Jan Leike: "I'm excited to join @AnthropicAI to continue the superalignment mission!"

By defun 🔸 @ 2024-05-28T18:08 (+35)

This is a linkpost to https://x.com/janleike/status/1795497960509448617

OllieBase @ 2024-05-29T07:45 (+8)

I'm grateful you shared this since it's probably interesting to the community, but flagging I'm a little nervous about the EA Forum filling up with cross-posts from X. I think that would make it less interesting and unique.

defun @ 2024-05-29T08:37 (+7)

I share your concern.

The main reason I share some Twitter links is that I think they're relevant (this seems to be confirmed by the positive votes). 

Deciding which big lab to join is a very important decision that many EAs are making.

(Personally, I often find myself going to Twitter because there are very relevant news updates regarding AI safety that aren't shared in the forum)

OllieBase @ 2024-05-29T08:54 (+10)

I agree that upvotes are an update that the content was useful and that might mean I'm just off here.

I guess I think relevancy or interest isn't the only thing we should be tracking; I think the forum adds value because, unlike Twitter, the posts and replies here are usually more thoughtful and more carefully investigated. To put it another way, if I just wanted relevant news and updates, I'd go to Twitter. Here, I'd hope for more commentary (e.g. "Jan has joined Anthropic. Here are some thoughts I have about this and how they might be relevant for EAs deciding which lab to join"). 

I'm not saying you personally should have done this and, again, I am grateful you made the effort to share this.

defun @ 2024-05-29T09:05 (+5)

Totally agree!

I would love to read a thoughtful analysis (I'm not qualified to do it).

My main question is: "Should EAs still try to join OpenAI?" (80,000 Hours is still recommending it)

Geoffrey Miller @ 2024-05-29T15:59 (+7)

I think, at this point, EAs (including 80k Hours) publicly boycotting OpenAI, and refusing to work there, and explaining why, clearly and forcefully, would do a lot more good than trying to work there and nudge them from the inside towards not imposing X risks on humanity. 

defun @ 2024-05-29T09:09 (+3)

Maybe there should be a separate space for "relevant EA news"? So that EAs don't have to ~waste time on Twitter to be up to date.

Personally, I wish it existed.

defun @ 2024-05-29T09:47 (+10)

A potential solution would be for the forum to have a tag called something like "News" and then with the "Customize feed" feature let people choose between hide/reduce/default.

Sarah Cheng @ 2024-05-29T11:08 (+11)

Appreciate the suggestion! Just wanted to mention that this already exists - users can add the "News relevant to effective altruism" topic to posts, and they can change how posts in this topic are weighted in their frontpage list.

defun @ 2024-05-29T11:22 (+1)

Oh, sorry! I literally searched for "News" but didn't see it ðŸ¤¦ðŸ¤¦

Thanks!

Linch @ 2024-06-03T01:33 (+2)

Great minds! :)

OllieBase @ 2024-05-29T09:15 (+5)

You might like David Nash's Monthly Overload of Effective Altruism.

It has a combination of forum/EA org updates and relevant news.

I can also recommend Shakeel's Transformer newsletter for AI-specific news.