digg/ai
1d ago

NVIDIA Trains Billion-Parameter Models Using Evolution Strategies Without Gradients

——0——
Original post
ChrSzegedy
Christian Szegedy#43@CHRSZEGEDY

I wonder if this will be more impactful than https://openai.com/index/evolution-strategies/ was.

6:25 PM · May 15, 2026 View on X
QUOTE POSTChristian Szegedy#43Christian Szegedy@CHRSZEGEDY

I wonder if this will be more impactful than https://openai.com/index/evolution-strategies/ was.

1:25 AM · May 16, 2026 · 54.9K Views
QUOTE POSTKyle Cranmer#477Kyle Cranmer@KYLECRANMER

Interesting. See clarifying comment from @visceralnair. Also, happy to see @shimon8282 in author list .

2:25 PM · May 16, 2026 · 1.1K Views
NVIDIA Trains Billion-Parameter Models Using Evolution Strategies Without Gradients · Digg