Oddbean new post about | logout
 nostr:npub1efvh0km06s5fqtz2mczx9yutnc6f4ksn39f6z0nz5n8hgde6x7vqp04jx2 nostr:npub14pk7ymgjn7pxxahyn08taf4f6hm5avg34nh9w70wmdu24ndxs4zs2y80yc IMO LLMs are better because they have a huge amount of data of text that makes sense, in a way that is pretty readily codified (grammar, orthography). That’s probably why the first successful uses of generative AI were to write scholarly articles. Visual is an order of magnitude harder to make sense.