| Not checkable |
Not checkable |
3 Apr 2025 |
And I get lots of praise when I read a book and often lots of money, and that’s a really good incentive.
Not checkable · 3 Apr 2025 |
|
| Not checkable |
Not checkable |
3 Apr 2025 |
And I think right now if we tried it, we would run into the combinatorial explosion.
Not checkable · 3 Apr 2025 |
|
| Not researched |
Factual |
3 Apr 2025 |
A blogger called LouKeep was widely liked about five years ago, wrote about 10 posts, and people still refer to those posts.
Factual · 3 Apr 2025 |
Other |
| Unresolvable |
Prediction |
3 Apr 2025 |
In a world of limitless prosperity after AI, the default outcome is that people engage in mindless consumerism.
Prediction · 3 Apr 2025 |
Philosophy, ethics and society |
| Not checkable |
Not checkable |
3 Apr 2025 |
I think it’s actually better at the style on a word-to-word, sentence-to-sentence level, than it is at planning out a blog post.
Not checkable · 3 Apr 2025 |
|
| Not researched |
Factual |
3 Apr 2025 |
GPT-3, asked whether bugs are real, gave a hedging answer that we can never truly tell what is real.
Factual · 3 Apr 2025 |
AI capabilities |
| Not researched |
Factual |
3 Apr 2025 |
The AI 2027 scenario discusses many technical questions about superintelligence but barely speculates about what happens in society.
Factual · 3 Apr 2025 |
Their own work or organisation |
| Unresolvable |
Prediction |
3 Apr 2025 |
Converting car factories to robot factories with superintelligence could take less time than the WWII bomber conversion.
Prediction · 3 Apr 2025 |
AI, the economy and work |
| Not checkable |
Not checkable |
3 Apr 2025 |
I think one thing that you didn’t mention in there was the data efficiency.
Not checkable · 3 Apr 2025 |
|
| Not checkable |
Not checkable |
3 Apr 2025 |
But imagine that you have to make a business out of you and your hundred closest friends who you agree with on everything.
Not checkable · 3 Apr 2025 |
|
| Not checkable |
Not checkable |
3 Apr 2025 |
And the way that the AI intelligence explosion dynamics work are just so weird that in order to have nothing happen, you need to have a lot…
Not checkable · 3 Apr 2025 |
|
| Not checkable |
Not checkable |
3 Apr 2025 |
I think it might be a crazy outcome to make a lot of money as a blogger.
Not checkable · 3 Apr 2025 |
|
| Unresolvable |
Prediction |
3 Apr 2025 |
If everything goes right, superintelligence-led car-to-robot factory conversion could be about three times faster than WWII bomber conversions, about a year.
Prediction · 3 Apr 2025 |
AI, the economy and work |
| Not checkable |
Not checkable |
3 Apr 2025 |
But nothing we’re saying is that different from what has already happened.
Not checkable · 3 Apr 2025 |
|
| Not checkable |
Not checkable |
3 Apr 2025 |
First of all, I agree if there’s only one way to do something that makes it much harder, and maybe that one way takes very long, we’re…
Not checkable · 3 Apr 2025 |
|
| Not checkable |
Not checkable |
3 Apr 2025 |
I’m not completely convinced that we don’t get something like alignment by default.
Not checkable · 3 Apr 2025 |
|
| Not checkable |
Not checkable |
3 Apr 2025 |
I kind of want to call you out on the claim that groups of humans don’t plot against other groups of humans.
Not checkable · 3 Apr 2025 |
|
| Not checkable |
Not checkable |
3 Apr 2025 |
One is that they’re too stupid to understand their training.
Not checkable · 3 Apr 2025 |
|
| Not checkable |
Not checkable |
3 Apr 2025 |
And I think that that has been the most consistently wrong prediction of all.
Not checkable · 3 Apr 2025 |
|
| Not checkable |
Not checkable |
3 Apr 2025 |
There’s also cases where it’s very obvious who’s part of what group.
Not checkable · 3 Apr 2025 |
|
| Not checkable |
Not checkable |
3 Apr 2025 |
The literal Cambrian explosion, the agricultural revolution, the industrial revolution, has phase changes.
Not checkable · 3 Apr 2025 |
|
| Not checkable |
Not checkable |
3 Apr 2025 |
I don’t buy that as a real limiting factor for most people.
Not checkable · 3 Apr 2025 |
|
| Not checkable |
Not checkable |
3 Apr 2025 |
I think with race, it’s very easy to know whether you’re black or white, and so there have been many cases of one race kind of conspiring…
Not checkable · 3 Apr 2025 |
|
| Unresolvable |
Prediction |
3 Apr 2025 |
Once AI is superintelligent, it will probably have access to much better simulations than exist now.
Prediction · 3 Apr 2025 |
AI capabilities |
| Not checkable |
Not checkable |
3 Apr 2025 |
And the answer is: combination of that speed up and the speed up in serial thought speed.
Not checkable · 3 Apr 2025 |
|
| Not checkable |
Not checkable |
3 Apr 2025 |
And I’m going to say we all believe it, although I can imagine people getting really conspiratorial about it and this not working.
Not checkable · 3 Apr 2025 |
|
| Not researched |
Factual |
3 Apr 2025 |
Scott Alexander said he had been thinking about "this" for a decade to a decade and a half. In the surrounding remarks, "this" refers to the subject of a specific scenario story that touched on the arms race with China.
Factual · 3 Apr 2025 |
AI timelines and forecasting |
| Not checkable |
Not checkable |
3 Apr 2025 |
It just seems like applying those two things.
Not checkable · 3 Apr 2025 |
|
| Not checkable |
Not checkable |
3 Apr 2025 |
And then the example I gave was that humans also can’t do this.
Not checkable · 3 Apr 2025 |
|
| Not checkable |
Not checkable |
3 Apr 2025 |
Once we can really understand the weights of an AI on a deep level, then we have a lot of alignment techniques open up to us.
Not checkable · 3 Apr 2025 |
|
| Not checkable |
Not checkable |
3 Apr 2025 |
It’s probably the weirdest couple of years that have ever been.
Not checkable · 3 Apr 2025 |
|
| Not checkable |
Not checkable |
3 Apr 2025 |
Kind of looks like you asked ChatGPT to summarize the past five years of AI progress, and you got something with a couple of…
Not checkable · 3 Apr 2025 |
|
| Not researched |
Factual |
3 Apr 2025 |
GPT-4 does not give hedging answers to 'are bugs real?' and says that they obviously are.
Factual · 3 Apr 2025 |
AI capabilities |
| Not checkable |
Not checkable |
3 Apr 2025 |
And also you can correct people when they’re wrong, which is a very important side benefit.
Not checkable · 3 Apr 2025 |
|
| Not researched |
Factual |
3 Apr 2025 |
Scott Alexander's p(doom) is the lowest of anyone on the AI 2027 team.
Factual · 3 Apr 2025 |
Their own work or organisation |
| Not checkable |
Not checkable |
3 Apr 2025 |
One of them is just continuing the next token prediction work.
Not checkable · 3 Apr 2025 |
|
| Unresolvable |
Prediction |
3 Apr 2025 |
Superintelligences will be somewhat distributed throughout the entire economy, trying to expand everything.
Prediction · 3 Apr 2025 |
AI timelines and forecasting |
| Not checkable |
Not checkable |
3 Apr 2025 |
So for me, the most salient example is the etymology of words.
Not checkable · 3 Apr 2025 |
|
| Not checkable |
Not checkable |
3 Apr 2025 |
But we know the people involved, we know who’s in the government, we know who’s leading all of these labs.
Not checkable · 3 Apr 2025 |
|
| Not checkable |
Not checkable |
3 Apr 2025 |
It seems like he’s a genius, he’s thought about this for years, probably at some point, he heard a couple of Indian words and a couple of…
Not checkable · 3 Apr 2025 |
|
| Not checkable |
Not checkable |
3 Apr 2025 |
This is actually part of our misalignment story, is that if the AI is sufficiently misaligned, then yes, we can tell it it has to follow…
Not checkable · 3 Apr 2025 |
|
| Not checkable |
Not checkable |
3 Apr 2025 |
So a regulation saying there has to be whistleblower protection.
Not checkable · 3 Apr 2025 |
|
| Not checkable |
Not checkable |
3 Apr 2025 |
Like, I think in order to have nothing ever happen, you actually need a lot to happen.
Not checkable · 3 Apr 2025 |
|
| Unresolvable |
Factual |
3 Apr 2025 |
Speaking about Elon Musk and SpaceX, he said he thought that in 2000 people would not have expected that someone could move two or five times faster than NASA with fairly limited resources.
Factual · 3 Apr 2025 |
Economics and markets |
| Not checkable |
Not checkable |
3 Apr 2025 |
So one of these models is just a hyperbola. Everything is along the same curve.
Not checkable · 3 Apr 2025 |
|
| Unresolvable |
Prediction |
3 Apr 2025 |
AIs will be solving alignment because even misaligned AIs want to align their successors.
Prediction · 3 Apr 2025 |
AI safety and alignment |
| Not researched |
Factual |
3 Apr 2025 |
Eli Lifland is a member of Samotsvety, described as the world's top forecasting team.
Factual · 3 Apr 2025 |
Field building and careers |
| Not checkable |
Not checkable |
3 Apr 2025 |
We could have a different copy of the superintelligence optimizing every single part full-time.
Not checkable · 3 Apr 2025 |
|
| Not checkable |
Not checkable |
3 Apr 2025 |
I think it is plausible to imagine a future society that works the same way.
Not checkable · 3 Apr 2025 |
|
| Not checkable |
Not checkable |
3 Apr 2025 |
Everything we’ve talked about has happened before.
Not checkable · 3 Apr 2025 |
|
| Unresolvable |
Prediction |
3 Apr 2025 |
Scott Alexander put his own p(doom) at about 20%, which he described as the lowest of anyone on his team.
Prediction · 3 Apr 2025 |
AI safety and alignment |
| Not checkable |
Not checkable |
3 Apr 2025 |
But I do think that we have a pretty good chance of surviving.
Not checkable · 3 Apr 2025 |
|
| Not checkable |
Not checkable |
3 Apr 2025 |
You get more lottery tickets for these discoveries when you are intelligent, when you have good technology, when you’re doing good work.
Not checkable · 3 Apr 2025 |
|
| Not checkable |
Not checkable |
3 Apr 2025 |
If you actually want to understand an issue in depth, you can’t use deep research. You gotta do it on your own.
Not checkable · 3 Apr 2025 |
|
| Not checkable |
Not checkable |
3 Apr 2025 |
I am also worried about UBI from a different perspective.
Not checkable · 3 Apr 2025 |
|
| Not researched |
Factual |
3 Apr 2025 |
Scott Alexander said he thought that when Grok was asked who is the worst spreader of misinformation, it refused to respond "Elon Musk".
Factual · 3 Apr 2025 |
AI safety and alignment |
| Not checkable |
Not checkable |
3 Apr 2025 |
Right now, I just write about it. I think writing about it is important, but I don’t know.
Not checkable · 3 Apr 2025 |
|
| Not checkable |
Not checkable |
3 Apr 2025 |
Yeah, I think it might be useful to think of our timelines as being like 2070, 2100.
Not checkable · 3 Apr 2025 |
|
| Not checkable |
Not checkable |
3 Apr 2025 |
I admit that my p(doom) is literally just p(doom) and not p(doom or oligarchy).
Not checkable · 3 Apr 2025 |
|
| Not researched |
Factual |
3 Apr 2025 |
Scott Alexander said that, in the predictions of AI milestones he has seen, such as Katja Grace's expert surveys, forecasters have almost always been too pessimistic about how fast AI would advance.
Factual · 3 Apr 2025 |
AI timelines and forecasting |
| Not checkable |
Not checkable |
3 Apr 2025 |
And so after dozens of viral hits, then you have a fan base.
Not checkable · 3 Apr 2025 |
|
| Mostly true |
Prediction |
3 Apr 2025 |
In the AI 2027 scenario, the government sees via cyber warfare that AIs are as capable as the best human hackers but deployable at huge scale.
Prediction · 3 Apr 2025 |
AI policy and governance |
| Not researched |
Factual |
3 Apr 2025 |
Current AI is much less data efficient than humans.
Factual · 3 Apr 2025 |
AI capabilities |
| Not checkable |
Not checkable |
3 Apr 2025 |
Would all of those journalists have become bloggers if there was no such thing as mainstream media? I’m not sure.
Not checkable · 3 Apr 2025 |
|
| Not checkable |
Not checkable |
3 Apr 2025 |
And like I feel like the company leaders have so far made slightly better noises about caring about alignment than the government leaders…
Not checkable · 3 Apr 2025 |
|
| Not checkable |
Not checkable |
3 Apr 2025 |
And if you look at why this happens, why this doesn’t happen, it tends to happen in cases where one group has an overwhelming advantage.
Not checkable · 3 Apr 2025 |
|
| Not checkable |
Not checkable |
3 Apr 2025 |
But just running them through these environments again and again, training on the successes, training against the failures, kind of…
Not checkable · 3 Apr 2025 |
|
| Not checkable |
Not checkable |
3 Apr 2025 |
When it is one smooth process, the smooth process is often a hyperbola that shoots to infinity in weird ways.
Not checkable · 3 Apr 2025 |
|
| Not checkable |
Not checkable |
3 Apr 2025 |
I have always wanted to get more involved in the actual attempt to make AI go well.
Not checkable · 3 Apr 2025 |
|
| Not checkable |
Not checkable |
3 Apr 2025 |
To me it does not seem like the same kind of problem as inventing all human institutions from the Paleolithic onward.
Not checkable · 3 Apr 2025 |
|
| Not checkable |
Not checkable |
3 Apr 2025 |
There are lots of ways that cheating and doing bad things can improve your success.
Not checkable · 3 Apr 2025 |
|
| Unresolvable |
Prediction |
3 Apr 2025 |
Other countries, especially China, will develop superintelligence around the same time, so AI will probably not stay confined to data centers.
Prediction · 3 Apr 2025 |
AI policy and governance |
| Unresolvable |
Prediction |
3 Apr 2025 |
AIs would spend some time, which he put at about a year (possibly more or less), getting the facilities they need for experiments, such as automated biology labs, running to clear bottlenecks.
Prediction · 3 Apr 2025 |
AI, the economy and work |
| Unresolvable |
Prediction |
3 Apr 2025 |
Superintelligence will be able to figure out how to build a Dyson sphere faster than unassisted IQ 100 humans would.
Prediction · 3 Apr 2025 |
AI timelines and forecasting |
| Not checkable |
Not checkable |
3 Apr 2025 |
I think of superintelligence as partly when they become so much more data efficient than humans that they are able to build on cultural…
Not checkable · 3 Apr 2025 |
|
| Unresolvable |
Prediction |
3 Apr 2025 |
Because of an AI research progress multiplier, the technological level of the year 2100 might be reached earlier than people expect.
Prediction · 3 Apr 2025 |
AI timelines and forecasting |
| Not checkable |
Not checkable |
3 Apr 2025 |
I feel like if it’s literally the superintelligent AI that is never wrong, then we have solved some of these coordination problems.
Not checkable · 3 Apr 2025 |
|
| Not researched |
Factual |
3 Apr 2025 |
AI systems are currently getting more reliable.
Factual · 3 Apr 2025 |
AI capabilities |
| Not researched |
Factual |
3 Apr 2025 |
In the AI 2027 scenario, the AI comes to understand its goal system and pretends to want moral things while humans are watching.
Factual · 3 Apr 2025 |
AI safety and alignment |
| Unresolvable |
Prediction |
3 Apr 2025 |
Training failures where AIs are not trained to do what developers intended will get much worse as AIs become agents.
Prediction · 3 Apr 2025 |
AI safety and alignment |
| Not checkable |
Not checkable |
3 Apr 2025 |
I think that’s just a really big speed up.
Not checkable · 3 Apr 2025 |
|
| Not checkable |
Not checkable |
3 Apr 2025 |
So these AIs will have access to all human knowledge, they will have read management books in some sense, they’re not starting blind.
Not checkable · 3 Apr 2025 |
|
| Resolved true |
Factual |
3 Apr 2025 |
Nvidia originally came from gaming.
Factual · 3 Apr 2025 |
Economics and markets |
| Not researched |
Factual |
3 Apr 2025 |
In 2021 Daniel wrote 'What 2026 Looks Like', a forecast of the next five years of AI progress.
Factual · 3 Apr 2025 |
AI timelines and forecasting |
| Not checkable |
Not checkable |
3 Apr 2025 |
My guess is it’s a lot worse than for startup founders.
Not checkable · 3 Apr 2025 |
|
| Resolved true |
Factual |
3 Apr 2025 |
Scott Alexander said he thinks the AI alignment community did not really expect LLMs.
Factual · 3 Apr 2025 |
AI safety and alignment |
| Not researched |
Factual |
3 Apr 2025 |
Scott Alexander blogged on Livejournal for several years, where people liked his blog but it was just another Livejournal blog.
Factual · 3 Apr 2025 |
Their own work or organisation |
| Not checkable |
Not checkable |
3 Apr 2025 |
I think I was the most boring normie liberal in the world before encountering LessWrong.
Not checkable · 3 Apr 2025 |
|
| Not checkable |
Not checkable |
3 Apr 2025 |
So like we said before, I think there’s a big difference between discontinuous and very fast.
Not checkable · 3 Apr 2025 |
|
| Not checkable |
Not checkable |
3 Apr 2025 |
And if it says like, “by the way, try to manipulate the government into doing this or that”, then we know that something bad has happened…
Not checkable · 3 Apr 2025 |
|
| Unresolvable |
Prediction |
3 Apr 2025 |
In future, once they have superintelligence and want robots, AI companies like OpenAI will be able to buy up many factories.
Prediction · 3 Apr 2025 |
AI, the economy and work |
| Not researched |
Factual |
3 Apr 2025 |
Scott Alexander said he thinks "we are kind of starting to move away" from LLMs toward reinforcement learning agents.
Factual · 3 Apr 2025 |
AI capabilities |
| Not checkable |
Not checkable |
3 Apr 2025 |
I realize that’s a much weaker thing than we’re positing, but it was done just with the human brain with a lot fewer resources than we’re…
Not checkable · 3 Apr 2025 |
|
| Not checkable |
Not checkable |
3 Apr 2025 |
I think if you were to try this, you would get something that was so obviously an AI house style that it would use the word ‘delve’ or…
Not checkable · 3 Apr 2025 |
|
| Not checkable |
Not checkable |
3 Apr 2025 |
that team of ethnobotanists would succeed in fewer than 50,000 years.
Not checkable · 3 Apr 2025 |
|
| Not checkable |
Not checkable |
3 Apr 2025 |
I think if you wrote a really stupid blog post, it could point out the correct objections to it.
Not checkable · 3 Apr 2025 |
|
| Not checkable |
Not checkable |
3 Apr 2025 |
I’m a little bit concerned that AI will make it much easier to break anonymity.
Not checkable · 3 Apr 2025 |
|
| Not checkable |
Not checkable |
3 Apr 2025 |
I think you can either be the sort of person who can go on a Dwarkesh podcast and get the New York Times to write about you, or you can do…
Not checkable · 3 Apr 2025 |
|
| Not researched |
Factual |
3 Apr 2025 |
AIs currently lie to people all the time.
Factual · 3 Apr 2025 |
AI safety and alignment |
| Not researched |
Factual |
3 Apr 2025 |
Algorithmic progress in AI already doubles roughly every year.
Factual · 3 Apr 2025 |
AI capabilities |