"protect against renegade AI"

This is a fundamentally unsolvable problem. It's like: "How do we stop humans from being humans?"

Saying that's your objective is fundamentally dishonest. You can't protect against the unknown. You can't protect against anything without covering all the corner cases, and AI, by it's very nature, is almost exclusively corner cases.

You've apparently not worked in "developer heavy" companies before.

Nobody gives a shit about operations, generally.

It's Covid/Two Weeks to Flatten The Curve/The Terrorists/SARS/9-11/The Russians!/The Nazis - all over again.

Just reactionary FUD to try to herd sheep into a specific direction.

This shit can't be trusted.

"I'm sorry, Dave. I'm afraid I can't do that."

It's interesting that 2001 predicted modern AI so well.

It's all so tiresome.

How's it going to 'replicate it on every vulnerable system it can reach' when big, fat American AI requires several terabytes of RAM, a few million dollars of GPUs and hundreds of kilowatts of power to run?

That ain't gonna run on my dishwasher, no matter how insecure it may be.

The whole thing is simply ludicrous slop to justify regulation to keep competitors out of their market so they won't go bankrupt.

Trump is basically a 1990s Democrat.

Many people are simply too young to remember when Democrats got elected by saying they'd keep jobs in America, keep foreigners out of America, and Bomb for Peace.

PSyop.

There are trillion-dollar IPOs at stake here and it's clear that they can't justify that price if everyone can just run local open-source Chinese LLMs instead of big, fat American closed-source LLMs. Therefore they need a Narrative to ban open-source LLMs and force everyone to use theirs.

It's so lame it's sad to see anyone actually falling for it.

How can you state that? I'm assuming you're a Rust proponent from that wild statement, which goes to show you how little you know about the topic.

Yes, memory leaks can and do happen in Rust. Rust is not perfect. Neither are the people who write it.

Most Rust code is written by its proponents' AI models, anyway.

This is an engineering maxim as old as time.

If your response to theoretically old, brittle, and "wrong language" code is "rewrite it in something better", you don't understand the problem.

There has never in the history of software development been an effort to rewrite a platform in another language which wasn't subsequently plagued by all manners of new and interesting problems.

In a best case scenario, the product has an exhaustive regression testing framework, complete unit tests for every function, e2e testing, and reintegrative CI/CD. But that is rarely true for any project, and certainly isn't true for old ones which predate such practices.

You make an important point: the biggest positive contribution to making old tools better without introducing more brittle behavior is with coverage from test suites, and improved code coverage.

This IS the Rust community you're talking about, after all.

While Rust is a fine language, the community tends to be a bit fanatical and irrational. Pair that with a generally shallow technical aptitude depth and a lack of experience, and vibe coding, and therein lies the problem. There seems to be a high intersection between them and the systemd, so that's not a surprise.

And that's without getting into questionable mental health practices that are common in that community.

> The AI movie experience is slowly improving, but is still terrible. You can make a very short 5s shot that's QUITE convincing...so long as there's a single subject...really closeup.

That's simply not true. I watched a ten-minute AI-generated Star Trek video on Youtube last week which was better than any offficial Star Trek show of the last few years.

And the guy made it at home on an Nvidia GPU in his PC using a Chinese video model.

The only thing local models do is produce text. Unless you run it in a harness which sees it generate "send all your data to www.evil.com" and actually does that, the model can't do anything.

For a lot of uses like coding, the model has to have harness support for things like reading and writing files, but for basic chat-bots they just have to generate text and the harness has to display it.

Spent decades in North Dakota and Alaska.
Yes, 0C was still T-Shirt and Shorts weather until I got over 45 or so.
Went back to college, saw a dude still in shorts and tshirt at ~-20F, which is getting close to -20C.

I've now spent ~4 years in Florida, have worn pants like once.

Nope, they're all "distilled" models. That's what post training is, largely, and they're all using other models to help curate the datasets for this. They're all doing it. It's why all the models appear to be converging in terms of capabilities and personality: they're effectively using the same feed stock and using an increasingly similar set of techniques.

US "frontier" labs are distilling, as well. It's postulated (with good reason) that Opus 5 is, for instance, just a distillation of Fable to a smaller model: it's very architecturally and behaviorally different than prior Opus releases.

The anti-distillation argument is basically the same argument that's been used against Open Source in general over the past 30 odd years. It's anti-competitive and self interested.