skrunch

Glad I'm not the only one, I was half expecting cards from "Aperture Science" or "Black Mesa" to show up in there.

It's a shame there's the "This site is a parody" footer, I'd be curious if LLMs were to actually fall for this without it, especially if that were to depend on model size.

Would love this with CS1.6 voices: "GO GO GO!", "The bomb has been planted", "Need backup"

Except the image benchmarks are compared against 2.0, which seems suspicious that they would casually drop to an older model for those.

Quick, someone let James Hoffman know!

> PyTorch offers a way to export to ONNX but you will encounter various errors. [1]

I mean sure, there are limitations, but this is greatly exaggerating their impact in my experience. I'd be curious to hear from anyone where these have been serious blockers, I've been exporting PyTorch models to ONNX (for CV applications) for the last couple of years without any major issues (and any issues that did pop up were resolved in a matter of hours).

On the fence about this. On the one hand I agree that sensitive data should stay local, but on the other hand this would make their very valuable asset (the model) public.

The visualisation linked to at the end goes into some more detail and was definitely worth the extra 5 minute read for me!

There's also a fantastic little cocktail bar called Ladies and Gents in Kentish Town, if you ever find yourself in the area.

From the article:

"In an attempt to understand how the virus works, researchers at the University of Otago in New Zealand discovered in 2017 that the virus binds to a receptor called ANTXR1, which is active in more than 60 percent of the cancers in humans."