🧵 View Thread
🧵 Thread (11 tweets)
@RamonDarioIT https://t.co/sTmC9DUtUZ
@nabeelqu I think grad level math is harder, but great prose (vs just “clear” communication) comes out of sensations and emotions that the models don’t yet have the capacity for. Maybe it will take artificial biological AIs with bodies and serotonin transmitters to write good prose.
@nabeelqu It's interesting that nobody (major) has made a Github Copilot equivalent for prose... like forget styling sentences and just help people complete them https://t.co/6JOy3EUaTO
Making a 'LiveWriter' prototype: AI-based autocomplete for prose I think this is a high-utility use case that's surprisingly under-explored I can't really put it online for free cause of the token $ but can put the code on Github I guess (or let users use their own API keys) https://t.co/0NIr61S3kA
having done both of these things i can confirm, yes. graduate level math looks hard from the outside because of the jargon / symbolism but that's just a matter of unfamiliar language. high quality prose is, almost by definition, very readable so it doesn't look hard. but writing well involves this very global use of one's whole being to prioritize what is relevant, interesting, entertaining, clarifying, etc. and ignore what is not, whereas math can successfully be done in this very narrow autistic way