Disclaimer: Just skimmed the paper.

Their definition of "self-replication" seems to be just to copy the model weights to another place and start it on an open port - no unexpected capability.

To see the performance of AI agents on more complex AI R&D tasks, I recommend METR's recent publication: https://metr.org/AI_R_D_Evaluation_Report.pdf 

Having a human-centric core in EA seems robustly good to me, so we're sure to make the world a better place even if our other ideas don't pan out.

$1 increase in women’s income

What kind of income is this? Monthly, yearly, lifetime?

Very interesting, thanks for the analysis!

To me especially Figure 3 is important, it's insane how much food goes unharvested and how bad animals are at producing food.

Is there also a version that uses kcal instead of tons and that breaks down animal products into meat and vegetarian?

Minor correction, it's Alternative für Deutschland.

I don't know much about painkillers in LMICs, but in my understanding the US prescribes many more opioids than other western countries (like Germany, where I'm from). So this idea/problem might be specific to the US.

@mincho: How does Suzetrigine compare to existing non-opioid painkillers?

For anyone interested in how to go from research to large-scale production, there recently was an article about that in Works in Progress: https://worksinprogress.co/issue/getting-materials-out-of-the-lab/

Update: We'll be joining forces with the Qualia Research Institute, who are in Berlin this week: https://qualiacomputing.com/2024/05/17/qri-in-germany/

I think we'll get along great and there's a big overlap. Scott:

The Qualia Research Institute is doing absolutely picture-perfect mad science.

Just an info, nothing actually changes!

I think it would still make sense to add it because I assume many people (like me) will add it manually anyway, giving you more free-text to analyze. :)

Why is there no "Science" as default option under "Do you watch YouTube videos on any of these broad topics?"?