Does this resolve NO in the case where the dominant power at Anthropic, EOY 2027, is Claude, because Askell is not Claude? Or does it resolve YES because Askell influenced Claude and thus Claude's influence is partly her influence?

@MartinRandall if she still works for the company, then I assume we should judge her "power" relative to that of the other human employees, even if none of the humans have very much. But the ongoing effects of her work definitely shouldn't count; the question is about Anthropic's corporate politics.

Thanks for filling my order, folks! Now talking my book: Askell has been there like five years, so her initial vest is done. And now she is getting high profile media coverage. She will have millions of incentives to leave.

Feels like Eliezer's approach here is counterproductive. I dunno about how things work at Anthropic, but if someone were to try this on me - (1) start a twitter beef with my manager, (2) suggest that he should be jealous of my success, (3) insult his integrity, (4) say "you better not fire that guy! I'll be real mad!"... then I'd feel like they were hurting my ability to retain my job, rather than helping.

@DanB lol I don’t think Dario would care

idk I feel like Elezier really enjoys the idea of saying I told you so and thus tends to be overconfident

I guess this market is kind of similar to the Elon Trump markets from a year ago (maybe not), in that it’ll stay steady around 75-80%

Basically it’s asking if there’s a divergence between two powerful people who have every reason to not separate, but I guess it might happen like Elons America party, I don’t think any AI company from Askell would do anything meaningful (probably less or similar to SSI or TM at best but they had a ton of technical people and investment, in the scenario Anthropic is worth ~1T but it loses “alignment” how much appetite would there be for a less effective AI)

I don't think this would convince me of your hypothesis from Twitter. If Askell leaves, which I consider plausible, I'd guess it was for any of the other reasons one happens to leave a company, rather than concluding it was the specific reason you theorized.

@ZaneMiller of course there are many possible reasons that many things happen but also most people who leave companies are not askell leaving anthropic. she is in a relatively special position.

@ZaneMiller I agree the hypothesis isn't convincing, at least by Yudkowsky's standard. If the idea was "How can I increase the chances that Askell stays a power at Anthropic? I'll just tell Dario why he'll fire her. He's gonna want to prove me wrong." though... I'd say this was well worth a shot.

On a related note, started reading her dissertation and its fascinating.

@Yakushi12345

@Yakushi12345 I'm sure this is very fascinating but it's just endless pages of incomprehensible math to a pleb like me

@bens I'm not that far yet, and definitely hard to follow.

But a lot of it is just laying out how utilitarian intuitions get really weird under infinite situations

@Yakushi12345 Thanks, I've been looking for good takes on population ethics, this might be worth to "delve into".

My initial concern with the described scenario was that it wasn't obvious that Dario has the independence to pull a trick like that off. If Amanda Askell is being cynically maneuvered against, she could, in principle, bring it to the attention of Anthropic's board of directors, or the Anthropic Long-Term Benefit Trust.

But there seems to be scant information about the LTBT available online. This article claims that "the Trustees must consult with and consider the views of the company’s directors and CEO regarding appointments of Trustees"

One interpretation of this would be that Dario, as CEO, has veto power over LTBT trustee appointments, which have an extremely short term (1 year). Anthropic's agreement with the LTBT can only be changed by a supermajority of voting shareholders; if the threshold is set high enough, it's possible that Dario holds enough voting stock that he may be able to unilaterally block any changes.

Even if not, it's worth noting that one third of the board of directors at present is Dario and his wife, and it is unclear to me if they occupy the seats selected by the LTBT.

Thus, it is possible that Dario effectively has, or will have, effectively unilateral control over Anthropic, which would make it trivial for him to make a move like this.

@Anthem A substantial amount of my concern about Anthropic is that I've never been able to get any straight answers about who holds actual de-facto voting power, and if there's any actual check on Amodei that he couldn't beat trivially if he wanted to -- never mind beating it with the same level of effort, cunning, and system violation that Altman used to coup OpenAI. It's always just "Oh I trust our managers, they'd never do anything wrong like that, we're all friends at our company".

I this about the Elon tweet?

@FergusArgyll
Indirectly, but it's directly from https://x.com/allTheYud/status/2023574055601983603

@JoshSnider Ah, thanks.

10k at 69% on YES. Happy to negotiate a larger bet somewhere in that vicinity.

context tweet:

@Bayesian I admire the specificity of this scenario but uhhh

Why do the criteria have this bit about "significantly demoted / sideways-promoted"? This seems like a pretty subjective criterion for a market where the creator is betting and I read the tweet as just saying she will no longer be employed.

Edit: this comment was wrong

@DavidHiggs Which one? The rest of the tweets in the thread don't mention the possibility of Askell taking on a different role.

@BoltonBailey huh, you’re right I just re-read it and it’s not in the original tweet thread. I could’ve sworn he mentioned it somewhere outside this market but I guess that’s just the fallibility of human memory for you