L
2

Is fine-tuning big models really better than training small ones from scratch?

I keep seeing people drop $50k on fine-tuning GPT-4 for niche tasks when a tiny model trained on their own 500 records would actually outperform it, but then others argue fine-tuning saves months of dev time - which side is actually right for small teams?
2 comments

Log in to join the discussion

Log In
2 Comments
the_jennifer
Friend of mine spent thousands on a big model and his cat knocked the server offline.
2
the_nancy
the_nancy10d agoMost Upvoted
Did the cat survive the server crash?
6