I've tried to read this, but somehow the article doesn't really come to a point. Is there any point besides Yud not being perfect, and having connections to people who openly talk about their sex life (and that it's somewhat non-standard)?
Can "ASI will inevitably kill us all" be falsifiable before we create ASI? If not, then what separates it from religion?
Not sure how much I agree with that position, but that's just how I understood it.
No, but there are decent proxies, like "Alice who is X standard deviations smarter than Bob will win >50% of the time at game Y even given a disadvantage that would make her chances <5% if she was playing against someone equally smart".
I've thought about this analogy for less than five minutes, so it's far from perfect. Point is, ASI being made of silicon isn't very central to the problem. There are lots of analogies that are falsifiable, and "ASI will kill us all" can be decomposed into a simple and bulletproof composition of falsifiable beliefs, even if it is too specific to be itself directly falsifiable.
How?
Suppose that for the following 20 or even 100 years nothing happens. The AIs miraculously stay as they are today, the further progress turns out to be impossible, there will be no more agents hacking various websites, no cyber warfare involving AI agents, nothing. There will be the year 2026, forever.
How would any of that impact the lifestyle of anyone in the rationalist community? Please be more specific with your accusation.
(Seems like the rationalists are simultaneously the dark lords of TESCREAL, pulling the strings of the world's most powerful billionaires and politicians, but also irrelevant weird nerds who can be ruined at any moment by a smart blogger explaining how their opinions are silly, or even that they know each other.)
The rational methods used to establish a large P(doom).
To definitively falsify the claim, humanity would need to live alongside superintelligence forever without it wiping us out. That doesn't mean we can't be pretty darn sure though, in the same way that we can be pretty darn sure that nuclear war could end civilization.
The rationalist doom and sex cult uses convoluted mathematics inspired language to justify their faith-- but at the end of the day it's just faith no matter how they dress it up.
In other words, it's not Pascal's wager if the very bad outcome is what you naturally expect to happen.
And what I just wrote applies to not just Eliezer and his allies: I don't know a single person or a single published argument that has ever used Pascal's-wager-like argumentation to argue that AI "progress" needs to be stopped. Eliezer is the author of a blog post titled "Pascal's mugging", but that blog post and any follow-up discussion are completely silent on the dangers of AI.
I don't think we get a second shot.
"The world will regress to a pre-industrial-revolution lifestyle" seems to be a retreat from "civilization will end" (not that I agree with it).
>travel will be even more risky than in the Dark Ages thanks to fallout
The radioactive contamination caused by Chernobyl has persisted for decades. In contrast, the radionuclides created by nuclear explosions are of a very different nature and persist for weeks. 3 weeks after the end of the explosions, there will be no place except ground-zero craters (i.e., very small areas) where the levels of radioactivity remain at deadly levels. (But I definitely agree with you that during those critical 3 weeks, almost all people who decide to travel will end up dying unless they know a lot about nuclear war and have good radiation detectors.)
To repeat my overall point: the worst-case scenario for the outcome of AI research should be compared to an asteroid strike by an extremely large asteroid, not to nuclear war—because nuclear war is too tame in its effects to make a good comparison.
The reason I got involved in this discussion is that some people use the possibility of nuclear war as one reason for us to go forward with AI research as fast as possible. Such people argue that a sufficiently mature AI technology would eliminate the possibility of nuclear war. I think that is a fallacious argument become AI "progress" is the more potent of the two dangers.
(doesn't mean that it's not a religion or it is, but rather I mean it being "falsifiable" or not, doesn't carry any information to that regard)
and if you put something trivially true into the blank the implication seems logical, and the argument feels powerful. but simply putting a true claim at the end doesn't mean that everything before follows from this claim. (and even if they do the argument is not a total description of all the causal chains that influence the outcomes.)
of course it's hard to do a careful survey of the facts and weigh them and then deduce some likelihoods for possibilities. because some aspects are unprecedented (investment ferocity, AI results/progress) while others are very much precedented (collapse of investment bubbles, doomsayers' wrong predictions).
Case in point - Bitfinex was and is absolutely a criminal enterprise and scam, and it didn't have any backing for years in row. But it didn't fail, because many different criminals in the world had decided that it is quite convenient tool and put a monetary IV line in it. And now ruzzians are buying kamikaze drones with it, selling oil to India with it, drug cartels are shuffling around their money with it and so on. Does it mean it wasn't and isn't bad? Of course no. They simply found a rich sponsor to bail them out.
Or the whole 2008 crisis is another popular example.
But note, you have very concrete things to claim about Bitfinex. Criminal enterprise, scam, organized crime funding, facilitating arms dealing and embargo avoidance, and so on.
Then you throw in 2008. It's definitely popular but example of what? Bad thing that didn't fail? (Things did fail, hence the crisis? Or you mean the too big to fail attitude?)
Policy proposals (eg. "coordinated AI safety R&D should be exempt from antitrust regulations" or "TARP made money on paper, but the financial sector should pay for the total cost of the crisis") cannot be divorced from the provenance of the source (sex cult, goldbugs, anticapitalists), but the same is true of bluesky screeds.
Same logic with pure AI corpos and adjacent dependents. If the doomsayers would ever be right about them being financially unsustainable, then most likely these companies would never get even close to the bankruptcy, simply because they would be bailed by external entities. So we would never be sure if OpenAI and Anthropic were "bad" or not, financially. Because either critics are wrong, or any corporate deficits would be covered by budget money (maybe foreign budget, it happens too).
The big question is, who's competing to sell us our salvation this time?
> This all most recently came together over the weekend, and they got Dario Amodei and Sam Altman and Elon Musk and Demis Hassabis all onto the same page to say they wanted to WAIVE ANTI-TRUST LAWS so that they could "pace", or slow down, research.
> I feel about this roughly the same way I would feel about all the pharmaceutical companies getting together to say they wanted to waive anti-trust laws so they could form a cartel to ban medical research:
> these people are the enemies of humanity.
> Also: Much of AI research IS medical research. They are very literally proposing to stop a pretty good chunk of medical research in its tracks so that they don't have to fear the paperclipper.
> There are many regulations I would be in favor of, but they only ever want to form a cartel or a world government against the paperclipper.
> No other industry has ever framed any regulation in terms of "stop", "pause", or "slow down". Those are insane ways to consider any form of policy. Policy should prevent bad outcomes and encourage good ones, it is never about some specific speed.
> This industry frames its regulations in this way because it has, at its core, a millenarian death cult that worships and fears AI as the cause of the end of the world, that feels compelled to create AI, and that at least somewhat wants to destroy AI as a field.
METR, AISI, MIRI, SBF, FDX, redwood research, yudkowsky, nate soares, anthropic.
frankly they are connected in something almost resembling a conspiracy (XD).
lower the confidence score of any institute connected on account of their conflict of interest with this religion.