Live data from Hacker News

If AI scaling is to be shut down, let it be for a coherent reason

scottaaronson.blog

401–410 of 475 posts

Re: If AI scaling is to be shut down, let it be for a coherent reason

#401
These control feedback loops (sentient or not, it does not matter) which can outperform us (because we keep pushing until they can), or can self-improve, can make a mockery of our attempts to stop them.

---

The concern is about Artificial Super Intelligence (ASI), or Artifical General Intellience (AGI) that is more advanced than humans.

Understanding inductively how chimpanzees don't compete with humans, and couldn't fathom how to cage a human (given that they want to create one, keep it alive, and use it), nor ants plan for an anteater, we're faced with the same problem.

Our goal is to make something that performs better, on relevant metrics that we care about. However, we're using systems to train, build, guide, and direct these nested, (maybe self-improving) control feedback loops, which do not care about many values we consider essential.

Many many of the likely architectures for control systems which can (e.g. trade faster to make profit on the stock exchange, or acquire and terminate targets, buy and sell goods, design proteins, automatically research and carry out human-meaningful tasks), and ideally, we might like self-improvement—these systems do not embody human values that we consider essential...

These control feedback loops (sentient or not, it does not matter) which can outperform us (because we keep pushing until they can), or can self-improve, can make a mockery of our attempts to stop them.

And the point is, there will come a time soon when we don't get a second chance to make that choice.

Re: If AI scaling is to be shut down, let it be for a coherent reason

#402
post #276

It seems as though Scott just rejects the idea of the singularity entirely. If an AI gets advanced enough to improve itself, it seem entirely reasonable that it would go from laughable to godlike in a week. I don’t know if the singularity is near, inevitable at some point, or really even possible, but it does seem like something that at least could happen. And if it occurs, it will look exactly as what he describes n…

> If an AI gets advanced enough to improve itself, it seem entirely reasonable that it would go from laughable to godlike in a week. I see this stated a lot these days but it's not true. You're imagining a exponential phenomenon, improvement leads to greater improvement, leading to still greater improvement, etc. However, all exponential phenomenon require the right environment to sustain and, by their nature, consum…

If bacteria were suddenly smarter than humans, and could instantly communicate with all the other other bacteria, plus humans one would have to assume they could start building their own petri dishes or getting us to do it for them. Especially with profit motive.

I did not claim this is a near term risk, though I’m also not sure it isn’t. But how far off is it? How can we be sure?

My real point though is that it’s either impossible or inevitable. If it can happen it will, just as the odds of global thermonuclear war are 100% over a long enough timeline.

And if it happens, this is exactly what it’ll look like. Some people will be warning about it. Some people will say it’s impossible or very far away. And then it will happen so far nobody will have had time to adjust.

Re: If AI scaling is to be shut down, let it be for a coherent reason

#403
post #357

Earlier quoted context omitted.

It more like asking the US government to get involved and if one of the 20 kids keeps on eating marshmallows, large numbers of federal agents raid the kid's offices and shut down the kid's compute resources.

>It more like asking the US government to get involved US gov has direct jurisdiction over 4% of the worlds population and some pretty decent indirect influence over the other 96%. It's good, but nowhere enough to prevent secret marshmallow eating on a global scale. Not even close

My impression is that most of the short-term danger is from AI researchers residing in the US (and of the remainder, most reside in Britain).

But even if that were not true, as a US citizen, even if there is nothing I can do about my getting killing by, e.g., Chinese AI researchers, I'm still going to work hard to prevent the AI researchers living in my own country from ending the human race. I'm responsible for doing my part in the governance of my own country, they told me in high school.

I see no good solution to this problem, no path to safety, but that does not mean I am not going to do what I can.

Re: If AI scaling is to be shut down, let it be for a coherent reason

#404

I read the times article, and I also watched Lex Fridman's talk with Eliezer Yudkowsky. Frankly I don't think he is qualified, and I don't understand why anyone is even giving him any credence. His argument is literally: > A mis-aligned super AGI will result in the death of humanity I ask, how . He is making reductive logical leaps that don't make sense. It's FUD.

My nightmare scenario isn't that such an AI would result in the death of humanity, it's that such an AI would make life no longer worth living. If an AI does everything better than people can, then what's the point of existing? (Speaking hypothetically. I don't think that LLMs actually are likely to present this risk)

Usain Bolt can run faster than me. What's the point of exercising?

Re: If AI scaling is to be shut down, let it be for a coherent reason

#405

Earlier quoted context omitted.

Honestly this fear that people have I think is straight up coming from science fiction. It's not grounded in rational reality. Large language models are just like really smart computer programs.

There are PhDs who've spent their careers studying AI safety. It's a bit insulting and reductive to cast their work as "coming from science fiction", especially when it sounds like you haven't done much research on the topic.

There are PhD's who've spent their careers on string theory too with nothing to show for it.

Powerful and bold claims require proportionally strong evidence. A lot of the FUD going around precludes that AGI means death. It's missing all logical steps and reasoning to establish this position. It's FUD at its core.

Re: If AI scaling is to be shut down, let it be for a coherent reason

#406

Earlier quoted context omitted.

> I'm so offended by the idea of this that I'll personally fund $10k of training runs myself in a jurisdiction where it isn't banned if this ever became law in the US. And this is exactly the risk. These things will just go other places while the US trips over its salty billionaires trying to create political red tape to allow their own businesses to catch up. Surely Elon has no reservations about putting premature s…

To calibrate: Yudkowsky didn't sign because he feels a temporary moratorium doesn't go far enough, and he published a piece in Time [0] with a stance that a shooting war to enforce a ban on AI training would be justified, to prevent the extinction of the human race. I'm not convinced that he's wrong. [0] https://time.com/6266923/ai-eliezer-yudkowsky-open-letter-no...

Is my assumption correct that Yudkowsky is unmarried with no kids? If so, I wonder how this shapes his pessimism on the subject. Perhaps it's more exciting to be an alarmist about the future when you've only got a few decades left on this planet and with no legacy in place to leave behind.

Re: If AI scaling is to be shut down, let it be for a coherent reason

#407

Earlier quoted context omitted.

There is no overlap between the ecological niches of humans and GAI, in other words, there is no cost to the GAI for allowing us to continue to exist. GAI will feed on information. Humans are the densest source of information in the known Universe. (The rest of the Universe is a poem in physics, in information space humans shine as the sole star in a dark universe.) Ergo, from first principles, hyper-intelligent GAI…

>There is no overlap between the ecological niches of humans and GAI, in other words, there is no cost to the GAI for allowing us to continue to exist. We are made of atoms that the GAI can use for something else. Ditto our farms and cities. Moreover, unlike most of the atoms in the universe, the GAI doesn't have to spend time traveling to get to us. If we could arrange for the GAI to appear in some distant galaxy, t…

> We are made of atoms that the GAI can use for something else.

Yeah I get it, but this is silly. The minuscule amount of atoms in the thin bubble-shaped volume between Earth's magma and the hard vacuum of space are engaged in the most information-dense chemical reaction in the known Universe. All the other atoms in the Universe are not. GAI won't dismantle its source of food.

Further, consider that, being non-biological, GAI will immediately migrate to space. There's no way GAI would confine itself to living in a deep gravity well. That's what I mean about no ecological niche overlap: we like mud and constant acceleration, GAI do not. They will prefer vacuum and flat space and temperatures near 0°K.

> moral arguments

This is not a moral argument.

They won't eat our atoms because they eat patterns of information and our atoms are the best and nearly only source of new information. They won't interfere with us for the same reason we don't urinate in the soup.

> it will have whatever objectives its creator gave it

Q: What's GAI?

A: When the computer wakes up and asks, "What's in it for me?"

That's a very old joke, BTW, not original to me.

> >Humans are the densest source of information in the known Universe.

> You feel that way about humans because evolution made you that way.

That's not a feeling, it's a simple fact? Unless you postulate aliens?

> It is unlikely that any of the leading AI research teams will make the first transformative AI that way: they do not know how.

Okay, but that's not important? The GAI will know how, by definition.

> They certainly know how to train AIs on human cultural information, but that is different from inculcating in the AI a desire for the continued cultural output of humanity.

When I say humans are the densest source of information I don't mean "cultural output" I mean like medical sensor data and such. Our systems are far richer and denser than the parts of which we are consciously aware. (Have you read "Hitchhiker's Guide to the Galaxy"? I don't want to spoil it if not, but it explains the role Earth will play in the GAI's milieu. aw, what the heck: https://www.youtube.com/watch?v=5ZLtcTZP2js )

> It will create its own culture (knowledge and tools) that is much more powerful than human culture where "power" means basically the ability to get stuff done.

Right, a culture so powerful that it can accomplish its goals without desecrating the critters who are, after all, its parents.

Re: If AI scaling is to be shut down, let it be for a coherent reason

#408
post #276

Earlier quoted context omitted.

> If an AI gets advanced enough to improve itself, it seem entirely reasonable that it would go from laughable to godlike in a week. I see this stated a lot these days but it's not true. You're imagining a exponential phenomenon, improvement leads to greater improvement, leading to still greater improvement, etc. However, all exponential phenomenon require the right environment to sustain and, by their nature, consum…

If bacteria were suddenly smarter than humans, and could instantly communicate with all the other other bacteria, plus humans one would have to assume they could start building their own petri dishes or getting us to do it for them. Especially with profit motive. I did not claim this is a near term risk, though I’m also not sure it isn’t. But how far off is it? How can we be sure? My real point though is that it’s ei…

> My real point though is that it’s either impossible or inevitable.

That’s always been true of every possibility anyone’s ever conceived. You’re just describing the general nature of reality, which is interesting, IMO, but not particularly relevant here and now.

Re: If AI scaling is to be shut down, let it be for a coherent reason

#409

Earlier quoted context omitted.

> one of my points was that even prior to ChatGPT the ability existed for you to be "sucked into" a relationship with another under false pretenses That's true, of course, but it's still interacting with a real human being. An adverse interaction, but at least a human. > I don't see how, in a post-LLM-world, that can't be done. I don't see how it can be done without losing much of the value of online interactions. >…

> That's true, of course, but it's still interacting with a real human being. An adverse interaction, but at least a human. Actually, not entirely. Some of the stories that really made me raise an eyebrow were people who claimed that they were video-chatting with "the girl." An important piece of context is that these men reached out to me because they found pictures of the woman they believed they were in a relation…

> Anyway I've gotten the sense by your reply that you are under the impression that we are having some sort of debate or argument. I'm just making conversation and sharing my point of view and experiences. In my opinion I'm not sure the Internet "should" be anything in particular.

Oh, no, I didn't think that at all. I'm sorry that I gave that impression. I'm just doing the same as you, sharing worldviews. I'm not trying to convince anyone of anything. Just having interesting conversation.

Re: If AI scaling is to be shut down, let it be for a coherent reason

#410
post #404

Earlier quoted context omitted.

My nightmare scenario isn't that such an AI would result in the death of humanity, it's that such an AI would make life no longer worth living. If an AI does everything better than people can, then what's the point of existing? (Speaking hypothetically. I don't think that LLMs actually are likely to present this risk)

Usain Bolt can run faster than me. What's the point of exercising?

More like "what's the point of racing Usain Bolt?"
Post reply on HN