True AI Could Ruin Video Games, Not Make Them Better

Because YOU will determine whether you win or not at the outset

Wolfenstein 3D (ca. 1992)

Humans love to play games. Be it an ancient board game like Go or the newest video game, we seek them out to fill our spare time. Some of us love the challenge of competing against another human in an epic battle of strategy or luck. Some of us enjoy the adrenaline that, now that our food comes from a store or restaurant and giant predators aren’t invading our villages each night, comes from dealing with a kind of stress we no longer routinely face each day. Some of us just enjoy the companionship over a shared experience. There are lots of reasons why we humans love games, and artificial intelligence (AI) will ruin them all.

Eventually, AI will equal human intelligence. And, shortly after that, it will surpass it. When AI can “think” better and faster than humans, we will have achieved what experts call “high-level machine intelligence,” and experts expect we have a 50/50 chance of achieving that by the year 2050.

But we don’t even need to wait that long. An AI from IBM, Deep Blue, surpassed the best human chess grandmaster, Garry Kasparov, in 1997, 20 years ago. AI from Google recently demolished Lee Sedol, the foremost human Go player. These are just two of the highest-profile examples and it reveals a clear conclusion: Once trained in the rules and strategy of a game, AI will trounce the human mind.

Yet, despite this, game makers of all kinds are toying with the idea of adding artificial intelligence to games. First-person shooters have long been ripe for enhancement by incorporating better algorithms to make the computer-controlled baddies harder to frag. Even board games are getting in on the action by adding in smartphone app-based capabilities.

But the AI that today’s videogames have incorporated hasn’t been “true” AI. In contrast with computational behemoths like Deep Blue and AlphaGo, these videogames have employed “intelligence” not much more sophisticated than a few conditional if-then statements.

But that could easily change. As AI platforms, like Google’s TensorFlow, become more easily integrated into software and web services, the expertise needed to embed these technologies in everything from digital doorbells (facial recognition to announce who’s on your stoop) to dating sites (how much of a profile is credible?) to, yes, videogames will rapidly diminish. And it will (eventually) ruin videogames.

Why? Imagine every game with a virtually unbeatable opponent. Those shifty figures in your first-person shooter are now so crafty, they seem to anticipate your plan before you even execute it (because they do…based on a mashup of your previous play-throughs as well as a “big data” mining of play styles, tactics, and decisions by everyone who has every played the game). AI, if implemented at full power, will make every game frustratingly unbeatable by the humans who have chosen to play.

So, game makers will resort to their tried-and-true approach they’ve been relying on for decades: The user-chosen difficulty level.

Far Cry 3 (2013)

But with true AI integrated into the game, the hardest level will be completely untenable for a human. In fact, it will be so lopsided that even the hardest “Deathwish” setting will have to be greatly watered down for the human to have a chance of enjoying the game, let alone surviving long enough to win it.

As a result, all the different skill levels will necessarily be “fake” competition. Imagine a grown adult with genius IQ playing chess against a toddler. There is literally no way the child can win if the adult tries even a little bit. This is an understatement of the situation we’ll find ourselves with true AI in videogames.

So, whether you win or not — the level of competition the game throws at you — will largely be predetermined by the skill/difficulty selection you make when starting to play. Sure, you might screw up and end the game earlier than you expect. Or, you might do well enough to win. But, in the back of your mind, you will know that the game really wasn’t even close to breaking a sweat. There was no real competition here…the AI was just playing with you, not against you.

And maybe that will be fine for some people. But a lot of us will find it difficult to thoroughly enjoy our victories knowing it wasn’t really ever a game to begin with. The enjoyment that comes from “winning” is quite different than the enjoyment that comes from a pre-designed and curated entertainment experience, and a lot of people will realize that. For them, this will ruin games. For others who choose not to think too much about it, they might still enjoy their distractions.

Many might end up resorting to games where there is no AI involved to compete against, or even to shape the gameplay. And that could further fuel the rising interest in tabletop games, where the only intelligence you’re playing against is sitting across from you eating chips and using some faulty, inefficient wetware to determine whether or not to play that card.

One caveat: The only hope we have of true AI making games better is if we teach it how to curate a game session so that the human player(s) enjoy it more while still feeling like they “won.” This is possible. Just like a successful Dungeon Master in a role-playing game can create a storyline that involves serious challenges while also balances the difficulty against the players' (and their characters’) abilities, an AI overseeing a video game might be able to adjust the game to keep it both from becoming overwhelming and tedious. If we can accomplish that, AI might just become the best thing that’s ever happened to gaming. But that will take a fundamentally different approach than what game developers have taken so far. Fingers crossed they will.

One clap, two clap, three clap, forty?

By clapping more or less, you can signal to us which stories really stand out.