A very astute framing by Ted Chiang—large language models as a form of lossy compression for text.
When we’re dealing with sequences of words, lossy compression looks smarter than lossless compression.
A lot of uses have been proposed for large language models. Thinking about them as blurry JPEGs offers a way to evaluate what they might or might not be well suited for.
I’m sure that by now you’ve already seen the infamous email from Richard Stallman—free software’s own worst enemy—detailing his somewhat eccentric approach to speaking at conferences.
I particularly like the memetic variation of The Stallman Dialogues. There’s a real genius in the way that it quotes passages from the email verbatim.
Y’know, I’m supposed to have a Skype call with Andy sometime next week about my upcoming talk and workshop at Build (tickets are still available for the workshop, by the way). I’m very tempted to channel my inner Stallman for the duration of our conversation.
Meeting that sad animal is not an agreeable surprise.
Hillman Curtis's new film about David Byrne and Brian Eno will be premiering at Southby. Should be fun.