Pennomi

Sounds like it’s time to leak a bunch of politician nudes to show them why encryption matters.

Okay but these models are actually pretty interesting when you dive in. Useful yet? Not by a long shot (I’m guessing decades, not years of work). But interesting nonetheless.

You know who wouldn’t have missed a word there? ChatGPT.

(Not saying it would’ve written a better article but AI is typically good at not making silly typos like humans.)

And let’s be honest, a certain percentage of junior devs never do learn. That’s always been the case, AI or not.

Star Trek’s WWIII happened from 2026-2053, I believe. So we’re still on track.

The thing I miss most from Reddit is all the niche technical communities. So much knowledge is contained in those.

Agreed with the other commenter, you need to verify that you have proper GPU access. Eevee will always crash if not given a GPU.

Eevee requires a GPU in ways unlike the old default renderer. If it’s virtualized there’s all kinds of things that might be wrong with getting access to a GPU.

The open paper they published details the algorithms and techniques used to train it, and it’s been replicated by researchers already.

It’s trivial compared to the compute they dedicate to AI models. Like, not even a rounding error.

The hell? There’s no reason to use plain HTTP instead of HTTPS.

And symmetric encryption is wildly irresponsible as well.

Yep, a better thing to do would be to have your browser just lie about its fingerprint on every request. Why yes, I do live in Antarctica!

I did not know that. There’s a bunch of news articles going around claiming that even the creators of the models don’t understand them and that they are some sort of unfathomable magic black box. I assumed you were propagating that myth, but I was clearly mistaken.

No one understands how LLMs work, not even on a basic level.

Well that’s just false.

10/1 to celebrate the 101 key keyboard apparently? That’s what the linked article mentioned.