Saturday in Italy

I skipped last weeks Saturday roundup of links to write a review of the Apple Vision Pro. I may not have quite the stamina to write a big one today either. But I wanted to put a few things out here today, not as much filler, so much as things I found interesting this week. But I’m feeling a little under the weather so I’m going to not probably spend as much time on it today as normal.

I came across this longer than average article about Conan O’Brien at NYT this morning. He has a new travel show on Max, but the article got more into his longer arc. Some things I already knew, but I didn’t know about this recent appearance of his on Hot Ones — the YouTube show where guests are interviewed while they eat around ten increasingly hot chicken wings. I was speechless watching Conan on this. Actually speechless isn’t the right word — I was very loudly laughing. People are calling Conan “unhinged” in this episode by the end of it. That’s the only word for it. Absolutely worth the 27 minutes it takes to watch it. There’s really nobody like Conan.

Changing subjects, Ilya Sutskever who was an OpenAI cofounder and who was instrumental in getting Sam Altman fired last winter and who also left the company last month announced this week the name of his new company. The name of the company — Safe Superintelligence — and its website are pretty lackluster and unsexy and while I don’t fully understand what it’ll actually do except that it’s clearly devoted to protecting humanity from rogue artificial intelligence, I am glad he’s found his place in this emerging ecosystem. NYT has an article about it. My understanding is that we owe ChatGPT-4’s existence to Ilya, and for that he has my respect forever.

This is an article about how there were some other papers around the time of Imbens and Angrist’s classic Econometrica on the LATE theorem that also were getting to that same place. The authors own 1994 paper also appeared to introduce LATE. But the main article I first linked to is both about LATE as well as about “simultaneous discoveries.” I’ve been introduced in the idea of simultaneous discoveries for a while but on a different topic. Michael Spence’s signaling model seemed to have precedent in David Lewis. Ten or so years ago, I wrote this timeline of Spence and Lewis trying to understand how both of them independently came up with that model. The easy answer is just to say “because of Thomas Schelling”, but it’s more than that. Still, these simultaneous discoveries are interesting.