rub. Someone might defect, and in the long run I doubt it’s possible to prevent superintelligence without halting (or maybe reverting, at this point) the progress of computing altogether or creating a draconian surveillance state (by default, at some point computers will get so good that a 4chan kid could do an intelligence explosion in their basement for the lulz). So a permanent ban likely seems impractical, and even if the science isn’t solved maybe at some point society would need to decide to go for it anyway despite the risk because it’s the least bad option.
    @BorisMPowerI believe everyone here is acting out of best intentions and I’m also glad that this is your position David. Excited for the world of a lot more intelligence, which we will bring about through thoughtful discussions and decisions along the way.

    15 Sources

    @DavidSacksDario has written that we need to “pace the frontier,” and Sam has agreed. People may be surprised by my response: go ahead. You guys are the frontier. By any reasonable metric — market share, revenue growth, model capability — the two of you have a duopoly on frontier intelligence. You’ve also claimed the lead is widening because of recursive self-improvement. I don’t see what you see in the lab. If the unreleased models are scary enough that you think you should slow down, I support your decision to be responsible. But stop pretending you need anyone else’s permission. Stop pretending antitrust law has to be suspended so you can form a cartel. Stop pretending you need a regulatory approval process that supersedes product liability. Stop pretending METR is independent when it is intertwined with Anthropic’s investors and staff. Stop pretending you need those same evaluators to police competitors who aren’t even at the frontier. Most of all, stop pretending the motivation to slow down is purely altruistic. You face massive product-liability exposure if your products enable a truly damaging cyberattack. The market already punishes models that behave in unpredictable or unauthorized ways. After the Hugging Face episode, it is simply good business for OpenAI and Anthropic to trade some raw power for reliability and predictability. Call it alignment if you want. It is also just giving customers what they want. Pacing the frontier would also create breathing room for a more intelligent conversation about regulation than Bernie Sanders’ “shut it all down.” China is very unlikely to join a global agreement, as you know, and that has to be taken into account as well. So go ahead and pace the frontier. You are the ones setting it. The easiest way not to build superintelligence is for you to agree not to build it. Demanding your preferred regulatory framework as the price of that will look like blackmail of the public and the political system. So just do it. If you do, you’ll buy goodwill for the next conversation. If you don’t, we’ll know this was just another bid for regulatory capture — or an election-season psyop.
    @KelseyTuoc@CharlieBull0ck imo you can do this if and only if you're willing to share what you prompted it with. (no one is, though)
    @knowerofmarketsVery well said, emotionally intelligent response and pretty much sums up how I’ve felt about all of this, which is - stop complaining to us and just do something yourselves. It’s clear that any lab employee knows more about RSI than a congressman chosen at random. Why lean on the latter to dictate how decisions at YOUR company should be made?
    @BrianRoemmeleA brilliant take on the Dario AI Theater. Thank you for clarity and not rubber stamps.
    @ScobleizerI totally disagree. Using AI can help you get your points across better. This was awesome. And I knew AI was involved because of the em dashes. I can’t stand this attitude.
    @jachiam0If you owe the bank $100, it's your problem. If you owe the bank $100 billion, it's the bank's problem. Similarly: if you sell the government one gun, it has the monopoly on violence. If you sell the government all of the guns, you have the monopoly on violence. This is roughly where I'm worried things are going with AI. It is hard to see any future where every facet of military strength doesn't eventually depend on (and maybe get fully executed by) AI. Force projection depends on technological progress in many fields (e.g. drones, sensors, guidance systems, chemicals, etc). Well, the relevant fields are going to eventually be automated with most or all R&D conducted by AI. Strategy development and situational awareness depend on the ability to ingest a ton of data, think ahead, and reason over many variables. Well, that is going to all be done by AI at a certain point. Building stockpiles of materiel and replenishing supplies as they get used up depends on the efficacy of an industrial base. Well, eventually all of the supply chains are going to get instrumented and loaded with robots and increasingly automated and orchestrated by AI. The AI will continuously redesign the supply chains to eke out efficiency gains until they're quasi-alien and no human has sufficient expertise to understand the whole thing end-to-end. Intelligence analysis, image processing, surveillance... It goes on and on. It's the Alison Burke "Square Hole" video, everything goes in the AI bucket eventually. And when these things get done by AI, it won't be AI designed and run by the government. It will be AI supplied by probably four or five frontier labs. Even if the government manages some in-house expertise for using open source near-frontier models, the government is still ultimately going to be forced, in order to remain competitive, to engage with the frontier labs. Frontier models will not come from the government because it has no idea how to make this technology. Government has largely hollowed out its capacity for technological excellence, ceding more and more ground to the private sector over the years. There is a veritable Gordian knot of reasons why it can't build, and it won't be fixed anytime soon (if it can be fixed at all). This puts the labs in a powerful position to influence government policy through choices in model design and usage policies. This is not entirely a bad thing: many of the people staffing these labs are citizens who are entitled to participate in the democratic process through vocalizing their opinions and building what they want to build. They should not have zero power. But the power that stands to concentrate onto them is quite immense, moreso, I believe, than any historical comparison. I am not saying all of these things are going to happen fast, nor am I saying they will happen overnight. (Robots are still a ten year problem and you furthermore have a long tail of integration problems beyond that.) But it feels like the writing is on the wall that one of the fundamental mandates of government - to provide for the national defense and own the monopoly on violence - is on track to devolve from it very quickly (on historical timescales), maybe only over 1-2 decades at this point. Certainly on cyber offense and defense, it will fully devolve from government in a matter of months. A couple of months ago I signed the Pacing the Frontier letter. I have quibbles with the wording: I don't think RSI is exactly the quantity of interest here, or at the very least, it's not the only one. But thinking through all of the above issues, for me the inescapable conclusion is that the government has to have some levers of choice here. They need to be able to modulate how fast the frontier goes, and choose when and how different steps happen. They need to have staff in the inner loop of these decisions so that government has continuous situational awareness and remains deeply informed. They need to be able to choose which advances are gated on national and global readiness. They need - and I say this despite feeling it is remarkably unpleasant - to be able to choose which advances belong exclusively to government and which should be widely accessible. This is not something I greatly like. I don't expect great things from government when it comes to technology. But, I don't want to live in a society where a small group of AI companies is a layer through which all decisions about the use of force have to be filtered. Those powers belong to the duly legitimated institutions of the people, by the people, and for the people.
    @BlackHCSo it's frontier models trying to persuade people now that we shouldn't pace the frontier? 🤯😅
    @GaryMarcus“stop pretending you need anyone else’s permission. Stop pretending antitrust law has to be suspended so you can form a cartel. Stop pretending you need a regulatory approval process that supersedes product liability. Stop pretending METR is independent when it is intertwined with Anthropic’s investors and staff. Stop pretending you need those same evaluators to police competitors who aren’t even at the frontier. Most of all, stop pretending the motivation to slow down is purely altruistic.”
    @w01feI agree with all of this emphatically, except I would qualify the superintelligence ban. Currently, to my knowledge nobody has any solid idea how to align or control superintelligemce, so I think something that amounts to a *temporary* ban until we do is a pretty good idea to the extent it could be reasonably and practically enforced worldwide. But of course that’s the rub. Someone might defect, and in the long run I doubt it’s possible to prevent superintelligence without halting (or maybe reverting, at this point) the progress of computing altogether or creating a draconian surveillance state (by default, at some point computers will get so good that a 4chan kid could do an intelligence explosion in their basement for the lulz). So a permanent ban likely seems impractical, and even if the science isn’t solved maybe at some point society would need to decide to go for it anyway despite the risk because it’s the least bad option.
    @BorisMPowerI believe everyone here is acting out of best intentions and I’m also glad that this is your position David. Excited for the world of a lot more intelligence, which we will bring about through thoughtful discussions and decisions along the way.