Mainly, I have a question that I could not find an answer for anywhere, about an AI machine learning game.

Is it possible for me to create an AI which can learn a game that I did not myself create, without any source codes or APIs? And I'm talking about simple games such as chrome t-rex, snake, etc. I was told this is possible using image processing but I am still not so sure. I am referring to just open up a game, not a downloaded one, a one available at google, such as the mentioned above, run my code, and it will start playing and learning the game. Just creating 1 algorithm to learn multiple easy 2D google games, is such a thing possible with image processing?


1 Answer 1


As far as I know you might archieve this with Reinforcement Learning. But there are some limitations in this approach: You have to train your algorithm a long time for even learning to press a button. The second limit is that you have to define the possible actions, that the agent might choose. And remember: Every single thing your system has to learn will affect your training time.

You can dive into this topic by searching for the openai gym, which will give you a good overview about what is done right now in order to build AIs which play games. A good book about this topic is Reinforcement Learning - An Introduction by Barto and Sutton.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.