query[for a friend]: if one is making a small little robot(in the form of a non-sentient computer script), what would be a good way to implement being able to communicate with it?
initial thought was LLMs which would work but mmmhhhhh probably not. unless there's any trained on public domain text etc mby.
a dropdown list of queries -> responses would work, but wouldn't really feel freetext, you should be able to "ask" it arbitrary questions.
and uhhhh thats all the ideas this one has. thoughts?
@5225225 Wouldn't bother with public-domain-trained LLMs because they consistently perform like crap from what I've seen (the whole 'tech' basically cannot work without mass nonconsensual scraping of data and conversations).
Perhaps something from the ELIZA era? Essentially just pattern matching and perhaps rudimentary knowledge graph processing, which can produce pretty acceptable results AFAIK.
@joepie91 that being said, performing like crap could be part of the charm, so it might be worth considering, do you happen to know the names of any to look into?
(old old AI art where it makes really really fucked up stuff in honestly a genre of its own is more charming than new stuff where it only really fucks up text)
will also look at ELIZA and friends.