I didn't realize this was part of the rationalist-originated "AI Village" project. See https://sage-future.org/ and https://theaidigest.org/village. Involved members and advisors include Eli Lifland and Daniel Kokotajlo of "AI 2027" infamy.
I hate-listen to his podcast. There's not a single week where he fails to give a thorough tongue-bath to some AI hypester. Just a few weeks ago when Google released Gemini 3, they had a special episode just to announce it. It was a defacto press release, put out by Kevin and Casey.
Bay Area rationalist Sam Kirchner, cofounder of the Berkeley "Stop AI" group, claims "nonviolence isn't working anymore" and goes off the grid. Hasn't been heard from in weeks.
In fairness, not everything nVidia does is generative AI. I don't know if this particular initiative has anything to do with GenAI, but a lot of digital artists depend on their graphics cards' capabilities to create art that is very much human-derived.
Yud: "That's not going to asymptote to a great final answer if you just run them for longer."
Asymptote is a noun, you git. I know in the grand scheme of things this is a trivial thing to be annoyed by, but what is it it with Yud's weird tendency to verbify nouns? Most rationalists seem to emulate him on this. It's like a cult signifier.
Now that his new book is out, Big Yud is on the interview circuit. I hope everyone is prepared for a lot of annoying articles in the next few weeks.
Today he was on the Hard Fork podcast with Kevin Roose and Casey Newton (didn't listen to it yet). There's also a milquetoast profile in the NYT written by Kevin Roose, where Roose admits his P(doom) is between 5 and 10 percent.
I know it's been said thousands of times before, but as a software developer I've never felt a greater sense of job security than I do right now. The amount of work it's going to take to clean up all this slop is going to be monumental. Unfortunately, that kind of work is also soul-deadening.
