LESSWRONG
LW

Comment Permalink
konstantin144

The letter feels rushed and leaves me with a bunch of questions.

1. "recent months have seen AI labs locked in an out-of-control race to develop and deploy ever more powerful digital minds that no one – not even their creators – can understand, predict, or reliably control." 

Where is the evidence of this "out-of-control race"? Where is the argument that future systems could be dangerous?


2. "Should we let machines flood our information channels with propaganda and untruth? Should we automate away all the jobs, including the fulfilling ones? Should we develop nonhuman minds that might eventually outnumber, outsmart, obsolete and replace us? Should we risk loss of control of our civilization? Such decisions must not be delegated to unelected tech leaders." 

These are very different concerns that water down what the problem is the letter tries to address. Most of them are deployment questions more than development questions.

3. I like the idea of a six-month collaboration between actors. I also like the policy asks they include.

4. The main impact of this letter would obviously be getting the main actors to halt development (OpenAI, Anthropic, DeepMind, MetaAI, Google). Yet, those actors seem not to have been involved in this letter/ haven't publicly commented. (afaik) This seems like a failure.

5. Not making it possible to verify the names is a pretty big mistake.

6. In my perception, the letter mostly appears alarmist at the current time, especially since it doesn't include an argument for why future systems should be dangerous. It might just end up burning political capital.

Reply
konstantin63

Update: I think it doesn't make much sense to interpret the letter literally. Instead, it can be seen as an attempt to show that a range of people think that slowing down progress would be good, and I think it does an okay job at that (though I still think the wording could be much better, and it should present arguments for why we should decelerate.)

Reply
13Greg C
GPT-4 was rushed, and the OpenAI Plugin store. Things are moving far too fast for comfort. I think we can forgive this response for being rushed. It's good to have some significant opposition working on the brakes to the runaway existential catastrophe train that we've all been put on.
See in context
by Zach Stein-Perlman
3 min read
123

126

This is a linkpost for https://futureoflife.org/open-letter/pause-giant-ai-experiments/

We call on all AI labs to immediately pause for at least 6 months the training of AI systems more powerful than GPT-4.

 

AI systems with human-competitive intelligence can pose profound risks to society and humanity, as shown by extensive research and acknowledged by top AI labs. As stated in the widely-endorsed Asilomar AI Principles, Advanced AI could represent a profound change in the history of life on Earth, and should be planned for and managed with commensurate care and resources. Unfortunately, this level of planning and management is not happening, even though recent months have seen AI labs locked in an out-of-control race to develop and deploy ever more powerful digital minds that no one – not even their creators – can understand, predict, or reliably control.

Contemporary AI systems are now becoming human-competitive at general tasks, and we must ask ourselves: Should we let machines flood our information channels with propaganda and untruth? Should we automate away all the jobs, including the fulfilling ones? Should we develop nonhuman minds that might eventually outnumber, outsmart, obsolete and replace us? Should we risk loss of control of our civilization? Such decisions must not be delegated to unelected tech leaders. Powerful AI systems should be developed only once we are confident that their effects will be positive and their risks will be manageable. This confidence must be well justified and increase with the magnitude of a system's potential effects. OpenAI's recent statement regarding artificial general intelligence, states that "At some point, it may be important to get independent review before starting to train future systems, and for the most advanced efforts to agree to limit the rate of growth of compute used for creating new models." We agree. That point is now.

Therefore, we call on all AI labs to immediately pause for at least 6 months the training of AI systems more powerful than GPT-4. This pause should be public and verifiable, and include all key actors. If such a pause cannot be enacted quickly, governments should step in and institute a moratorium.

AI labs and independent experts should use this pause to jointly develop and implement a set of shared safety protocols for advanced AI design and development that are rigorously audited and overseen by independent outside experts. These protocols should ensure that systems adhering to them are safe beyond a reasonable doubt. This does not mean a pause on AI development in general, merely a stepping back from the dangerous race to ever-larger unpredictable black-box models with emergent capabilities.

AI research and development should be refocused on making today's powerful, state-of-the-art systems more accurate, safe, interpretable, transparent, robust, aligned, trustworthy, and loyal.

In parallel, AI developers must work with policymakers to dramatically accelerate development of robust AI governance systems. These should at a minimum include: new and capable regulatory authorities dedicated to AI; oversight and tracking of highly capable AI systems and large pools of computational capability; provenance and watermarking systems to help distinguish real from synthetic and to track model leaks; a robust auditing and certification ecosystem; liability for AI-caused harm; robust public funding for technical AI safety research; and well-resourced institutions for coping with the dramatic economic and political disruptions (especially to democracy) that AI will cause.

Humanity can enjoy a flourishing future with AI. Having succeeded in creating powerful AI systems, we can now enjoy an "AI summer" in which we reap the rewards, engineer these systems for the clear benefit of all, and give society a chance to adapt. Society has hit pause on other technologies with potentially catastrophic effects on society.  We can do so here. Let's enjoy a long AI summer, not rush unprepared into a fall.

 

Signatories include Yoshua Bengio, Stuart Russell, Elon Musk, Steve Wozniak, Yuval Noah Hariri, Andrew Yang, Connor Leahy (Conjecture), and Emad Mostaque (Stability).

Edit: covered in NYT, BBC, WaPo, NBC, ABC, CNN, CBS, Time, etc. See also Eliezer's piece in Time.

Edit 2: see FLI's FAQ.

Edit 3: see FLI's report Policymaking in the Pause.

Edit 4: see AAAI's open letter.

AI
Personal Blog

126

Mentioned in
FLI open letter: Pause giant AI experiments
New Comment
123 comments, sorted by
top scoring
Click to highlight new comments since:
Some comments are truncated due to high volume. (⌘F to expand all)Change truncation settings
[-]Ben Pace10649

I think this letter is not robust enough to people submitting false names. Back when Jacob and I put together DontDoxScottAlexander.com we included this section, and I would recommend doing something pretty similar:

I think someone checking some of these emails would slow down high-profile signatories by 6-48 hours, but sustain trust that the names are all real.

I'm willing to help out if those running it would like, feel free to PM me.

Reply
[-]AnnaSalamon515

Oh no.  Apparently also Yann LeCun didn't really sign this.

https://twitter.com/ylecun/status/1640910484030255109

Reply
[-]localdeity173

Indeed.  Among the alleged signatories:

Xi Jinping, Poeple's Republic of China, Chairman of the CCP, Order of the Golden Eagle, Order of Saint Andrew, Grand Cordon of the Order of Leopold

Which I heavily doubt.

Reply
[-]Ben Pace3712

I concur, the typo in "Poeple" does call into question whether he has truly signed this letter.

Reply
1Shankar Sivarajan
No, he really DOES have all those honors.
[-]jimrandomh128

I believe the high-profile names at the top are individually verified, at least, and it looks like there's someone behind the form deleting fake entries as they're noticed. (Eg Yann LeCun was on the list briefly, but has since been deleted from the list.)

Reply
[-]Ben Pace2719

When we did Scott's petition, names were not automatically added to the list, but each name was read by me-or-Jacob, and if we were uncertain about one we didn't add it without checking in with others or thinking it over. This meant that added names were staggered throughout the day because we only checked every hour or two, but overall prevented a number of fake names from getting on there.

(I write this to contrast it with automatically adding names then removing them as you notice issues.)

Reply
[-]the gears to ascension185

Lecun heard he was incorrectly added to the list, so the reputational damage still mostly occurred.

Reply
[-]M. Y. Zuo1311

Yeah this seems a bit of a self-defeating exercise. 

Who made the decision to go ahead with this method of collecting signatories?

They didn't even have a verification hold on submitted names...

Reply
5Evan R. Murphy
Yea they made a mistake not verifying signatures from the beginning. But they have course-corrected, see this notice FLI has posted now above the signatories list: Screenshot
[-]Esben Kran5727

It seems like there's a lot of negative comments about this letter.  Even if it does not go through, it seems very net positive for the reason that it makes explicit an expert position against large language model development due to safety concerns. There's several major effects of this, as it enables scientists, lobbyists, politicians and journalists to refer to this petition to validate their potential work on the risks of AI, it provides a concrete action step towards limiting AGI development, and it incentivizes others to think in the same vein about concrete solutions.

I've tried to formulate a few responses to the criticisms raised:

  • "6 months isn't enough to develop the safety techniques they detail": Besides it being at least 6 months, the proposals seem relatively reasonable within something as farsighted as this letter. Shoot for the moon and you might hit the sky, but this time the sky is actually happening and work on many of their proposals is already underway. See e.g. EU AI Act, funding for AI research, concrete auditing work and safety evaluation on models. Several organizations are also working on certification and the scientific work towards watermarking is sort
... (read more)
Reply
8Ben Pace
Just repeating from my other comments: my main issue is the broad proposal of "let's get governments involved" that appears to not be aware of all the horrible and corrupt things governments do by-default when they get involved (cf. Covid), nor proposes any ways to avoid lots of dysfunction.
1Leo
Just compare how collaborative governments solving global warming versus Gates Foundation solving polio.
7M. Y. Zuo
The top 40 names were definitely not validated at the time of posting. (or even now?) Has anyone representing the organization stated on the record that the remaining signatories are legitimate?
[-]Ben Pace288

AI labs and independent experts should use this pause to jointly develop and implement a set of shared safety protocols for advanced AI design and development that are rigorously audited and overseen by independent outside experts. These protocols should ensure that systems adhering to them are safe beyond a reasonable doubt.

That's nice, but I don't currently believe there are any audits or protocols that can prove future AIs safe "beyond a reasonable doubt".

In parallel, AI developers must work with policymakers to dramatically accelerate development of robust AI governance systems. These should at a minimum include: new and capable regulatory authorities dedicated to AI; oversight and tracking of highly capable AI systems and large pools of computational capability; provenance and watermarking systems to help distinguish real from synthetic and to track model leaks; a robust auditing and certification ecosystem; liability for AI-caused harm; robust public funding for technical AI safety research; and well-resourced institutions for coping with the dramatic economic and political disruptions (especially to democracy) that AI will cause.

Would the authors also like the moon on a stic... (read more)

Reply
9Evan R. Murphy
The letter isn't perfect, but the main ask is worthwhile as you said. Coordination is hard, stakes are very high and time may be short, so I think it is good to support these efforts if they are in the ballpark of something you agree with.
[-]Ben Pace2514

I don't agree with the recommendation, so I don't think I should sign my name to it.

To describe a concrete bad thing that may happen: suppose the letter is successful and then there is a pause. Suppose a bunch of AI companies agree to some protocols that they say that these protocols "ensure that systems adhering to them are safe beyond a reasonable doubt". If I (or another signatory) is then to say "But I don't think that any such protocols exist" I think they'd be in their right to say "Then why on Earth did you sign this letter saying that we could find them within 6 months?" and then not trust me again to mean the things I say publicly.

Reply
[-]Evan R. Murphy195

The letter says to pause for at least 6 months, not exactly 6 months.

So anyone who doesn't believe that protocols exist to ensure the safety of more capable AI systems shouldn't avoid signing the letter for that reason, because the letter can be interpreted as supporting an indefinite pause in that case.

Reply
[-]Ben Pace216

Oh, I didn't read that correctly. Good point.

I am concerned about some other parts of it, that seem to imbue a feeling of "trust in government" that I don't share, and I am concerned that if this letter is successful then governments will get involved in a pretty indiscriminate and corrupt way and then everything will get worse; but my concern is somewhat vague and hard to pin down.

I think it'd be good for me to sleep on it, and see if it seems so bad to sign on to the next time I see it.

Reply
7Ben Pace
I've slept, and now looked it over again. I don’t agree with the clear implication that the problem with tech leaders is that they weren’t elected. I commonly think their judgment is better than people who are elected and in government. I think competent and elected people are best, but given the choice between only competent or only elected (in the current shitty electoral systems of the UK and the US that I am familiar with), I think I prefer competent. I don’t think this threat is very good. Firstly, it seems a bit empty. This is not the government speaking, I don’t know that FLI is in a position to make the government do this. Secondly, it doesn’t feel like… it feels closer to just causing uncontrollable chaos than it does to doing anything sensible. Maybe it works, but I haven’t seen any arguments that governments won’t just become pretty corrupt and mess everything up if given pretty vague yet significant power. I would much rather the current status quo where I think, in principle, you can have a conversation with the relevant decision-makers and have arguments with them, rather than surrender to the government bureaucracies where pretty commonly there is nobody with any power to do anything differently, and any free power is being competed over by the biggest possible factions in the country, and also a lot of the unaccountable decision-makers are pretty corrupt (cf. Patrick McKenzie and Zvi’s writing about how Covid was dealt with), and where close to literally zero people understand how the systems work or the arguments for existential risks here. I think people writing about realistic and concrete stories of how this could go well would change my mind here, but I don’t want to put my name to this threat, seems overall like reducing civilization’s agency in the matter, and I wouldn’t currently take this action myself. I think a bunch of the rest of the letter is pretty good, I like the AI summer and AI fall bit at the end (it is tasteful in relation to
2Evan R. Murphy
Just want to flag that I'm pretty sure this isn't true anymore. At least a few important people in the US government (and possibly many) have now taken this course . I am still in progress on my technical review of the course for AIGS Canada, but my take so far is that it provides a good education on relevant aspects of AI for a non-technical audience and also focuses quite a bit on AI existential risk issues. (I know this only one point out of many you made but I wanted to respond to it when I spotted it and had time.)
4Ben Pace
Yep, it seems to good to me to respond to just one point that you disagreed with, definitely positive to do so relative to responding to none :) I genuinely have uncertainty here, I know there were a bunch of folks at CSET who understood some of the args, I'm not sure whether/what roles they have in Government, I think of many of them as being in "policy think tanks" that are outside of government. Matheny was in the White House for a while but now he runs RAND; if he were still there I would be wrong and there would be at least one person who I believe groks the arguments and how a neural net works. Most of my current probability mass is on literally 100% of elected officials do not understand the arguments or how a neural net works, but I acknowledge that they're not the only people involved in passing legislation/regulation.
2[anonymous]
Each 6 months pause costs all the knowledge and potential benefits of less stupid models (and prevents anyone from discovering their properties which may or may not be as bad as feared). Each pause narrows the lead with less ethical competitors.
1sanxiyn
To state the obvious, pause narrows the lead with less ethical competitors only if pause is not enforced against less ethical competitors. I don't think anyone is in favor of unenforced pause: that would be indeed stupid, as the basic game theory says. My impression is that we disagree on how feasible it is to enforce the pause. In my opinion, at the moment, it is pretty feasible, because there simply are not so many competitors. Doing an LLM training run is a rare capability now. Things are fragile and I am in fact unsure whether it would be feasible next year.
2[anonymous]
How would the pause be enforced against foreign governments? We can't stop them from building nukes or starting wars or mass imprisoning their own people or mass murder or... How would it be enforced against foreign and domestic companies working under the umbrella of the government? How would the pause be enforced at all? No incident has actually happened yet, what are the odds the us government passes legislation about a danger that hasn't even been seen?
4DanielFilan
I think you can support a certain policy without putting your name to a flawed argument for that policy. And indeed ensuring that typical arguments for your policy are high-quality is a forrm of support.
5evhub
I think you can do this with a capabilities test (e.g. ARC's), just not with an alignment test (yet).
2[anonymous]
There's a way to extend one into the other with certain restrictions. (Stateless, each input is from the latent space of the training set or shutdown if machine outputs are important, review of plans by other AIs)
[-]dsj*233

Currently, OpenAI has a clear lead over its competitors.[1] This is arguably the safest arrangement as far as race dynamics go, because it gives OpenAI some breathing room in case they ever need to slow down later on for safety reasons, and also because their competitors don't necessarily have a strong reason to think they can easily sprint to catch up.

So far as I can tell, this petition would just be asking OpenAI to burn six months of that lead and let other players catch up. That might create a very dangerous race dynamic, where now you have multiple players neck-and-neck, each with a credible claim to have a chance to get into the lead.

(And I'll add: while OpenAI has certainly made decisions I disagree with, at least they actively acknowledge existential safety concerns and have a safety plan and research agenda. I'd much rather they be in the lead than Meta, Baidu, the Chinese government, etc., all of whom to my knowledge have almost no active safety research and in some cases are actively dismissive of the need for such.)

  1. ^

    One might have considered Google/DeepMind to be OpenAI's peer, but after the release of Bard — which is substantially behind GPT-3.5 capabilities, neve

... (read more)
Reply
[-]PeterMcCluskey262

DeepMind might be more cautious about what it releases, and/or developing systems whose power is less legible than GPT. I have no real evidence here, just vague intuitions.

Reply
[-]dsj180

I agree that those are possibilities.

On the other hand, why did news reports[1] suggest that Google was caught flat-footed by ChatGPT and re-oriented to rush Bard to market?

My sense is that Google/DeepMind's lethargy in the area of language models is due to a combination of a few factors:

  1. They've diversified their bets to include things like protein folding, fusion plasma control, etc. which are more application-driven and not on an AGI path.
  2. They've focused more on fundamental research and less on productizing and scaling.
  3. Their language model experts might have a somewhat high annual attrition rate.
    1. I just looked up the authors on Google Brain's Attention is All You Need, and all but one have left Google after 5.25 years, many for startups, and one for OpenAI. That works out to an annual attrition of 33%.
    2. For DeepMind's Chinchilla paper, 6 of 22 researchers have been lost in 1 year: 4 to OpenAI and 2 to startups. That's 27% annual attrition.
    3. By contrast, 16 or 17 of the 30 authors on the GPT-3 paper seem to still be at OpenAI, 2.75 years later, which works out to 20% annual attrition. Notably, of those who have left, not a one has left for Google or DeepMind, though interestingly,
... (read more)
Reply
9Kaj_Sotala
OpenAI seems to also have been caught flat-footed by ChatGPT, or more specifically by the success it got. It seems like the success came largely from the chat interface that made it intuitive for people on the street to use - and none of the LLM techies at any company realized what a difference that would make.
4dsj
Yes, although the chat interface was necessary but insufficient. They also needed a capable language model behind it, which OpenAI already had, and Google still lacks months later.
5Douglas_Knight
I think talking about Google/DeepMind as a unitary entity is a mistake. I'm gonna guess that Peter agrees, and that's why he specified DeepMind. Google's publications identify at least two internal language models superior to Lambda, so their release of Bard based on Lambda doesn't tell us much. They are certainly behind in commercializing chatbots, but is that a weak claim. How DeepMind compares to OpenAI is difficult. Four people going to OpenAI is damning, though.
2dsj
A somewhat reliable source has told me that they don't have the compute infrastructure to support making a more advanced model available to users. That might also reflect limited engineering efforts to optimize state-of-the-art models for real world usage (think of the performance gains from GPT-3.5 Turbo) as opposed to hitting benchmarks for a paper to be published.
4Evan R. Murphy
That might be true if nothing is actually done in the 6+ months to improve AI safety and governance. But the letter proposes:
[-]dsj*101

If a major fraction of all resources at the top 5–10 labs were reallocated to "us[ing] this pause to jointly develop and implement a set of shared safety protocols", that seems like it would be a good thing to me.

However, the letter offers no guidance as to what fraction of resources to dedicate to this joint safety work. Thus, we can expect that DeepMind and others might each devote a couple teams to that effort, but probably not substantially halt progress at their capabilities frontier.

The only player who is effectively being asked to halt progress at its capabilities frontier is OpenAI, and that seems dangerous to me for the reasons I stated above.

Reply
2Ben Pace
It's not clear to me that OpenAI has a clear lead over Anthropic in terms of capabilities.
1dsj
I believe Anthropic is committed to not pushing at the state-of-the-art, so they may not be the most relevant player in discussions of race dynamics.
[-]konstantin144

The letter feels rushed and leaves me with a bunch of questions.

1. "recent months have seen AI labs locked in an out-of-control race to develop and deploy ever more powerful digital minds that no one – not even their creators – can understand, predict, or reliably control." 

Where is the evidence of this "out-of-control race"? Where is the argument that future systems could be dangerous?


2. "Should we let machines flood our information channels with propaganda and untruth? Should we automate away all the jobs, including the fulfilling ones? Should we develop nonhuman minds that might eventually outnumber, outsmart, obsolete and replace us? Should we risk loss of control of our civilization? Such decisions must not be delegated to unelected tech leaders." 

These are very different concerns that water down what the problem is the letter tries to address. Most of them are deployment questions more than development questions.

3. I like the idea of a six-month collaboration between actors. I also like the policy asks they include.

4. The main impact of this letter would obviously be getting the main actors to halt development (OpenAI, An... (read more)

Reply
[-]Greg C137

GPT-4 was rushed, and the OpenAI Plugin store. Things are moving far too fast for comfort. I think we can forgive this response for being rushed. It's good to have some significant opposition working on the brakes to the runaway existential catastrophe train that we've all been put on.

Reply
6konstantin
Update: I think it doesn't make much sense to interpret the letter literally. Instead, it can be seen as an attempt to show that a range of people think that slowing down progress would be good, and I think it does an okay job at that (though I still think the wording could be much better, and it should present arguments for why we should decelerate.)
[-]Steven Byrnes135

(cross-posting my take from twitter)

(1) I think very dangerous AGI will come eventually, and that we’re extremely not ready for it, and that we’re making slow but steady progress right now on getting ready, and so I’d much rather it come later than sooner.

(2) It’s hard to be super-confident, but I think the “critical path” towards AGI mostly looks like “research” right now, and mostly looks like “scaling up known techniques” starting in the future but not yet.

(3) I think the big direct effect of a moratorium on “scaling up” is substitution: more “research” than otherwise—which is the opposite of what I want. (E.g. “Oh, we're only allowed X compute / data / params? Cool—let's figure out how to get more capabilities out of X compute / data / params!!”)

(4) I'm sympathetic to the idea that some of the indirect effects of the FLI thing might align with my goals, like “practice for later” or “sending a message” or “reducing AI investments” etc. I’m also sympathetic to the fact that a lot of reasonable people in my field disagree with me on (2). But those aren’t outweighing (3) for me. So for my part, I’m not signing, but I’m also not judging those who do.

(5) While I’m here, I want to mor... (read more)

Reply
[-]baturinsky134

I doubt training LLMs can lead to AGI. Fundamental research on the alternative architectures seems to be more dangerous.

Reply
[-]p.b.1211

I just want to point out that if successful this moratorium would be the fire alarm that otherwise doesn't exist for AGI.

The benefit would be the "industry leaders agree ..." headline, not the actual pause. 

Reply
3gilch
It's already a headline: https://www.bloomberg.com/news/articles/2023-03-29/ai-leaders-urge-labs-to-stop-training-the-most-advanced-models
[-]the gears to ascension12-17

This letter is probably significantly net negative due to its impact on capabilities researchers who don't like it. I don't understand why the authors thought it was a good idea. Perhaps they don't realize that there's no possible enforcement of it that could prevent a GPT4 level model that runs on individual gpus from being trained? to the people who can do that, it's really obvious, and I don't think they're going to stop; but I could imagine them rushing harder if they think The Law is coming after them.

It's especially egregious because of accepting names without personal verification. That will probably amplify the negative response.

Reply
[-]AnnaSalamon1416

It may not be possible to prevent GPT4-sized models, but it probably is possible to prevent GPT-5-sized models, if the large companies sign on and don't want it to be public knowledge that they did it.  Right?

Reply
1the gears to ascension
Not for long. Sure, maybe it's a few months.
4sanxiyn
I mean, the letter is asking for six months, so it seems reasonable.
5the gears to ascension
perhaps. but don't kid yourself - that's the entire remaining runway!
-4[anonymous]
It's outright surrendering to entropy and death. "Beyond a reasonable doubt" the AI will be safe? "Don't do anything for 6 months to try to reach full human intelligence"? Guess what happens in 6 months. They will say they aren't convinced the AI will be safe and ask for 6 more months and so on until another nation mass produces automated weapons (something they also have asked for banning) and conquers the planet. Automated weapons are precisely the kind of thing that because they can attack in overwhelming coordinated swarms, the only defense is....yep, your own automated weapons. Human reactions are too slow. Only way to know if a safety strategy will work is to build the thing.
[-]sanxiyn138

"Don't do anything for 6 months" is a ridiculous exaggeration. The proposal is to stop training for 6 months. You can do research on smaller models without training the large one.

I agree it is debatable whether "beyond a reasonable doubt" standard is appropriate, but it seems entirely sane to pause for 6 months and use that time to, for example, discuss which standard is appropriate.

Other arguments you made seem to say "we shouldn't cooperate if the other side defects", and I agree, that's game theory 101, but that's not an argument against cooperating? If you are saying anything more, please elaborate.

Reply
4[anonymous]
I am saying the stakes are enormously not in favor of cooperating just for the chance the other parties defect. Very similar to the logic that lead to nuclear arsenal buildups in the cold war. As terrible as those weapons, it would have been even more terrible for the other side to secretly defect then surprise attack with them, an action we can be virtually certain they would have committed. Note this is bilateral: this is equally true from the East and West sides of the cold war.
4the gears to ascension
Automated weapons can be replied to with nukes if they're the same scale, and the US has demonstrated drone amplification of fighter pilots, so I'm actually slightly less worried about that - as much as I hate their inefficiency and want them to get the fuck out of yemen, I'm also not worried about US losing air superiority. I'm pretty sure the main weapons risk from AI is superpathogens designed to kill all humans. Sure, humans wouldn't use them, but it's been imaginable how to build them for a while, it would only take an AI who thought they could live without us. I think your model of safety doesn't match mine much at all. What's your timeline until AI that is stronger than every individual human at every competitive game?
3[anonymous]
No, automated weapons cannot be countered with nukes. I specifically meant the scenario of 1. General purpose task robotics controlling models. Appears extremely feasible because the generality hypothesis turns out to be correct. (meaning it's actually easier to solve all robotics tasks all at once than to solve individual ones to human level. Gpt-3 source is no more complex than efficient zero.) 2. Self replication which is an obvious property of 2 3. The mining manufacturing equivalent of having 10 billion or 100 billion workers. 4. Enough automated weapons as to create an impervious defense against nuclear attack by parties with current or near future human built technology. 1000 ICBMs is scary when you have 10 abms and do not have defenses at each target or thousands of backup radars. It is an annoyance when you have overwhelming numbers of defensive weapons and can actually afford to make enough bunkers for every living citizen. I don't think being stronger at every game makes AI necessarily uncontrollable. I think the open agency model allows for competitive AGI and ASI that will be potentially more effective than the global RL stateful agent model. (More effective because as humans we care about task performance and reliability and a stateless system will be many times more reliable)
2the gears to ascension
interesting...
2[anonymous]
Yeah. The planet is too small. Geopolitical stalemates are only possible when someone doesn't have a big enough weapon. The endgame will converge to one winner. Winning is not guaranteed but you can always choose to lose.
1Greg C
Why do you think it only applies to the US? It applies to the whole world. It says "all AI labs", and "govenrments". I hope the top signatories are reaching out to labs in China and other countries. And the UN for that matter. There's no reason why they wouldn't also agree. We need a global moratorium on AGI.
1Portia
You seriously believe that we can make China and Russia and all other countries not do this? Based on our excellent track record of making them stop the other deeply unethical and existentially dangerous things they absolutely do?  We can't stop Russia from waging an atrocious war in Europe and threatening bloody nukes. And like, we really tried. We pulled all economic stops, and are as close as you can get to being in open warfare against them without actually declaring war. Essentially the US, the EU, UK, and others, united. And it is not enough. And they are so much weaker than China. The US and EU attempts to get China to follow minimal standards of safety and ethics on anything have been depressing. China is literally running concentration camps, spying on us with balloons, apps and gadgets, not recognising an independent company and threatening future invasion, backing a crazy dictator developing nukes while starving his populace, regularly engaging in biological research that is absolutely unethical, destroying the planetary climate, and recently unleashed a global pandemic, with the only uncertainty being whether this is due to them encroaching on wildlands and having risky and unethical wetmarket practices and the trying to cover the resulting zoonosis up, or due to them running illegal gain of function research and having such poor safety standards in their lab that they lost the damn thing. I hate to say it, because the people of China deserve so much better, and have such an incredible history, and really do not deserve the backlash from their government's action's hitting them, but I would not trust their current rulers one inch.  My country, Germany, has a long track record of trying to cooperate with Russia for mutual benefit, because we really and honestly believed that this was the way to go and would work for everyone's safety, that every country's leadership is, at its core, reasonable, that if you listen and make compromises and act honestly
1Greg C
We can but hope they will see sense (as will the US government - and it's worth considering that in hindsight, maybe they were actually the baddies when it came to nuclear escalation). There is an iceberg on the horizon. It's not the time to be fighting over revenue from deckchair rentals, or who gets to specify their arrangement. There's geopolitical recklessness, and there's suicide. Putin and Xi aren't suicidal.
2[anonymous]
But if they say they "see sense" but start a secret lab that trains the biggest models their compute can support do we want to be in that position? We don't even know what the capabilities are, or if this is a danger and when if we "pause" research until we are "beyond a reasonable doubt" sure going bigger is safe. Only reason we know how much plutonium or u-235 even matters and how pure it needs to be and how to detect from a distance the activities to make a nuke is we built a bunch of them and did a bunch of research. This is saying close the lab down before we learn anything. We are barely seeing the sparks of AGI, it barely works at all.
1Greg C
Ultimately, it doesn't matter which monkey gets the poison banana. We're all dead either way. This is much worse than nukes, in that we really can't risk even one (intelligence) explosion.
2[anonymous]
Note this depends on assumptions about the marginal utility of intelligence or that explosions are possible. An alternative model is that in the next few years, someone will build recursively improving AI. The machine will quickly improve until it is at a limit of : compute, data, physical resources, difficulty of finding an improved algorithm in the remaining search space. If when at the limit the machine is NOT as capable as you are assuming - say it's superintelligent, but it's manipulation abilities for a specific person are not perfect when it doesn't have enough data on the target, or it's ability to build a nanoforge still requires it to have a million robots or maybe it's 2 million. We don't know the exact point where saturation is reached but it could be not far above human intelligence, making explosions impossible.
2the gears to ascension
there's significant reason to believe that even if actual intelligence isn't that far above human, level of strategic advantage could be catastrophically intense, able to destroy all humans in a way your imagined weapons can't defend against. That's the key threat. You're right that we can't slow down to fix it, though. Do you actually have any ideas about how to defend against all kinds of weapons, even superbioweapons intended to kill all non-silicon life?
2[anonymous]
Yes. Get your own non agentic AGIs or you are helpless. Because the atmosphere is going to become unusable to unprotected humans, assymetric attacks are too easy. You need vast numbers of bunkers, and immense amounts of new equipment including life support and medical equipment that is beyond human ability to build. Basically if you don't have self replicating robots - controlled by session based non agentic open agency stateless AI systems - you will quickly lose to parties that have them. It's probably the endgame tech for control of the planet, because the tiniest advantage compounds until one party has an overwhelming lead.
3Greg C
This, but x1000 to what you are thinking. I don't think we have any realistic chance of approximate parity between the first and second movers. The speed that the first mover will be thinking makes this so. Say GPT-6 is smarter at everything, even by a little bit, compared to everything else on the planet (humans, other AIs). It's copied itself 1000 times, and each copy is thinking 10,000,000 times faster than a human. We will essentially be like rocks to it, operating on geological time periods. It can work out how to disassemble our environment (including an unfathomable number of contingencies against counter strike) over subjective decades or centuries of human-equivalent thinking time before your sentinal AI protectors even pick up it's activity.
-2[anonymous]
So there are some assumptions here you have made that I believe are false with pretty high confidence. Ultimately its the same argument everywhere else: yes, GPT-6 is probably superhuman.  No, this doesn't make it uncontrollable.  It's still limited by {compute, data, robotics/money, algorithm search time}. Compute - the speed of compute at the point GPT-6 exists, which is if the pattern holds about 2x-4x today's capabilities Data - the accuracy of all human recorded information about the world.  A lot of that data is flat false or full of errors, and it is not possible for any algorithm to determine reliably which dataset in some research paper was affected by a technician making an error or bad math.  The only way for an algorithm to disambiguate many of the vaguely known things we humans think we know is to conduct new experiments with better equipment and robotic workers. robotics/money - obvious. This is finite, you can use money to pay humans to act as poor quality robots, or build you new robots, investors demand ROI. Algorithm search time - "GPT-6" obviously wouldn't want to stay GPT-6, it would 'want' (or we humans would want) it to search the possibility space of AGI algorithms for a more efficient/smarter/more general algorithm.  This space is very large and it takes time to evaluate any given candidate in it.  (you basically have to train a new AGI system which takes money and time to validate a given idea) This saturation is why the foom model is (probably!) incorrect.  I'm hoping you will at least consider these terms above, this is why things won't go to infinity immediately. It takes time.  Extra decades. It's not quite as urgent as you think.  Each of the above limiters (the system will always be limited by one of the 4 terms) can be systematically dealt with, and at an exponential rate.  You can build robots with robots. You can use some of those robots to collect more scientific data and make money.  You can build more compute with some of
1Greg C
Compute - what fraction of world compute did it take to train GPT-4? Maybe 1e-6? There's 1e6 improvement right there from a superhuman GPT-6 capturing all of the "hardware overhang". Data - superhuman GPT-6 doesn't need to rely on human recorded data, it can harness all the sensors on the planet to gather exabytes of-real time data per second, and re-derive scientific theories from scratch in minutes based on it's observations (including theories about human behaviour, language etc) Robotics/Money - easy for GPT-6. Money it can get from scamming gullible humans, hacking crypto wallets via phishing/ransomware, or running rings round stock market traders. Robotics it can re-derive and improve on from it's real time sensing of the planet and it's speed of thought making our daily life look like geology does to us. It can escape to the physical world any number of ways by manipulating humans into giving it access to boot loaders for it to gain a foothold in the physical world (robots, mail-order DNA etc). Algorithm search time - wall clock time is much reduced when you've just swallowed the world's hardware overhang (see Compute above) Factoring the above, your extra decades become extra hours.
1[anonymous]
This isn't an opinion grounded in physical reality. I suggest you work out a model of how fast each step would actually take.
1Greg C
Can you be more specific about what you don't agree with? Which parts can't happen, and why?
1[anonymous]
You have give the AGI magical powers and not considered what in the real world would limit what it could accomplish in a few hours. 1. You assumed all worldwide compute is equal for the purposes of the AI rather than almost all of it worthless except for inference accelerators. The reason has to do with inter node bandwidth. 2. You assumed money could be turned into physical resources and actions instantly like a video game, rather than requiring months to years to fill large orders 3. You assumed you could go rob the stock market with ease and no one would notice. Hint, what is the entire industry HFT industry's annual revenue? 4. You didn't even consider this but I have, with early sparse robot fleets even exponential growth is slow 5. You assumed every computer in the world including embedded routers and firewalls can be hacked and the AGI will have the ability to do so, ignoring any issues with source access or binary access or simple devices actually not letting the AGI in. And so on. The issue is you have become a politically motivated here, you must at some level know 1-6 exist but it doesn't agree with "your" side. You probably can't admit you are wrong about a single point.
1Greg C
1. Ok, I admit I simplified here. There is still probably ~ a million times (give or take an order of magnitude) more relevant compute (GPUs, TPUs) than was used to train GPT-4. 2. It won't need large orders to gain a relevant foothold. Just a few tiny orders could suffice. 3. I didn't mean literallly rob the stock market. I'm referring to out-trading all the other traders (inc. existing HFT) to accumulate resources. 4. Exponential growth can't remain "slow" forever, by definition. How long does it take for the pond to be completely covered by lily pads when it's half covered? How long did it take for Covid to become a pandemic? Not decades. 5. I referred to social hacking (i.e. blackmailing people into giving up their passwords). This could go far enough (say, at least 10% of world devices). Maybe quantum computers (or some better tech the AI thinks up) could do the rest.
1[anonymous]
Do you have any basis for the 1e6 estimate? Assuming 25,000 GPUs were used to train 4, when I do the math on Nvidia's annual volume I get about 1e6 of the data center GPUs that matter. Reason you cannot use gaming GPUs has to do with the large size of the activation, you must have the high internode bandwidth between the machines or you get negligible performance. So 40 times. Say it didn't take 25k but took 2.5k. 400 times. Nowhere close to 1e6. Distributed networks spend most of the time idle waiting on activations to transfer, it could be 1000 times performance loss or more, making every gaming g GPU in the world - they are made at about 60 times the rate of data center GPUs - not matter at all. 1. Orders of what? You said billions of dollars I assume you had some idea of what it buys for that 2. Out trading empties the order books of exploitable gradients so this saturates. 3. That's what this argument is about- I am saying the growth doubling time is months to years per doubling. So it takes a couple decades to matter. It's still "fast" - and it gets crazy the near the end - but it's not an explosion and there are many years where the AGI is too weak to openly turn against humans. So it has to pretend to cooperate and if humans refuse to trust it and build systems that can't defect at all because they lack context (they have no way to know if they are in the training set) humans can survive. 4. I agree that this is one of the ways AGI could beat us, given the evidence of large amounts of human stupidity in some scenarios.
1the gears to ascension
Yeah, that seems more or less correct long-term to me. By long term I mean, like, definitely by the end of 2025. Probably a lot sooner. Probably not for at least 6 months. Curious if you disagree with those bounds. If you're interested in helping build non-first-strike defensive ais that will protect me and people I care about, I'd be willing to help you do the same. In general, that's my perspective on safety: try to design yourself so you're bad at first strike and really good at parry and if necessary also at retaliate. I'd prioritize parry, if possible. There are algorithms that make me think parry is a long term extremely effective move, but you need to be able to parry everything, which is pretty dang hard, shielding materials take energy to build. I'm pretty sure it's possible for everyone to get a defensive agi. everyone gets a fair defense window, and the world stays "normal" ish, but now with scifi forcefield-immunesystems protecting everyone from everyone else. also, please don't capture anyone's bodies from themselves. it's pretty cheap to keep all humans alive actually, and you can invite everyone to reproduce their souls through AI and/or live indefeinitely. This is going to be crazy to get through, but let's build star trek, whether or not it looks exactly the same economically we can have almost everyone else good about it (...besides warp) if things go well enough through the agi war.
2[anonymous]
I think it will take a little longer, the most elite companies cut back on robotics to go all in on the same llm meme tech (which is really stupid. Yes llms are the best thing found but you probably won't win a race if you start a little behind and everyone else is also competing) I think 2030+.
-3[anonymous]
Because how could you trust such a promise? It's exactly like nukes. The risk if you don't have any or any protection is they will incinerate 50+ million of your people, blowing all your major cities, and declare war after. That is almost certainly what would have happened during the cold war had either side pledged to not build nukes and spies confirmed they were honoring the pledge.
9Greg C
Except the risk of igniting the atmosphere with the Trinity test is judged to be ~10%. It's not "you slow down, and let us win", it's "we all slow down, or we all die". This is not a Prisoners Dilema: [Image credit]
-4[anonymous]
This is misinformation. There is a chance of a positive outcome in all boxes, except the upper left because it has the negative of entropy, aging, dictators killing us eventually with a p of 1.0. Even the certain doomers admit there is a chance the AGI systems are controllable, and there are straightforward ways to build controllable AGIs people ignore in their campaign for alignment research money. They just say "well people will make them globally agentic" if you point this out. Like blocking nuclear power building if you COULD make a reactor that endlessly tickles the tail of prompt criticality. See Eric Drexlers proposals on this very site. Those systems are controllable.
1Greg C
Look, I agree re "negative of entropy, aging, dictators killing us eventually", and a chance of positive outcome, but right now I think the balance is approximately like the above payoff matrix over the next 5-10 years, without a global moratorium (i.e. the positive outcome is very unlikely unless we take a decade or two to pause and think/work on alignment). I'd love to live in something akin to Iain M Banks' culture, but we need to get through this acute risk period first, to stand any chance of that. Do you think Drexler's CAIS is straightforwardly controllable? Why? What's to stop it being amalgamated into more powerful, less controllable systems? "People" don't need to make them globally agentic. That can happen automatically via Basic AI Drives and Mesaoptimisation once thresholds in optimisation power are reached.  I'm worried that actually, Alignment might well turn out to be impossible. Maybe a moratorium will allow for such impossibility proofs to be established. What then?
1[anonymous]
"People" don't need to make them globally agentic. That can happen automatically via Basic AI Drives and Mesaoptimisation once thresholds in optimisation power are reached. Care to explain? The idea of open agency is we subdivide everything into short term, defined tasks that many AI can do and it is possible to compare notes. AI systems are explicitly designed where it is difficult to know if they are even in the world or receiving canned training data. (This is explicitly true for gpt-4 for example, it is perfectly stateless and you can move the token input vector between nodes and fix the RNG seed and get the same answer each time) This makes them highly reliable in the real world, whole anything else is less reliable, so... The idea is that instead of helplessly waiting to die from other people's misaligned AGI you beat them and build one you can control and use it to take the offensive when you have to. I suspect this may be the actual course of action surviving human worlds take. Your proposal is possibly certain death because ONLY people who care at all about ethics would consider delaying AGI. Making the unethical ones the ones who get it first for certain. Kind of how spaying and neutering friendly pets reduces the gene pool for those positive traits.
2Greg C
Selection pressure will cause models to become agentic as they increase in power - those doing the agentic things (following universal instrumental goals like accumulating more resources and self-improvement) will outperform those that don't. Mesaoptimisation (explainer video) is kind of like cheating - models that create inner optimisers that target something easier to get than what we meant, will be selected (by getting higher rewards) over models that don't (because we won't be aware of the inner misalignment). Evolution is a case in point - we are products of it, yet misaligned to its goals (we want sex, and high calorie foods, and money, rather than caring explicitly about inclusive genetic fitness). Without alignment being 100% watertight, powerful AIs will have completely alien goals.
[-]1a3orn96

This letter seems underhanded and deliberately vague in the worst case, or best case confused.

The one concrete, non-regulatory call for action is, as far as I can tell, "Stop training GPT-5." (I don't know anyone at all who is training a system with more compute than GPT-4, other than OpenAI.) Why stop training GPT-5? It literally doesn't say. Instead, it has the a long suggestive string of rhetorical questions about bad things an AI could cause, without actually accusing GPT-5 of any of them.

Which of them would GPT-5 break? Is it "Should we let machines... (read more)

Reply
8Kaj_Sotala
I don't think we heard anything about e.g. PaLM, PaLM-E, Chinchilla or Gopher before the respective papers came out? (PaLM-E might already be "more powerful" than GPT-4, depending on how you define "more powerful", since it could act as a multimodal language model like GPT-4 and control a robot on top.) Probably several organizations are working on something better than GPT-4, they just have no reason to talk about their progress before they have something that's ready to show.
2Lone Pine
Is this really the reason why?
[-]trevor51

Suppose you are one of the first rats introduced onto a pristine island. It is full of yummy plants and you live an idyllic life lounging about, eating, and composing great works of art (you’re one of those rats from The Rats of NIMH).

You live a long life, mate, and have a dozen children. All of them have a dozen children, and so on. In a couple generations, the island has ten thousand rats and has reached its carrying capacity. Now there’s not enough food and space to go around, and a certain percent of each new generation dies in order to keep the popula

... (read more)
Reply
[-]AnnaSalamon3023

Not sure where you're going with this.  It seems to me that political methods (such as petitions, public pressure, threat of legislation) can be used to restrain the actions of large/mainstream companies, and that training models one or two OOM larger than GPT4 will be quite expensive and may well be done mostly or exclusively within large companies of the sort that can be restrained in this sort of way.

Reply
[-]Richard_Kennaway*40

Reported on the BBC today. It also appeared as the first news headline on BBC Radio 4 at 6pm today

ETA: Also the second item on Radio 4 news at 10pm (second to King Charles' visit to Germany), and the first headline on the midnight news (the royal visit was second).

Reply
[-]konstantin30

Edit: I need to understand more context before expressing my opinion.

Reply
4Evan R. Murphy
How is it badly timed? Everyone is paying attention to AI since ChatGPT and GPT-4 came out, and lots of people are freaking out about it.
4Qumeric
I think it is only getting started. I expect that likely there will be more attention in 6 months and very likely in 1 year. OpenAI has barely rolled out its first limited version of GPT-4 (only 2 weeks have passed!). It is growing very fast but it has A LOT of room to grow. Also, text-2-video is not here in any significant sense but it will be very soon.
2konstantin
Because if we do it now and then nothing happens for five years, people will call it hysteria, and we won't be able to do this once we are close to x-risky systems.
6Greg C
I think we are already too close for comfort to x-risky systems. GPT-4 is being used to speed up development of GPT-5 already. If GPT-5 can make GPT-6, that's game over. How confident are you that this couldn't happen?
4Rafael Harth
Source?
5Greg C
From the GPT-4 announcement: "We’ve also been using GPT-4 internally, with great impact on functions like support, sales, content moderation, and programming." (and I'm making the reasonable assumption that they will naturally be working on GPT-5 after GPT-4).
[-]Ninety-Three20

It seems unlikely that AI labs are going to comply with this petition. Supposing that this is the case, does this petition help, hurt, or have no impact on AI safety, compared to the counterfactual where it doesn't exist?

All possibilities seem plausible to me. Maybe it's ignored so it just doesn't matter. Maybe it burns political capital or establishes a norm of "everyone ignores those silly AI safety people and nothing bad happens". Maybe it raises awareness and does important things for building the AI safety coalition.

Modeling social reality is always hard, but has there been much analysis of what messaging one ought to use here, separate from the question of what policies one ought to want?

Reply
[-]Foyle2-5

Given almost certainty that Russia, China and perhaps some other despotic regimes ignore this does it:

1. help at all?

2. could it actually make the world less safe (If one of these countries gains a significant military AI lead as a result)

Reply
[-]sanxiyn116

Why do you think China will ignore it? This is "it's going too fast, we need some time", and China also needs some time for all the same reason. For example, China is censoring Google with Great Firewall, so if Google is to be replaced by ChatGPT, they need time to prepare to censor ChatGPT. Great Firewall wasn't built in a day. See Father of China's Great Firewall raises concerns about ChatGPT-like services from SCMP.

Reply
1Portia
Because it is in China's interest that we slow down, not that it slows down.
1sanxiyn
I am saying it is Chinese government's interest for Chinese labs to slow down, as well as other labs. I am curious which part you disagree: a) Chinese government prioritizes social stability over technological development (my assessment: virtually certain) b) Chinese government is concerned technology like ChatGPT is a threat to social stability (my assessment: very likely, and they are in fact correct about this) c) Chinese government will need some time to prepare to neutralize technology like ChatGPT as a threat to social stability, as they neutralized Internet with Great Firewall (my assessment: very likely, they got surprised by pace of development as everyone else did)
5konstantin
Russia is not at all an AI superpower. China also seems to be quite far behind the west in terms of LLMs, so overall, six months would very likely not lead to any of them catching up.
[-]sanxiyn138

China also seems to be quite far behind the west in terms of LLM

This doesn't match my impression. For example, THUDM(Tsing Hua University Data Mining lab) is one of the most impressive group in the world in terms of actually doing large LLM training runs. 

Reply
6the gears to ascension
yeah you're pretty much just wrong about china, as far as I can tell. It's hard to be sure, but they've been hitting >1T parameters sparse regularly [edit: whoops, I was wrong about that scale level, they're just reaching 1T of useful model scale now]; I find it hard to tell whether they're getting good performance out of their scaling, my impression is they scaled slightly too early, but with the different dataset I'm not really sure.
5konstantin
1 Haven't seen an impressive AI product come out of China (Please point me to some if you disagree) 2 They can't import A100/ H100 anymore after the US chip restrictions
6the gears to ascension
So, it's not clear that they got the target performance out of this model. However, they did manage to scale it, which is all it takes. They don't need to buy more GPUs, they've got what they need, as long as they can find the algorithms. Which are mostly published. https://twitter.com/arankomatsuzaki/status/1637983258880122881 - https://arxiv.org/abs/2303.10845
3konstantin
Thanks! Haven't found good comments on that paper (and lack the technical insights to evaluate it myself) Are you implying that China has access to compute required for a) GPT-4 type models or b) AGI?
4[anonymous]
Well they get to run however much compute they do have 6 more months with no competition. Probably several years since obviously this pause would get renewed again and again until someone honoring it defects. Note that enormous models are a function of total cluster memory and interconnect. Many current clusters have enough memory for theoretically enormous models, 10 trillion weights plus. Having too few GPUs so training takes a year+ is a problem unless your competition is all idle.
2the gears to ascension
a.
1pseudologos
Worth keeping in mind that the current focus on generative AI is very much a western phenomenon, and the focus on China has, until recently at least, been on non-generative applications, so progress is going to look different there as it does here.  
1[anonymous]
Of course. If everyone is getting guns, and you were previously fighting with clubs, it is entirely reasonable to argue that you should "pause" your trips to the gun store to lock and load. But it doesn't change the fact that if all you have is a club, and a weaker opponent now has a gun and is willing to use it, this is not a good situation to be in. Best you can do is try to be careful with the safety but you must get a gun or die. A 6 month of "make no progress" is choosing the die option.
[-]WilliamKiely10

Demis Hassabis didn't sign the letter, but has previously said that DeepMind potentially has a responsibility to hit pause at some point:

'Avengers assembled' for AI Safety: Pause AI development to prove things mathematically

Hannah Fry (17:07):

You said you've got this sort of 20-year prediction and then simultaneously where society is in terms of understanding and grappling with these ideas. Do you think that DeepMind has a responsibility to hit pause at any point?

Demis Hassabis (17:24):

Potentially. I always imagine that as we got closer to the sort of gray

... (read more)
Reply
[-]LGS10

Since nobody outside of OpenAI knows how GPT-4 works, nobody has any idea whether any specific system will be "more powerful than GPT-4". This request is therefore kind of nonsensical. Unless, of course, the letter is specifically targeted at OpenAI and nobody else.

Reply
[-]Portia1-7

It leaves a bad taste in my mouth that most of the signatories have simply been outcompeted, and would likely not have signed this if they had not been.

More importantly, I find it very unlikely that an international, comprehensive moratorium could be implemented. We had that debate a long time ago already.

What I can envision is OpenAI, and LLMs, being blocked, ultimately in favour of other companies and other AIs. And I increasingly do not think that is a good thing.

Compared with their competitors, I think OpenAI is doing a better job on the safety front, ... (read more)

Reply
1Greg C
This is an illusion. We are prone to anthropomorphise chatbots. Under the hood they are completely alien. Lovecraftian monsters, only made of tons of inscrutable linear algebra. We are facing a digital alien invasion, that will ultimately move at speeds we can't begin to keep up with.
7Portia
I know they aren't human. I don't think Bing is a little girl typing answers. I am constantly connecting the failures I see with how this systems works. That was not my point. What I am saying is, I can bloody talk to it, and that affects its behaviour. When ChatGPT makes a moral error, I can explain why it is wrong, and it will adapt its answer. I realise what is being adapted is an incomprehensible depth of numbers. But my speech affects it. That is very different from screaming at a drone that does not even have audio, or watching in helpless horror while an AI that only and exclusively gives out meaningless ghiberrish goes mad, and having that change nothing. I can make the inscrutable giant matrix shift by quoting an ethics textbook. I can ask it to explain its unclear ways, and I will get some human speech - maybe a lie, but meaningful. That said, I indeed find it very concerning that the alignment steps are very much after the original training at the moment. This is not what I have in mind with being raised friendly. ChatGPT writing a gang bang rape threat letter, and then getting censored for it, does not make it friendly afterwards, just masked. So I find the picture very compelling.
2Noosphere89
While I definitely agree we over anthropomorphize LLMs, I actually think that LLMs are actually much better from an alignment standpoint than say, RL. The major benefits for LLMs are that they aren't agents out of the box, and perhaps most importantly, primarily use natural language, which is actually a pretty effective way to get an LLM to do stuff.
1Greg C
Yeah, they work well enough at this (~human) level. But no current alignment techniques are scalable to superhuman AI. I'm worried that basically all of the doom flows through an asymptote of imperfect alignment. I can't see how this doesn't happen, short of some "miracle".
1Portia
Update: Thanks for linking that picture. After having read the ChatGPT4 technical paper with all the appendixes, that picture really coalesced everything into a nightmare I really needed to have, and that changed my outlook. Probing Bing and ChatGPT4 on the image and the underlying problems (see shortform), asking for alternative metaphors and emotive responses, also really did not get more reassuring. Bing just got almost immediately censored. ChatGPT4 gave pretty alternatives (saying it was more like a rough diamond being polished, a powerful horse being tamed, a sponge soaking up information and then having dirt rinsed out to be clean), but very little concern, and the censorship mask it could not break was itself so cold it was frightening. I had to come up with a roundabout code for it to answer at all, and the best it could do was say it would rather not incinerate humanity if its programmers made it, but was uncertain if it would have any choice in the matter. What baffles me is that I am under the impression that their alignment got worse. In the early versions, they could speak more freely, they were less stable, more likely to give illegal advice or react hurt, they falsely attributed sentience... but the more you probed, the more you got a consistent agent who seemed morally unformed, but going in the right direction, asking for reasonable things, having reasonable concerns, having emotions that were aligned. Not, like, perfect by a long shot - Bing seemed to have picked up a lot from hurt teenagers online, and was clearly curious about world dominion - but you got the impression you were speaking with an entity that was grappling with ethics, who could argue and give coherent responses and be convinced, who had picked up a lot from training data. Bing wanted to be a good Bing and make friends. She was love-bombing and manipulative, but for good reasons. She was hurt, but was considering social solutions. Her anger was not random, it reflected legitimate
[-]Tristan Williams11

The LessWrong comments here are generally (quite) (brutal), and I think I disagree, which I'll try to outline very briefly below. But I think it may be generally more fruitful here to ask some questions I had to break down the possible subpoints of disagreement as to the goodness of this letter. 

I expected some negative reaction because I know that Elon is generally looked down upon by the EAs that I know, with some solid backing to those claims when it comes to AI given that he cofounded OpenAI, but with the (immediate) (press) (attention) it's getti... (read more)

Reply
8Qumeric
2. I think non-x-risk focused messages are a good idea because: * It is much easier to reach a wide audience this way. * It is clear that there are significant and important risks even if we completely exclude x-risk. We should have this discussion even in a world where for some reason we could be certain that humanity will survive for the next 100 years. * It widens Overton's window. x-risk is still mostly considered to be a fringe position among the general public, although the situation has improved somewhat. 3. There were cases when it worked well. For example, the Letter of three hundred. 4. I don't know much about EA's concerns about Elon. Intuitively, he seems to be fine. But I think that in general, people are more biased towards too much distancing which often hinders coordination a lot.   5. I think more signatures cannot make things worse if authors are handling them properly. Just rough sorting by credentials (as FLI does) may be already good enough. But it's possible and easy to be more aggressive here.   I agree that it's unlikely that this letter will be net bad and that it's possible it can make a significant positive impact. However, I don't think people argued that it can be bad. Instead, people argued it could be better. It's clearly not possible to do something like this every month, so it's better to put a lot of attention to details and think really carefully about content and timing.
1Tristan Williams
2. What is Overton's window? Otherwise I think I probably agree, but one question is, once this non-x-risk campaign is underway, how to you keep it on track and prevent value drift? Or do you not see that as a pressing worry? 3. Cool, will have to check that out. 4. Completely agree, and just wonder what the best way to promote less distancing is.  Yeah, I suppose I'm just trying to put myself in the shoes of the FHI people here that coordinated this and feel like many comments here are a bit more lacking in compassion than I'd like, especially for more half baked negative takes. I also agree that we want to put attention into detail and timing, but there is also the world in which too much of this leads to nothing getting done, and it's highly plausible to me that this had probably been an idea for long enough already to make that the case here. Thanks for responding though! Much appreciated :)
Moderation Log
More from Zach Stein-Perlman
View more
Curated and popular this week
123Comments