It’s a fact that electronic brains are more and more prevalent in our world, doing all kinds of complex and repetitive work. One aspect of computer brains gaining in popularity is machine learning and artificial intelligence. They aim for life-like characteristics in machines. That could come in handy in games, since we all know how predictable pre-programmed AI can be. Well, that’s precisely what Blizzard and Google are working on, combining DeepMind and Starcraft 2.
Strategy games like StarCraft 2 demand logical thinking and focus to always be ahead of the curve. When you’re playing online, it’s a challenge because you’re playing against other human beings and not against someone that’s not pre-programmed.
On the other hand, if you’re playing against a computer, the only difficulty you can have is in the number of AI you encounter and their aggression and precision. Everything else is just code-based repetition and, often, following a set path.
A Brief About Artificial Intelligence
Artificial intelligence is nothing new. It’s been around for ages. Well, not literally ages, but since the invention of the first electronic computer in 1941. The first step towards modern AI, however, was in 1955. Norbert Wiener said that machines could, in theory, simulate intelligent behavior via feedback mechanisms. The theory sparked the creation of the first modern AI program, The Logic Theorist, designed in 1955 by Allen Newell and Herbert Simon.
In spite of the early efforts, the first person to do something noteworthy about AI was John McCarthy. In 1958, after several intelligent programs created by others, all based on Wiener’s theory, LISP(LISt Processing) finally saw the light of day. After the release of the revolutionary language that’s still in use today, John McCarthy became the father of modern AI.
In the case of DeepMind, they’ve taken things to a whole new level. Starting in 2010 as a small company doing AI research from their London office, DeepMind became a part of Google’s Alphabet group in 2014. Countless hours have been invested for a goal that’s truly something else; “We’re on a scientific mission to push the boundaries of AI, developing programs that can learn to solve any complex problem without needing to be taught how.”
The central principle behind artificial intelligence is machine learning. You know how you learn to do things by watching tutorials? That’s precisely what machine learning is, except it’s teaching the machine how to behave like a human would in certain situations.
Learn From Humans, Act Like Humans
In the case of our artificially intelligent DeepMind and StarCraft 2 collaboration, DeepMind initially had a pool of about 65,000 matches, which their bots used to learn how to behave. After that, Blizzard updated the pool with about half a million matches per month.
The results of the two years of the in-progress collaboration saw the light of day at the 2018 edition of BlizzCon. DeepMind’s bots managed to learn enough, to give them a 50% chance of beating Blizzards’ insane level AI. Blizzard also noticed that the bots often immediately worker rushed their opponents, which is a move popular with human players.
Since artificial intelligence is almost like the human brain, it’s learning curve features a geometric progression. The more it knows, the more it can learn. Blizzard chairman said: “The StarCraft games have emerged as a ‘grand challenge’ for the AI community as they’re the perfect environment for benchmarking progress against problems such as planning, dealing with uncertainty and spatial reasoning.”
I agree with this because bots can learn how humans think much faster than in any other environment. Games, especially online multiplayer strategies, are incredibly unpredictable. Our brains have to work extra hard to make the most out of a specific situation.
It looks like DeepMind and Blizzard made some considerable progress with this project. Stop by either their YouTube channel or StarCraft’s Twitch stream tomorrow at 10 am PT time/1pm ET for the update.
Two Sides Of The Artificial Intelligence Story
As with everything about machine learning, we have the mandatory two sides of the story. Reading through people’s opinions on implementing the technology, the all for it or all against it vibe is noticeable. There’s no middle ground.
I think that, if used as intended, the gaming industry of the future could benefit significantly from this. Just imagine having a game that is intelligent and the players are like people. The benefits are endless. Feel free to leave your views on the subject in the comment section.