That wasn't even remotely rigorous?

FWIW Zero-3 refers to a common strategy for sharding model components across GPUs (commonly called FSDP-2, Full Sharded Data Parallel). The "3" is the level of sharding (how much stuff to distribute across GPUs, e.g. just weights, versus optimizer state as well, etc.)

A Codon-based model is cool. I know NVIDIA is building quite a large one.

At GTC they showed an SAE they built on a smaller version of it, allowing you to see what their model learned: https://research.nvidia.com/labs/dbr/blog/sae/

It's important to note that pathological liars don't stop lying. In fact, when they're caught lying red-handed, they usually double down and lie even more.

I also assume these damage control type missives to be very misleading. Seen so many of these on HN over the years.

Also: you'd expect a compliance company to understand basic software licensing, especially the most popular.

The CEO is a clear scammer. How anyone trusts another word out of her math is beyond belief.

That reddit thread is brutal, knowingly making interviewees pay hundreds of dollars to interview in this economy is messed up.

Finally, my years of playing Starcraft have real-world use! Also: Everyone will soon bow to S. Korea :D

Just gotta worry about the network lag

Right? At least they're easy to identify (and subsequently close).

> At least they're easy to identify

For better or worse, I found this one different.

Usually I see a solid wall of black, but this one was actually readable with scripts disabled.