Yes, but the cousin also doesn't need sleep and can produce a lot of work

Shut the fuck up, Donny, you're out of your element.

You cannot get the training data out of a model unless it's so overtrained on a specific piece of it that it actually impairs its usefulness for other things. So no, those works are not in there.

However, you might still be creating derivative works with an LLM, because the standard for those is "substantial recognizable elements" and the exact meaning of that can be fuzzy.

I would think that an audit trail only needs temperature readings, not adjustments.

You don't want to know who made the adjustment? Nor the time the adjustment was made? What adjustment was made?

Might as well gut audit logs in Windows or Linux if all you want to know is a reading.

Maybe, kind of.

But there are two issues here which say no.

The first and more irrelevant argument is that LLMs aren't actually thinking, their process is not like ours just because they use neural networks, which bear only a passing resemblance to how our brain might work. We don't even know all of the details there and we STILL have enough info to know they aren't the same.

But the real argument is that copyright exists to protect the work of humans to induce them to make more work, so what the software is doing inside is wholly irrelevant.

That's before you even get into the actual weeds of copyright, like whether LLM output is even copyrightable at all, and if so, which copyright applies.

It could however be segregated to a VPN.

The managers could find that out by viewing security footage.

Management requires effort, it doesn't just magically happen.

Auditing only requires that information go one way, so no, and also you don't understand the argument you're making.

This is probably a good buy regardless. Interconnects have generally been going optical.

I hear that some of the AI generated ports to retro machines are as good as the originals.

No they absolutely are not. First, they chose to take the contract with the police. Second, they chose to employ an incompetent. The driver should have determined it unsafe to tow with his equipment.

Facts matter.

The fact is that while we know Trump raped at least one woman and it was proven in court, and we have every reason to believe he raped more (as well as children) he has not been CONVICTED of rape in court. And we should take care to state these facts correctly.

With that said, I have zero compunctions about stating that he raped women and children as facts. If he hadn't raped children he certainly would have sued about it.

And it pays way better than most of the jobs which are more dangerous than cop as well.

Back in the day some game designers would farm out the actual coding of their games. The Bitmap Brothers did that a lot, but still managed to make games like Gods and Speedball 2.

These days most indie devs use frameworks that can boil game development down to writing scripts.

I think if the designers know what they are doing, and particularly if they do the graphics and sound themselves, they could probably get a decent result with AI writing the code. Likely no worse than some of the crappy developers from back in the day, lazily posting Atari ST slop to the Amiga, for example.

We're going to go back to a model where we only trust word of mouth, because any other way of discovering things is too fraught.

The clever way to technically enable this would be through a cryptographically backed web of trust layer which could be applied to the top of web browsers, news readers, and email clients to provide trust scores.

There seems to be little interest in providing this. There's no money in it since nobody will trust such a system if it's monetized. So the community will have to do it. I don't really see another way forward that doesn't basically amount to giving up on self-discovery.