Recent. Yes, this is an important factor. Thanks for reminding. Because this question is mainly the college students who are deciding their paths 

Thanks for your opinion a lot

Hello Sam: Thanks very much on your warm reply and resoucres provided. And thanks for your willingness to talk with me, I might contact you later if I found there's topic that you'll be extremely helpful with

Absoultely agree. In fact, currently as an EA beginner who is doing major decision(see: https://forum.effectivealtruism.org/posts/fpJLTFH9entmNoFMM/the-value-of-non-ea-world-direct-work-vs-earning-to-give-for), I'm really suffering from lacking EAs to talk with(I can only talk to ChatGPT most of the times). I have spent a lot of hours on thinking how to network and improving my gmail writing, however it seems to me 90% of EAs are quite busy and don't have any time to talk with beginner. I hope in the future when I became more expertised, I can spend hours per week on talking to beginners.

Thank you for posting this. I'm also thinking about the probability of stable totalitarianism recently, and your post gave me some new perspectives. Especially the idea that "Lie detection or mind-reading technique may increase totalitarianism" and "P(T) is like a series of dice rolls for each country"

Intersting research. I think an important moral view that we should align AI with is that wild animal suffering is also a serious issue. In my intuition, AI might hold the view that there are a lot of happiness in the wild so spreading animals to the outer space would be something good, which is probably not true and might cause s-risks to happen.

This seems like a valuable resaerch and inspires me some ideas of how to use AI to reduce wild animal suffering(Although currently we don't have any clear interventions). Maybe in the future we could try to test how LLMs would response to digital suffering. Also another point would it be would AI support the idea of spreading wild animal to other planets?(It seems very bad in suffering focused ethics, however maybe the maintsream norm would support this idea which would make LLMs to support this)