Hacker News new | past | comments | ask | show | jobs | submit login

Very cool. I do wish it could be trained on Douglas’ books, since it has no knowledge of Zem:

> user: are there really sentient mattresses?

> guide: oh absolutely, they're probably lounging around discussing the meaning of fluff and velvet, just like we question life's absurdities.

> user: what are they called?

> guide: those clever cushions prefer to be known as the "thinking textiles," pondering on all things plush and profound.




It was. But the dataset for training is rather small and focuses only on the first book.

I might improve it in the future


I was sad that it didn't seem to know much about the Pan Galactic Gargle Blaster. Fun project though!


I asked it what the strongest drink ever was and it gave me pan galactic gargle blaster


Also you could achieve that with RAG quite easily. training is good mainly for the tone and style (which still isn't perfect) but better than most things out there that simply use the llm with no fine tuning




Join us for AI Startup School this June 16-17 in San Francisco!

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: