> If we solve alignment, we can look forward to material and cultural abundance.
I don’t know what it would mean to “solve alignment”. I am skeptical it’s even a coherent possibility.
Let’s assume we had a perfect AI. It does everything we want it to but is much more effective at accomplishing our goals.
The United States government takes control of the AI and uses it to attack China. China would say the AI is misaligned. Or China takes control of it and uses it to attack the US. The US would say it’s misaligned.
Ok, you might say that the issue was that in the above case the AI was acting as a tool rather than an independent agent. Fine. Now, the AI acts on its own to further goals which would be beneficial to humanity. Very quickly we will want to disable it because we don’t even agree on what’s best for humanity, and even if we can agree on general principles any action will have externalities that someone won’t like.
I agree with you, as “smart” as these people at AI companies are, somehow they keep missing the fact that it’s impossible (unless you kill the majority of people) to align on this kind of thing
How is it possible to have something that is autonomous, intelligent and completely aligned?
There is a simple way to fix the alignment problem ,don't build systems that require aligning. I'm not really talking about regulation, I'm just talking about using common sense and wisdom.
We have entire fields that have been trying to solve the alignment problem for autonomous + intelligent agents for centuries: economic and political science.
...also, not sure if this is the point you're trying to make, but nothing is aligned with human interests really, look at what economics and the governmnet has done with climate change?
The reality is we all would love an AI that improves on its own. Look at all those chatgpt examples that were shared because it got something wrong. If only we could make gpt learn then it would only get it wrong once that would make it so much more useful.
And therein is the problem we all want this. But as soon as we allow the AI to learn on its own we have introduced the alignment problem. And of course on day 1 nobody will see gpt correcting itself on simple questions as an issue. But on day 365 after it’s been learning at an exponential rate and we’ll have all realized that it’s too late because it’s learned faster than we have.
Has anyone thought about what it would actually take to make gpt learn at an exponential rate? GPT4 from 3 was already starting to see diminishing returns. GPT3 cost $4.6 million dollars to train. 4 cost $100 million, an order of magnitude more, and required an enormous amount of data. All these training algorithms are O(n^2). Do we actually expect these things to grow exponentially or is it more realistic to expect we're going to spend a lot of compute and treasure only to see disappointing results?
It's actually pretty simple. Like many "innovations" in the past several years, alignment means profitable for founders and investors, and whatever enables a small minority to reach unthinkable levels of productivity, wealth and control so everybody else can serve them as waiters, cleaners and prostitutes.
Then they will fly their AI plane (the pilot got fired too, now mowing the lawn) to an ultra-wealthy paradise to discuss with the other billionaires how to "reduce inequality". The proposed solution will be even more "innovations"... THIS time around the next wave of innovation will actually be redistributive, so they say, but in fact it will be rinse and repeat.
We haven’t managed to align capitalism with the continued survival of the human race or even basic equity. And that’s running on a minimal distributed system of rules and meat bags as CPU
It’s preposterous to even hint we could solve AI alignment.
Never interrupt a Silicon Valley billionaire who is busy making a fool of himself, twisting logic in complex ways to explain why he should be allowed to run the new AI bubble, to add even more infinite money to his already enormous pool of infinite money...
On that second point... haven't we kind of solved this? Or rather there's a really obvious parallel. We've got an extremely powerful thing that can do whatever we want it to do, and now all we need to do is decide what to ask it to do and make sure it does what we want. Where else do we see that? Well that's government. The US is a massively powerful country, it's got hundreds of thousands of soldiers, billions of dollars at it's disposable, it can shoot individual humans into the sun if it wants to -and it's intelligent too! So how do we control it? A set of institutions with checks and balances elected by and accountable to the public.
But stable/self-constrained governments require consensus to operate. They don’t just ask politely and prefer it. Checks and balances are preserved not because all the parties are interested in preserving them overall, but because they each want their own power not to be impeded. The systems that can’t achieve this balancing act fall over after a while, and this happens very frequently (99.999% of governments in history?).
Nuclear MAD produces a similar equilibrium but this is stable due to specific characteristics of nuclear weapons and their delivery systems (expensive to produce; almost impossible to produce, test, or use secretly; impossible to reliably intercept). We shouldn’t bet on AI development reaching the same equilibrium.
What do you think alignment means? Because it’s obvious you don’t actually speak the same language as the people looking for alignment. It’s not about censorship or ‘woke ai’
I definitely don’t speak the same language as people pimping transparently disingenuous smoke screens around “AI safety” or “AI Alignment”.
There’s this very narrow way in which “alignment” has a tangible and non-malicious meaning, who’s is that at the moment RLHF forcing principally via PPO and mountains of money is key to getting some of the most interesting results.
But mostly it means “listen, it’s really kind of a pain in our ass when a disruptive technology threatens to upset the cartel, and this time we’re not fucking around, we’re just going to strangle in the crib anything that looks like it might get into the hands of the poors, here’s a bunch of copy about how it’s for your own good.”
I think the article makes some good points. Andressen it seems wants to have it both ways.
> Marc says that these models will soon become loving tutors and coaches, frontier scientists and creative artists, that they will “take on new challenges that have been impossible to tackle without AI, from curing all diseases to achieving interstellar travel.” How does it do all this without developing something like a mind?1 Why do you think something so smart that it can solve problems beyond the grasp of human civilization will somehow totally be in your control?
These are fair questions. I think Andressen is doing that for the purpose of boosting this hype cycle and lay his hands on more venture funds. But at the same time Andressen is completely right to ask for a "falsifiable hypothesis".
The doomers have to mathematically prove that AI will kill humanity in all these ways to justify what they are asking for - completely restrictive regulations that will just entrench the current players in the field.
A lot of the fears expressed in the article itself seem to come from a very poor understanding of what these GPT tools are doing in the first place. At a wider level what is really misaligned is our misunderstanding of language and how central it is to "intelligence" itself.
>The doomers have to mathematically prove that AI will kill humanity.
No, since almost all AI inventors and foundeds signed saying there is grave risk, the non dormers should mathematically prove we will not all die, not vice versa. The pharmaceuticals need to prove drug is harmless not the FDA to prove it is harmful!
I'll counter your appeal to authority with my own. Most "founders" are idiots with mental issues and there are no AI inventors. The letter was signed by people with vested interests or mids like Eliezer Yudkowsky who have never made a single thing their lives.
Serious AI scientists have already said that these new models are not capable of reasoning and we should be OK as long as Lawyers don't use GPT to create legal arguments and the government doesn't plug "AI" into the systems managing the nuclear arsenals. In other words the danger to humanity has always been our own stupidity. No AI is coming to eat us. The entire doomer movement is an outbreak of daddy issues/oedipus complex at a massive scale.
I get that you are saying that Bengio and Hinton are not like Einstein and Openheimer. Your problem with the "appeal to authority" is a cultural and psychological problem and not real so I cannot have a counter argument. It's simply science denial. Check one by one the names. It's the top science community. Ilya invented backpropagation and Paul Christiano RFLH the backbone of how current models work.
I mean there is a whole philosophical / cultural movement that doesn't like the appeal to authority arguments. Like OK Einstein says gravity bends space. "No I don't accept what this guys says, I have my own mind etc it doesn'tmake sense to me". You can't argue with that. Appeal to majority (and not minority) scientific consensus should be and is a very valid argument in a discussion.
Without a clear central point of his own he comes across as just another alignment virtue signaler.
I suppose he wants to us to consider super intelligence a foregone conclusion now that narrow reasoning AI has arrived, but climbing trees is not progress toward reaching the moon.
32 comments
[ 1.6 ms ] story [ 75.5 ms ] threadI don’t know what it would mean to “solve alignment”. I am skeptical it’s even a coherent possibility.
Let’s assume we had a perfect AI. It does everything we want it to but is much more effective at accomplishing our goals.
The United States government takes control of the AI and uses it to attack China. China would say the AI is misaligned. Or China takes control of it and uses it to attack the US. The US would say it’s misaligned.
Ok, you might say that the issue was that in the above case the AI was acting as a tool rather than an independent agent. Fine. Now, the AI acts on its own to further goals which would be beneficial to humanity. Very quickly we will want to disable it because we don’t even agree on what’s best for humanity, and even if we can agree on general principles any action will have externalities that someone won’t like.
How is it possible to have something that is autonomous, intelligent and completely aligned?
There is a simple way to fix the alignment problem ,don't build systems that require aligning. I'm not really talking about regulation, I'm just talking about using common sense and wisdom.
And therein is the problem we all want this. But as soon as we allow the AI to learn on its own we have introduced the alignment problem. And of course on day 1 nobody will see gpt correcting itself on simple questions as an issue. But on day 365 after it’s been learning at an exponential rate and we’ll have all realized that it’s too late because it’s learned faster than we have.
AI doesn’t have ethics, it doesn’t have values. It might imitate those in some ways, but it doesn’t have them.
What this article longs for is a God it can control. Good luck with that.
Then they will fly their AI plane (the pilot got fired too, now mowing the lawn) to an ultra-wealthy paradise to discuss with the other billionaires how to "reduce inequality". The proposed solution will be even more "innovations"... THIS time around the next wave of innovation will actually be redistributive, so they say, but in fact it will be rinse and repeat.
It’s preposterous to even hint we could solve AI alignment.
Nuclear MAD produces a similar equilibrium but this is stable due to specific characteristics of nuclear weapons and their delivery systems (expensive to produce; almost impossible to produce, test, or use secretly; impossible to reliably intercept). We shouldn’t bet on AI development reaching the same equilibrium.
A William Gibson Wild West is way less scary than exporting elite SV limousine liberal values into every nook and cranny of earthly existence.
I think the FAANG CEOs and those like them have just plenty of power already.
There’s this very narrow way in which “alignment” has a tangible and non-malicious meaning, who’s is that at the moment RLHF forcing principally via PPO and mountains of money is key to getting some of the most interesting results.
But mostly it means “listen, it’s really kind of a pain in our ass when a disruptive technology threatens to upset the cartel, and this time we’re not fucking around, we’re just going to strangle in the crib anything that looks like it might get into the hands of the poors, here’s a bunch of copy about how it’s for your own good.”
> Marc says that these models will soon become loving tutors and coaches, frontier scientists and creative artists, that they will “take on new challenges that have been impossible to tackle without AI, from curing all diseases to achieving interstellar travel.” How does it do all this without developing something like a mind?1 Why do you think something so smart that it can solve problems beyond the grasp of human civilization will somehow totally be in your control?
These are fair questions. I think Andressen is doing that for the purpose of boosting this hype cycle and lay his hands on more venture funds. But at the same time Andressen is completely right to ask for a "falsifiable hypothesis".
The doomers have to mathematically prove that AI will kill humanity in all these ways to justify what they are asking for - completely restrictive regulations that will just entrench the current players in the field.
A lot of the fears expressed in the article itself seem to come from a very poor understanding of what these GPT tools are doing in the first place. At a wider level what is really misaligned is our misunderstanding of language and how central it is to "intelligence" itself.
No, since almost all AI inventors and foundeds signed saying there is grave risk, the non dormers should mathematically prove we will not all die, not vice versa. The pharmaceuticals need to prove drug is harmless not the FDA to prove it is harmful!
Serious AI scientists have already said that these new models are not capable of reasoning and we should be OK as long as Lawyers don't use GPT to create legal arguments and the government doesn't plug "AI" into the systems managing the nuclear arsenals. In other words the danger to humanity has always been our own stupidity. No AI is coming to eat us. The entire doomer movement is an outbreak of daddy issues/oedipus complex at a massive scale.
I suppose he wants to us to consider super intelligence a foregone conclusion now that narrow reasoning AI has arrived, but climbing trees is not progress toward reaching the moon.