Things I read and liked in March

Weeks when decades happened

1. Between the release of OpenAI’s GPT-4, the abrupt mainstreaming of AGI safety, and Harry Potter Balenciaga, it was impossible to keep up with AI this month. In my opinion, the best you can do is follow Zvi Moshowitz, who covers AI developments thoroughly and insightfully. His coverage includes: GPT-4, the ‘Sparks of AGI’ paper about GPT-4, the open letter on pausing AI, more developments).

2. An old George Saunders short story that is very good.

3. ettingermentum @ettingermentum

   nice argument. however i have already portrayed myself as the sleek FDR bullet train and you as the seething coal locomotive.

   

4. From the Qualia Research Institute: “This paper proposes the "Heavy-Tailed Valence” (HTV) hypothesis: the notion that human capacity for emotional experiences of pleasure and pain spans an incredibly wide range of intensity, a minimum of two orders of magnitude between the most mild and the most intense experiences”. Related: extreme bliss states.

5. What we owe the future

   Mαtt Thomαs @mattthomas

   

6. Rosie Campbell on her experience with weight loss drug semaglutide, and on why some people are judgy about the drug:

   “I suspect this is in part due to the naturalistic fallacy, and in part due to some misplaced notion that things are only virtuous if they are difficult. The evidence is quite clear that diet and exercise alone is not enough to help most people lose weight and keep it off, no amount of hammering on about it and shaming overweight people for not being virtuous enough is going to change that. I am overjoyed that there now seems to be a solution that actually works!”

7. A fascinating history of oral rehydration therapy, one of the most underrated miracles of modern times:

   “In the Dacca cholera hospital where Cash worked in the 1960s, patients sometimes arrived only after long, tortuous journeys by bicycle rickshaw. And for every patient who made it to the clinic, Cash knew there were others who would never survive the long trip. In the West, doctors weren’t coming up with new treatments for dehydration because they already had easy access to intravenous drips. For doctors like Cash, Pierce, and Phillips, however, it was clear there was a pressing need for a much simpler way of treating dehydration.”

8. The “dark Satanic mills” of factory farming: “I must admit that I am taken aback by the continued prevalence of horrific animal abuse in a nation where nearly two-thirds of the population identifies as Christian….One need not think that the consumption of meat is intrinsically impermissible (I certainly do not think so) in order to oppose factory farming.”

9. Bryan Caplan thought large language models were overhyped, so in January he bet Matthew Barnett that an AI would not get As on his midterm exams by 2029. GPT-4 has already gotten an A on one of them. Caplan: "Base rates have clearly failed me".

10. Related:

    tetraspace 💎 @TetraspaceWest

    "AI will never" it's going to in like 6 months

11. Natural selection favors AIs over humans.

12. An excellent essay about a machine learning researcher being surprised and changing his mind based on the continued progress of large language models. An open-mindedness that’s maddeningly rare: “Today, I see a similar pattern of resistance to taking LLM results seriously. Many of my colleagues' views on LLMs have not changed at all over the last couple of years.”

13. Related (now at 36 followers):

    Robert Long @rgblong

    one of my favorite AI takes of the month is from a dog with 14 followers

    

    Zlatý Retvítr @retvitr

    Please stop trying to show how #GPT4 is stupid by giving it riddles I cannot solve.

14. Programmer Simon Willison on how “AI-enhanced development makes me more ambitious with my projects”

15. Jonathan Birch foretells the absurdity we’ll feel navigating the uncertain sentience of AI systems: “I suspect that human lives are more absurd now, on average, than they ever have been, and that this upward trend will continue and accelerate…I believe AI has the potential to supercharge absurdity”

16. Lilian Weng’s characteristically masterful overview of prompt engineering.

17. Peter Wildeford: “Essentially, effective altruism contains three radical ideas that I don’t easily find in other communities. These three ideas are ideas I want to protect.”

18. Ava on the importance of having fun: “You have to do the thing you actually enjoy doing, not the thing you find conceptually exciting. You have to date the person you actually like, not the ideal of perfection you fetishize in your mind.”

19. Related classic Robin Hanson post on play: “While conscious planning can at times be important, what tends to matter more is finding a strong motivation, any strong motivation, to really get into whatever it is you are doing.”

20. Garrison Lovely on pescetarianism- includes an excellent survey of evidence for fish pain, sociality, and preferences.

21. C.S. Lewis on the desire to be an “insider”, to be in the “Inner Ring”: “As long as you are governed by that desire you will never get what you want. You are trying to peel an onion: if you succeed there will be nothing left. Until you conquer the fear of being an outsider, an outsider you will remain.”