Saw Wendell do this on the latest GN video and for some reason just knew it needed to be a thing.

That face when you are fighting for consumer rights but you are obsessed with AI.

Oof

I feel like wendell has one of the more balanced/pragmatic takes on the tech, and at the least he's a huge proponent of self hosted models instead of paying the big providers.

As he described it, the Terminator is coming whether we like it or not so we need a Summer Glau.

The cat is out of the bag and one of the ways to fight for our rights is by owning as much of the tech as possible. Deprive the big AI firms from the revenue they expect for the services they sell. Make them mid-enough-to-fail. If we let them make the money they expect to make we can't regulate them out of our lives without a major socioeconomic collapse like the great depression.

Don't forget self hosted models are made from the same stolen data, made in the same data centers, and are even more unreliable. They are just as bad.

That is so fetch.

Oh shit I forgot about that meme image macro.

Streets ahead

"Whoopsy daisy"

It is kind of unfortunate that he needs AI to anything security related

Whoopsy doodle!

In my opinion they are not just as bad. I've found small local models do a very good job of solving various small problems, while using minimal energy. At worst they're just regular bad. If you think mass data collection without consent to do stuff is a problem then its time to stop using just about any internet product with recommendations like YouTube or search engines as they all track data and scrape websites. There's always a line and I feel the ongoing bad actors that are AI companies that have growth at any cost and already made local models are wildly different in the evil scale.

...
Damnit that works on me

Selfhost Apertus open source model then.

Then for you it's okay the use of models trained with stolen data and, more importantly, works stolen from real artists, scientists, philosophers, etc., just because "iT's SeLf HoStEd"?

All of us are trained on stolen data too. Unless you invented your own language and came up with every idea you have from scratch, then we're all running on borrowed information. It's our shared inheritance as humanity.

That's why open-weight local models are the only way LLMs should be done. The data belongs to us, humanity. We made the data that they scraped, over thousands of years, and we're still doing it to this day every time we post a comment on a public forum.

It shouldn't be kept locked up in a vault where some oligarch can control its distribution and divvy it out for profit. It should be held in a public repo where anyone can download it, run it, tweak it, like any other open-source software. It belongs in the public domain.

Diffusion generation is different of course because the data is more complex, and stylistic details are often directly spoofed from the originals. It's taking magazine-cuttings and calling them original work. AI slop isn't art.

Language generation on the other hand is just... how humans have been speaking and sharing ideas since we began verbally communicating.

While using minimal energy to run, but not minimal energy to make.

Great, so we get a ethical shitty piece of tech.