Probably a big story on a technical side: Nvidia showed off their new technology and it is the fastest and angriest my feed has been in years.
I've been watching this story unfold all week, and it's deeply troubling. Some of the results are technically impressive, especially in making Starfield's janky "I <3 2006" character models look like they belong in a modern game, but there are so many other things to be concerned about that it negates the positives.
Some of the controversies have been well covered elsewhere, such as the algorithmic averaging out of artistic intent, and the scope for publishers to axe even more developers by letting them spam the AI button to turn undercooked games into a superficial replica of a finished product. It was especially tone deaf of Nvidia to use RE9's Grace as the poster girl for their AI yassification overlay, since the gaming community has rapidly connected with her flawed humanity, and turning her into a glossy pin-up model tramples on that. Then there's the fact that they needed two 5090s to get this working. If this is what maxed out DLSS5 looks like on 50-series GPUs, I dread to think how a pared back version would look on my new 5070, which is a hell of a thing to find myself thinking about a graphics card I bought a few weeks ago for more than the cost of a PS5. DLSS5 is clearly tech intended for the 60-series or later, which means any current gen implementation is likely to be using gamers as beta testers.
Something I've seen less discussion about is how obviously cherry-picked the trailer shots are. It's mostly static scenes with nothing stressing the usual gen AI pressure points like hand movement, facial expression changes, or fast action. Is there any chance of this keeping up with a melee action game, for example?
Overall, it feels like a cheap, cynical shortcut to an awkward estimation of next-gen visuals. For years we've been seeing diminishing returns in the increase in graphical quality bought with increased hardware power, and DLSS5 feels like a smoke-and-mirrors trick intended to look like a generational breakthrough at a glance. If I have any hope for future implementation of this tech, it's that we have currently only seen it slapped onto already completed games. It may integrate more smoothly with development workflows that take it into account from the outset. But who knows.