undefined | Better HN

0 pointsthrowaway3141555mo ago0 comments

> which presumably hasn't done a fresh pre-training over the web

What makes you think that?

> Did they figure out how to do more incremental knowledge updates somehow?

It's simple. You take the existing model and continue pretraining with newly collected data.

0 comments

A leak reported on by semi-analyses stated that they haven't pre-trained a new model since 4o due to compute constraints.

j / k navigate · click thread line to collapse

A leak reported on by semi-analyses stated that they haven't pre-trained a new model since 4o due to compute constraints.

j / k navigate · click thread line to collapse