>>725104812
On the hardware thing, you're right, but Nvidia is just straight building chips designed for LLM workloads now, so GPUs will be a separate, smaller market for now.
As for the latency issue, maybe it'll get to that point, but I don't see it happening that quickly. On the idea of Sony using AI to predict player movements, this already exists without AI in fighting games and is the basis of rollback networking. It's kinda interesting to read about if you're into that nerd shit, but basically it simulates and best-guesses what the player will do and will simulate a bunch of possible next-moves and then just needs confirmation which one to do. It's very fast, because the computer guesses and then confirms and executes on what it had guessed.
If we're talking modern "AI" as in LLM stuff, I don't see how AI will lessen latency, when it currently operates on the order of seconds to minutes, to reduce latency on something that needs to be done on a scale of milliseconds.
Pricing for this shit is still an issue, because whoever is supplying the hardware as a service, needs to keep it up to date and able to run games well. Then they need to buy licenses for the games and scalability across all regions, so how much do you need to pay to maintain cutting edge hardware, power to run it all, across multiple regions, state of the art networking, teams of researchers figuring out how to reduce latency to nearly nothing and then also buying licenses for every game a player wants on the system?
Do you charge players for the game licenses? How much do you charge them for the service? How do you make money for your company out of all this?
OnLive tried, Google tried, GaiKai tried, Nvidia is trying, Microsoft is trying, I think Sony is trying, some Switch games were streamed, but it's just not catching on. I don't think it's financially feasible when you could just sell users hardware directly. and make most problems go away
