All of lukasb's Comments + Replies

Regarding AI lab coordination, it seems like the governance teams of major labs are a lot better placed to help with this, since they will have an easier time getting buy in from their own lab as well as being listened to by other labs. Also, the frontier models forum seems to be aiming at exactly this. 

Thanks for writing this! If I don't want to sign-up to the finders course, are there any resources you would recommend for doing the one-hour lovingkindness sessions?

Agreed. My view here is pretty uninformed.

If taking your lawyer's advice, in this case, means being silent for 5-7 years, it seems like some people should speak openly and bear the costs.

Why are so many people disagreeing with this?

I'd be interested to hear what you think is going wrong with Paul's writing style, if you want to share.

Hm, yeah I guess my intuition is the opposite. To me, one of the central parts of effective altruism is that it's impartial, meaning we shouldn't put some people's welfare over other's. 

I think in this case it's particularly important to be impartial, because EA is a group of people that benefitted a lot from FTX, so it seems wrong for us to try to transfer the harms it is now causing onto other people.  

Maybe I'm misunderstanding bank runs, but as I understand it, they happen because 

• the institution that is holding other people's money doesn't have all that money in liquid form 
• they are unable to give it back if everybody tries to deposit it at once
• when this happens, the institution runs out of money and many people, who didn't withdraw their cash in time, lose all their deposits

I think the reason Richard listed #2 as a preference is that there might still be hope that FTX doesn't run out of money in the first place and no one loses their deposi... (read more)

Or better yet, at Y Combinator. 

For Level 3: Machine Learning, this document might be useful. It provides a quick summary/recap of a lot of the math required for ML.

This looks exciting! I plan to apply. 

One reaction I have looking at the syllabus is that it's too theoretical for me in the beginning. I feel like it would be better to have an applied component from the start. Maybe the first two weeks could be theory paired with writing a short distillation in the first week, getting feedback, and then refining it. The feedback loop of actually writing is probably by far the best way to improve distillation skills.  This is just an impression I have though, so I could be wrong.

Nice work! Just wanted to flag that the eaopp.com link is down for me. 

Personally, I don't have a problem with the title. It clearly states the central point of the post. 

Regarding the example, spending $5k on EA group dinners is really not that much if it has even a 2% chance to cause one additional career change.

How much of the impact generated by the career change are you attributing to CEA spending here? I'm just wondering because counterfactuals run into the issue of double-counting (as discussed here). 

Thanks!

I agree that there is an analogy to animal suffering here, but there's a difference in degree I think. To longtermists, the importance of future generations is many orders of magnitude higher than the importance of animal suffering is to animal welfare advocates. Therefore, I would claim, longtermists are more likely to ignore other non-longtermist considerations than animal welfare advocates would be.

Thanks for writing this! It seems like you've gone through a lot in publishing this. I am glad you had the courage and grit to go through with it despite the backlash you faced. 

What does TUA stand for?

I agree with Aaron! Given the little time you have, I would make the pitch as simple as possible.

Fair enough. I would guess you can usually have a higher impact through your career since you are doing something you've specialized in. But the first two examples you bring up seem valid.

For context here's a LessWrong post talking about the Paxlovid situation.

This seems like a good idea.

I submitted the following comment:

I urge the FDA to schedule its review of Paxlovid and to make the timeline 3 weeks or less, as it did with the COVID vaccine.

1000 people are dying of COVID in the US every day. With an efficacy of 89%, Paxlovid could prevent many of these deaths. The earlier Paxlovid is approved, the more lives will be saved.

Thank you for your consideration.

I wasn't sure what topic to put it under so I chose "Drug Industry - C0022." 

I like this framing a lot. I particularly like the idea of replacing the phrase "doing good" with "helping others" and "maximization" with "prioritization."

I understand the impulse to mention volunteering before donations and careers because people naturally connect it with doing good. But I think it would be misleading for the following reasons:

• As you said, there is currently very little emphasis on volunteering in EA
• In most cases, individuals can do much more good by changing their career path or donating

I  think we should be as accurate as we can w... (read more)

Is there evidence/theoretical reason to believe that not experimenting in governance leads a movement to become slow over time?

thank machine doggo

[Comment pointing out a minor error]  Also, great post!

I'm impressed with the success you guys had! I'm excited to see your organization develop.

Great post!  Thanks.

Good point. I'll bring this up with other group leaders.

This approach is compelling and you make a good case for it, but I think what Lynch said about how not supporting a movement can feel like opposing it is significant here. On our university campus, supporting a movement like Black Lives Matter seems obvious, so when you refuse to, it makes it looks like you have an ideological reason not to.

What is the best leadership structure for (college) EA clubs?

A few people in the EA group organizers slack (6 to be exact) expressed interest in discussing this.

Here are some ideas for topics to cover:

• The best overall structure (What positions should there be etc.
• Should there be regular meetings among all general members/ club leaders?
• What are some mistakes to avoid?
• What are some things that generally work well?
• How to select leaders

I envision this as an open discussion for people to share their experiences. At the end, we could compile the result of our discussion into a forum post.

In the beginning of the Christiano part it says

There can't be too many things that reduce the expected value of the future by 10%; if there were, there would be no expected value left.

Why is it unlikely that there is little to no expected value left? Wouldn't it be conceivable that there are a lot of risks in the future and that therefore there is little expected value left? What am I missing?

Thanks for pointing that out!