A Google AI System Can Now Beat You at Videogames on Atari 2600
Well, everyone among the big companies in the technology market are working, in one way or another, on artificial intelligence and how it can change our lives. Take the idea of driverless cars for example.
While this one might not be as life changing as a driverless car (at least at this point), but Google has invented a new AI system that has been taught to play videogames.
Remember when Google acquired the artifical intelligence startup DeepMind? This is the first project that they have worked on. This AI system we are talking about is still in its early stages but even at that, it has learnt how to play as many as 49 games on the classic Atari 2600.
Out of those 49 games, it has been able too beat the scores of 43 human players too!
I know the games on Atari 2600 were just basic and that can, in no way, compare to today’s 3D games with complex worlds, but it is a continuing process.
When the project was initially started, the AI system couldn’t even pass the first levels of Breakout. After a couple hundred attempts however, it has mastered the game. Same is the case will all the other videogames it has been taught.
Demis Hassabis, vice president of engineering at Google and co-founder of DeepMind stated this was the first ever step of its kind because it is the “first time anyone has built a single learning system that can learn directly from experience and manage a wide range of challenging tasks.”
Obviously, this is the first step on a long journey to a time where maybe this AI system could start playing games like Forza Motorsport 5 and maybe someday start driving our real cars.
It’s mastering and understanding the structure of these games, but we wouldn’t say yet it’s building conceptual knowledge or abstract knowledge. The ultimate goal here is to build smart, general-purpose machines, but we’re many decades off from doing that.
For now, it is learning form the environment and its past experience. In a given situation it can go back to past events and correct the mistakes it made – on a very basic level of course.
It would be great to see how in the coming months or years Google turns this AI system into something more complex with the help of DeepMind.