Friday, October 10, 2025

Hello. Should we just go ahead and get started?

Item 1: a link

Back in January, Julien Crockett interviewed sci-fi writer Ted Chiang (previously recommended in these pages, I think, but I can't find where) about AI. Chiang had written about AI for the New Yorker in 2023, putting forth the metaphor of the "blurry JPEG" to explain both why generative AI output is at least superficially compelling, but also why it inevitably contains inaccuracies (or "hallucinations").

A year or two later, Chiang remains reasonably skeptical about AI, but fair and curious, and I don't think he'd need to change a word from this interview based on anything that's happened since he sat down for it.

Do you think that LLMs will become useful tools that can reliably answer questions?

I don’t want to say LLMs are only good for entertainment; there are many respects in which LLMs are genuinely amazing. The fact that they can rephrase something in any style of prose is fascinating; no one would have predicted that statistical models of all the text on the internet would be capable of that. But predicting the most likely next word is different from having correct information about the world, which is why LLMs are not a reliable way to get the answers to questions, and I don’t think there is good evidence to suggest that they will become reliable. Over the past couple of years, there have been some papers published suggesting that training LLMs on more data and throwing more processing power at the problem provides diminishing returns in terms of performance. They can get better at reproducing patterns found online, but they don’t become capable of actual reasoning; it seems that the problem is fundamental to their architecture. And you can bolt tools onto the side of an LLM, like giving it a calculator it can use when you ask it a math problem, or giving it access to a search engine when you want up-to-date information, but putting reliable tools under the control of an unreliable program is not enough to make the controlling program reliable. I think we will need a different approach if we want a truly reliable question answerer.

Item 2: a list

Paul Thomas Anderson Feature Films, Ranked:

1. Phantom Thread
2. There Will Be Blood
3. Magnolia
4. One Battle After Another
5. Boogie Nights
6. Punch-Drunk Love
7. The Master
8. Inherent Vice
9. Licorice Pizza
10. Hard Eight

Item 3: a media recommendation

Item 4: word of the week

Maculate

Just once I'd like to leave Benihana without a shirt maculated with that mysterious brown dipping sauce.

Item 5: a photograph

See ya!

Thank you. Next week.