Galton first described this, both qualitatively and statistically, 100 years prior to Pinker. He made the "ancient VVIQ" Pinker mentions. Everyone touching this topic (including Pinker) knows Galton.

Excellent passage from Adorno about tension between institutional and non-institutional intellectuals, resonant with Heinlein's famous passage on specialization. (reposting for attribution, but I would choose different contemporary examples)

I've complained before that Microsoft regularly breaks the computer I use for my young kids. But it's not just a child-user problem; the gentleman below is sophisticated. Microsoft regularly pushes broken code to hundreds of millions of people already. AI will make them worse.

People are teasing Nadella over this but he is just explaining subtleties zoomer slang. "Slop" is what the AI produces. "Clanker" is the preferred term for the AI itself.

You've heard of OpSec and CogSec. I propose LikeSec: denying the network of AI crawlers access to your likeness so they can't impersonate you. Unfortunately (?) practicing LikeSec forces every public person to be a Vtuber. The culmination of pseud/anon culture perhaps.

Despite being "long" AI on the 5-10 year frame, I must admit the current moment feels like Crypto/NFT in 2021: - businesses obliviously insisting - large market segments yelling "please stop" - "no one asked for this" / solution-looking-for-problem

At the risk of seeming crass, and with apologies in advance to his mother — it's worth considering how Swartz was prosecuted versus the (so far) light touch in AI. The relevant laws were not quite the same; there isn't a perfect comparison. Still, one can't ignore. RIP Aaron.

I am occasionally asked to do this, for example, by people creating regulatory processes. I am not exaggerating even a bit. The expectations out there are crazy. The post below is delivered as a joke, but this is actually just what people think can be done with LLMs.

AI code is a new "FizzBuzz doubt" event — people overestimate the low end of employed-programmer skill. When Spolsky wrote in 2006, many doubted FizzBuzz could filter *anyone* from a programming job. Sadly, they were wrong. Now ask: could Opus replace people failing FizzBuzz?

When GEB was published (1979) perhaps 1% of people had ever used a computer, and 0.1% had ever seen recursive or adaptive programs. It's true that 45 years later we find computing metaphors for cognition ordinary, but this is because works like GEB made a persuasive novel case.