In a more honest world, I think our current "generative AI" technology trained on seas of human content would be seen and used more as an intelligent indexing system, than something to create "novel content" (which, it does a very poor job at).
=> More informations about this toot | More toots from demofox@mastodon.gamedev.place
I had a nice experience over break where I asked AI for a thread safe queue with some specific features, and it gave good results, but basically I'm sure there is human written code that looks identical or nearly so. Finding it is the hard part.
=> More informations about this toot | More toots from demofox@mastodon.gamedev.place
@demofox
My original take on LLM AI seems to be getting increasing agreement: they're quite good at dealing with natural language, which was always super hard (AI-hard), but because humans have always mixed up the topic of language with the topic of thinking, people keep misunderstanding that LLMs are good at thought -- which they are not.
=> More informations about this toot | More toots from dougmerritt@mathstodon.xyz
@dougmerritt huh, interesting
=> More informations about this toot | More toots from demofox@mastodon.gamedev.place
@demofox Nowadays where search engines are annoying to use, I tend to just look stuff up at the places I usually end up anyway, which is Stack Overflow for coding topics. Pretty high chance AI tools copy from there and searching on there is faster then searching on Google and hoping the result I need from Stack Overflow appears on there.
=> More informations about this toot | More toots from wildrikku@mastodon.gamedev.place This content has been proxied by September (ba2dc).Proxy Information
text/gemini