undefined | Better HN

0 pointsconcinds6mo ago0 comments

And it's a 4B model. I worry that nontechnical users will dramatically overestimate its accuracy and underestimate hallucinations, which makes me wonder how it could really be useful for academic research.

0 comments

2 comments · 1 top-level

DGoettlich6mo ago· 1 in thread

valid point. its more of a stepping stone towards larger models. we're figuring out what the best way to do this is before scaling up.

spicyusername6mo ago

If there's very little text before the internet, what would scaling up look like?

j / k navigate · click thread line to collapse