SoylentNews is people
SoylentNews
from the if-all-you-have-is-a-big-cloud-all-your-games-will-be-movies dept.
Stadia (Google's next gen gaming console) will use AI to achieve negative latency in games
Speaking with Alex Wiltshire in Edge magazine #338, Google's top streaming engineer claims the company is verging on gaming superiority with its cloud streaming service, Stadia, thanks to the advancements it's making in modelling and machine learning. It's even eyeing up the gaming performance crown in just a couple of years.
"Ultimately, we think in a year or two we'll have games that are running faster and feel more responsive in the cloud than they do locally," Bakar says to Edge, "regardless of how powerful the local machine is."
This would be achieved using Google's homegrown streaming tech, which it's been teasing ever since Stadia was first announced late last year with Project Stream. The company believes its tech is capable of overcoming the hurdles presented by over-the-web gaming, despite its extensive web of datacentres sitting potentially hundreds of miles away from a user.
Specifically Bakar notes Google's "negative latency" will act as a workaround for any potential lag between player and server. This term describes a buffer of predicted latency, inherent to a Stadia players setup or connection, in which the Stadia system will run lag mitigation. This can include increasing fps rapidly to reduce latency between player input and display, or even predicting user inputs.
Yes, you heard that correctly. Stadia might start predicting what action, button, or movement you're likely to do next and render it ready for you – which sounds rather frightening.
With enough latency, the game will play itself and the console will just stream the game-play movie. I have the feeling a Netflix subscription will be cheaper.
(Score: 2, Interesting) by Anonymous Coward on Sunday October 13 2019, @12:00AM (1 child)
It's probably not as dumb as it sounds. MS has been "cheating" for years by repositioning bullets to account for latency. Basically repositioning the bullet so that it hits the thing that appeared to be aimed at when the player pressed the button to fire, even if the player's display was slightly out of sync with what the weapon was actually aimed at when the button was pressed.
In this case, I doubt they're going to be trying to predict anything beyond the next second or two, as the reaction time of even bad gamers isn't going to require it. A typical reaction time isn't going to require more than a second of predictive capabilities and most likely less during games that would benefit from this. During that amount of time, there's a limited number of possible things for the player to do.
This isn't an easy thing to accomplish, but it's not impossible. In many cases, it's probably going to feel like lag messing with a key press did. Weird at first and after that people are probably going to get used to it.
(Score: 3, Insightful) by Mykl on Sunday October 13 2019, @09:42PM
This won't be predicting the next second or two - this will be predicting actions in the next 50-100ms or so. Lag is a huge issue for First Person Shooters and other twitch-reflex games. I'm OK in principle with the corrections that games perform today (as an earlier poster said, adjusting for what appeared on the players screen at the time of their action rather than the actual state of the server), but deciding to shoot an opponent because I 'probably' will do so really soon sounds dodgy.