alt Hacker NewsI don't agree with Yudkowsky, but I think there's certainly a chance that he's right about AI destroying humanity. I just don't think the likelihood of that happening is as high as he thinks it is. But there certainly is a chance.
The problem with trying to stop it is, how? Even if you killed every single AI company leader and every single top AI engineer, it would almost certainly just slow down the rate of progress in the technology, not stop it. The technology is so vital to national security that in the face of such actions, state security forces would just bring development of the tech under their direct protection Manhattan Project-style. Even if you killed literally every single AI engineer on the planet, it's pretty likely that this would just delay the development of the technology by a decade or so instead of actually preventing it.
The technology is pushed forward by a simple psychological logic: every key global actor knows that if they don't build the technology, they will be outcompeted by other actors who do build the technology. No key actor thinks that they have the luxury of not building the technology even if they wanted to not build it. It's very similar to nuclear weapons in that regard. You can talk about nuclear disarmament all you want but at the end of the day, having nuclear weapons is vital to having sovereignty. If you don't have nuclear weapons, you will always be in danger of becoming just the prison bitch of countries that do have them. AI seems that it is growing toward a similar position in the calculus of states' notional security.
I can think of no example in history of the entire world deciding to just forsake the development of a technology because it seemed like it could prove to be too dangerous. The same psychological logic always applies.
Replies
> I can think of no example in history of the entire world deciding to just forsake the development of a technology because it seemed like it could prove to be too dangerous. The same psychological logic always applies.
I wish they did before too.
> I can think of no example in history of the entire world deciding to just forsake the development of a technology because it seemed like it could prove to be too dangerous. The same psychological logic always applies.
Can't you? Haven't many (most?) countries agreed to nuclear disarmament? What about biological weapons? Even anti-personnel mines, I think?
[flagged]
>The technology is pushed forward by a simple psychological logic: every key global actor knows that if they don't build the technology, they will be outcompeted by other actors who do build the technology. No key actor thinks that they have the luxury of not building the technology even if they wanted to not build it.
I don't remember who, but someone made an interesting point about this around the time GPT-4 was released: If the major nuclear powers all understand this, doesn't that make nuclear war more likely the closer any of them get to AGI/ASI? After all, if the other side getting there first guarantees the complete and total defeat of one's own side, a leader may conclude that they don't have anything to lose anymore and launch a nuclear first strike. There are a few arguments for why this would be irrational (e.g. total defeat may, in expectation, be less bad than mutual genocide), but I think it's worth keeping in mind as a possibility.
Humanity agreed, for example, that growing ozone hole is dangerous for everyone, and worked together to ban production of gases that damage ozone layer. See Montreal Protocol International Treaty. It was highly effective. Training powerful AIs isn’t different.
> I don't agree with Yudkowsky, but I think there's certainly a chance that he's right about AI destroying humanity. I just don't think the likelihood of that happening is as high as he thinks it is. But there certainly is a chance.
This is the rhetorical trick that LessWrongers (Yudkowsky's site) have settled on for decades: They have justified everything around the premise that there's a chance, however small, that the world will end. You can't argue that the world ending is a bad thing, so they have their opening for the rest of their arguments, which is that we need to follow their advice to prevent the world maybe ending. They rebut any counterarguments by trying to turn it into a P(doom) debate where we're fighting over how likely this outcome is, but by the time the discussion gets there you've already been forced to accept their argument. Then they push the P(doom) argument aside and try to argue that it doesn't matter how unlikely it is, we have a morally duty to act.