Quoting: EhvisThe one thing I'm wondering. Nvidia stated that they used two 5090s for the demos, one for the game and one for the dlss processing. They also stated that it will be releasing this autumn. How is that going two work? Did they really pay off a bunch of studios for demo games so that those 7 people in the world that have a spare 5090 can use it?

We don't know why it was running on the other card, it's not necessarily for performance reasons, it's possible that the cards need a firmware update that currently borks other things and this was the only workaround they had on hand. We've seen in previous conferences that they rush announcements out with little warning to the teams working on them.

View this comment - View article - View full comments

What's weird is that the RE Requiem example is the only one that looks like an "AI-ifyer", the others really just look like a Reshade addon (at 15x the computational cost)

View this comment - View article - View full comments

Quoting: tesfabpelLike, this video on Crimson Desert "denoiser" (how is that just a denoiser??):

Actually the point of ray reconstruction/regeneration is that the older denoiser techniques were less faithful to the intent of the developer by overly smoothing lighting features.
This is just an ML model's conception of photorealism regardless of the underlying art direction.🤮

View this comment - View article - View full comments

I just don't know why people think these things are reliable. I just asked ChatGPT if you can use options separately (pacman -Syu -p) and it said -p shouldn't be used with Syu because it's for querying/installing packages. It's a dry-run option.

View this comment - View article - View full comments

Quoting: eggroleIf the AI generated code can't be found without labelling it, what is the problem?

If the code was crap or buggy and you could reasonably sniff it out that would make sense. But if the only way for you to tell if it is AI generated is the person behind it announces it, then the AI code is literally indistinguishable from the human generated code.

If he never labeled the stuff in the first place, no one would have ever noticed or cared.

LLMs, by definition, model language. They're designed to look right. Whether they are right is purely circumstantial and that requires the code to be deliberately differentiated from written code. It requires greater scrutiny because it hides its mistakes through language that we're susceptible to. It's called the Eliza effect, people have a bad habit of attributing human traits to things.

View this comment - View article - View full comments

How is it cheaper to implement something once instead of implement something only in specific parts of one country? 😅

View this comment - View article - View full comments

Quoting: ExplosiveDiarrhea

Quoting: GoEsrEU laws don't apply in the US, yet every new iPhone has a USB-C port despite Apple yelling and spitting like a Visigoth against it.

And that's because the EU smartphone market is bigger than the USA, by a couple hundreds million users actually...
It only made sense for Apple to move away from their proprietary stuff in this case: splitting their product lines just for one tiny, in comparison, market was out of the question from an economical standpoint.

Apple were the ones fighting against it, as I said. The point is, per your original comment, that laws applied in important markets like the EU and the US do tend to just be applied globally because it's easier and cheaper.

View this comment - View article - View full comments

You know, I never thought of it that way. You're right, this could massively backfire.

View this comment - View article - View full comments

EU laws don't apply in the US, yet every new iPhone has a USB-C port despite Apple yelling and spitting like a Visigoth against it.
Some projects/companies will do the minimum necessary to comply where they have to but far more will just apply them universally.
Interestingly the California bill at least is meant to provide information to applications, but I haven't seen any conversation from developers on what that implementation is supposed to look like.

View this comment - View article - View full comments

They're basically inventing a "pirate" industry where open-source software will have to implement these measures, then the community will create patches (cracks) to remove them. 😆

View this comment - View article - View full comments