Newmark's Door

One good thing: if actually 16 billion passwords--about two for every person on the planet--have leaked, presumably no future security breach will add significantly to the problem.

"Why AI is still making things up".

"Why ChatGPT Creates Scientific Citations — That Don’t Exist".

The author has written a series of energetically anti-Elon articles so I'm not sure how correct this piece is. But he raises an interesting question: what, exactly, is the value of having a human-sized, human-scale robot? For at least some uses a robot specialized to a single task would be much more efficient.

Interesting--post is free access as of 6/3--account of how all-of-a-sudden Substack as gone mainstream.

Interesting perspective that runs counter to the AI-is-going-to-replace-almost-all-white-collar-workers articles I've seen a lot of recently.

AI is not perfect yet.

A Twitter thread analyzing the "core claims around AI risk" and finds them wanting. (Link via Marginal Revolution.)

I believe it: "Long-term studies show stable loneliness levels over decades, not an 'epidemic'."

Discusses five job characteristics that lead to conjectured "high vulnerability," led by "short time-horizon tasks".