one trillion parameters
1 statements across 1 episodes · 0 bullish · 0 bearish · 1 people on the record · first statement Sep 27, 2023 by Christian Lanng · across every show →
Everything said about one trillion parameters, oldest first
Sep 27, 2023
Lanng: Marginal Public Data Differences Are Rounding Errors in Large Models
“It doesn't matter if you have 10% different data if you train on one trillion parameters. It doesn't matter because it's such a big data set that that actually is such a small, Rounding error in, in what you get out on the other side.”