What Use Humans?
In a world With Ultra Smart AI Based Bots, Are Humans Useful?
Not a day goes by that there isn’t yet another story discussing how an AI has bested a human in something – typically a game, which we all know are microcosms of the real world.
First chess, then Go a few months ago. Now the AI developers have set their sights on MMORPGs like World of Warcraft, with much more complex rules engines. Just the other day, a bot from OpenAI bested the best players in Dota 2, a multiplayer online battle area, where players, team up in teams of 5 to destroy the other teams Ancient (the center of the enemy base), a loose tower defense game – a bit like capture the flag.
While the rules for chess and Go are simplistic, strategies can be very complex. Online games like Dota 2 and World of Warcraft (or WoW) are orders of magnitude more complex, however, they also more accurately model the real world.
World of Warcraft and other similar games go out of their way to physically mimic reality, using physics engines mapped onto the real world. AI backed bots can very quickly master these games, even if they don’t have any special non-human skills and abilities, by simply cloning the game world, creating players which are copies of themselves, and playing themselves in the world, at many times the speed of the human player.
The bot can learn, then master the game, in very little time. Once the bot has mastered the rules of the game, it can very quickly dispatch the human players. It can have a full and complete picture of all the rules and ways of doing things, coming up with new, non-human solutions and strategies.
One can look at this and be completely convinced that given an accurate enough simulation of the real world, such AI bots can eventually do the same in the real world – being able to literally replace humans on many tasks in the real world because they can simply learn how to do them better and faster than humans can.
In this world, why would we need humans?
Before you start thinking that I’m advocating for Skynet, we must realize that human created worlds like Dota 2 and World of Warcraft may attempt to mimic the physical world, but they are only a small fraction of it.
Additionally, the other upside for the AI is that they can eventually gain full access to all the data in these manufactured worlds, using iteration and proactive to fully illuminate all the aspects of the world.
One might say that the bots can, in these limited worlds, create a “map” of the world which has 100% of the information about the world and its capabilities/physics. The only reason that they can create a 100% map of the world is that the world is human created and finite, and they are able to run uncountable simulations in that world to map out all the possible permutations of the world.
In this world, they can duplicate it, run it forwards and backward in time, experiment with strategies, and basically explore every possible corner of this world – because it is finite.
The real world, however, is infinite. Even if we blanketed the world with enough sensors to give the AI bots the ability to simulate the world well enough for them to map it, they will still enter and interact with the real world with an imperfect set of information.
This is where the humans come in.
Humans, in combination with AI Bots, can provide the more effective way to deal with the world. In fact, this is happening even now – are we not better humans already when technology reminds us to take our medication, buys our spouse flowers, or helps us to book meetings?
While we live in this infinite real world, AI bots will likely never be able to fully master it, like they do the worlds of Dota2 and WoW, simply because they can’t get a handle on all of it.
While it remains to be seen if simply having a high percentage of real world information is good enough for the bots to be able to take a useful place in the real world, the most powerful entity will likely remain a centaur: the AI augmented human (or some might say, the human augmented AI).
Humans will be useful after all.