What you need to know
- VP of engineering Madj Baker expects Stadia to outperform local gaming systems in 'a year or two.'
- This would be accomplished by making use of Google's machine learning technology to create 'negative latency.'
- In order to do this Google will attempt to buffer gameplay and even predict what buttons a player will press next.
Back when Google was testing out Stadia under the name "Project Stream", I was blown away by the responsiveness and quality the streaming video game service provided. However, that was when it was locked to 1080p at 60fps. Once Stadia launches, it is promising the same low-latency with 4K HDR resolution at 60fps.
Not only that, the VP of engineering Madj Bakar thinks Stadia will outperform local gaming systems in "a year or two."
That's some big talk for a company whose first gaming product won't even start shipping until November, but if any company can do it, it's Google.
To achieve this, Baker says that Stadia will use machine learning to buffer gameplay and even predict the player's next button presses. This will create what he refers to as "negative latency" to prevent any lag for the player.
In other words, it would act similar to how YouTube loads, only much more complicated because Stadia will have to predict what you'll do next. It should also prevent any performance hiccups users might experience due to their connection or own equipment.
While we might have to wait a couple of years to see if Baker's predictions come true, you won't have to wait much longer for Stadia. If you want to give the game streaming service a try, you can pre-order it now and check it out when it launches in November.
Personally I don't think this will be around in 2 years. But gaming industry is a hard one to guess so maybe it will be the next big thing.
What's the production time for AAA games... 2-3 years minimal with some taking as long as 10 years. If you have Stadia Entertainment currently working on AAA games and some third party developers currently working on exclusives too, how can you assume Stadia will shutdown in 2 years? Seems a little baseless other than you just believe the platform is a fail before it's even launched.
Hope they account for trolls somehow. I can see a troll campaign buying a game and purposefully jumping the first cliff (think Mario) a little to early. This will teach the AI To think people will want to press the button early and send many players to their doom.
I'd bet good money this "feature" will actually make the game perform significantly worse than on consoles. It also guarantees many people who are serious about competitiveness will avoid stadia like the plague.
Don't be so sure. Nobody does machine learning better than Google.
They use the same technology for many online competitive games that are server based. There's a fighting game company that sells technology for it as well. This is nothing new, but to say better than almost instant, that seems like a pipe dream. I would say akin to consoles would be a better goal.
This seems like you could easily generate a horrible feedback loop. As I'm learning a game, I'm going to progressively get better at it (in theory). But the machine learning is going to be making predictions based on my old patterns. Let's say that I start out by hitting a button 300ms earlier before I should. As I'm trying to retrain myself to delay, the machine learning is fighting me, sending that button push at the usual time, even though the actual press came 250ms later. In theory, the machine learning should eventually pick up the change, however, the feedback that I get from the game is that waiting that extra 250ms made no difference. So I wait 400ms, and it seems closer, so I wait 600ms, and now I'm 100ms late. Maybe once you completely master a game, you can eventually get that to zero out, but getting to that point seems like it's going to be frustrating as hell. And all this is BEFORE you throw an lag into the mix.
Thank you for signing up to Android Central. You will receive a verification email shortly.
There was a problem. Please refresh the page and try again.