mario_kart_tour_2

Courtesy of Nintendo

October 16, 2019

BONO | These Games Play Themselves

Print More

When Google’s next big endeavor, Stadia, was announced earlier this year (through a livestream I remember watching in the middle of a lecture hall during the last few minutes before class), it faced a lot of backlash. The idea was to release an online service — the “Netflix” of games — to stream popular video games without a console. Instead of requiring hardware or software, all users would hypothetically need would be a Google account and a good internet connection.

Right away, a problem emerges: How good an internet connection is “good”? With demanding games on their roster like Assassin’s Creed Odyssey, Borderlands 3 and Cyberpunk 2077, their servers have to be rock-solid to keep up. Games like those are being developed with more and more intensive graphics each year, and the concept of running them with no specialized hardware or downloaded software is a tall order.

How does the company plan to combat latency, the time it takes for a local copy of a game to receive information from a server? Usually, latency is only a problem in multiplayer online games, like Splatoon 2 or Overwatch — where even though the game structure itself is local to a player’s system, the interactions with other players are constantly streamed through an internet connection. Google is trying to make all games browser-based, where every aspect of the game would be streamed directly to the player, like Club Penguin and all the other Flash games from the 2000s (may they rest in peace). 2D renders of cartoon penguins are one thing, but what about complex 3D worlds made with the newest technology?

It’s worth noting that I myself tried Stadia as a beta tester last December when it was cryptically called “Project Stream,” a name I honestly kind of like better than Stadia. Using my home wi-fi and some luck, I was able to play Assassin’s Creed Odyssey for a decent amount of time. Honestly, I didn’t have too much of a problem with latency. The bigger problem was getting the game to connect in the first place, as it won’t let you so much as look at the game’s menu until your internet connection is stable enough. Once I was connected, though, I had a pretty good time playing the game. So much so that when the beta period ended, I tried to download and play the game normally on my laptop — to disastrous results. I guess my Dell XPS 13 isn’t powerful enough to run intensive graphics, so I never made it past the title screen. That’s why I, personally, am looking forward to seeing what becomes of Stadia, and whether it can live up to its promises — I’d be willing to put up with a little lag to run games that are otherwise out of my computer’s league.

Regardless, last week Google’s VP of engineering Madj Bakar told Edge Magazine that the Stadia team plans to shrink latency down in the next couple of years of service. Eventually, it will feature “negative latency,” a feature where Google’s servers predict the players’ button presses for them, using machine learning to eliminate the need for the game to even wait for user input before proceeding. It’s a concept that seems somewhat akin to handing someone an unplugged controller and telling them they’re controlling an NPC and led to even more community backlash. If games can just predict player input and operate on their own, are they truly interactive? Would the concept of these AI-driven games that play themselves ruin gaming?

I’m conflicted on the matter. On the one hand, it feels kind of disingenuous, not to mention something that sounds too good to be true. What kind of margin of error would this AI technology have, and what happens when the machine gets it wrong? It seems too early to know anything for certain, and I’m not sure why they would bring the technology’s distant future up now when the service has yet to launch. On the other hand, if the technology is good enough that it can really predict my clumsy button-presses and make me think it’s responding to my choices, does it truly matter? Plenty of games use illusions and tricks to render complex environments and details like shadows and particle effects; tricking the player is just a part of game design.

I can’t help but think about ways other games have already pulled this trick. Just the other day, at the Saint Motel concert, I was playing Mario Kart Tour between sets when the music started back up, interrupting my race. I locked my phone and enjoyed the concert, and afterwards, when I opened my phone, the race started up again as if I’d paused it. My car didn’t even swerve off the road. This would make sense if it were a normal, offline game, but Mario Kart Tour claims to be online, pitting players against each other. My opponents weren’t NPCs, they were players with usernames! How could they be still racing an hour later?

What Mario Kart Tour likely does is either give NPC racers real players’ usernames, or records players’ race performances and randomly select 11 performances to pit you against every time you start an online race. That would explain why I haven’t seen any players drop out or lag the way I would in an online console game like Splatoon. And for a mobile game, it makes sense — does it really matter that I’m not racing in real time, as long as I still have the experience of playing real Mario Kart? I almost don’t think so; I think I’m willing to be fooled if I can still feel like I’m getting the real deal.

Olivia Bono is a senior in the College of Arts and Sciences. She can be reached at ojb26@cornell.edu. On the Level runs alternating Tuesdays this semester.