Would a company like OpenAi, it the "average" person be able to make a bot, using machine learning, that can outperform the best e-sport players in complex games, like PubG, with the technology we have available today? And if so: what kind of models and approach would most likely be used?

  • $\begingroup$ Sure, take a look at papers published in Transactions on Computational Intelligence and AI in Games journal. If you employ learning, they are usually supervised and reinforcemnt. $\endgroup$ Apr 27 '18 at 7:48
  • $\begingroup$ @Media thanks for the link to the journal! Do you think a supervised learning, where it learn from game footage from an amateur (me) or by learning itself completly, using reinforced learning? $\endgroup$
    – vladasha
    Apr 29 '18 at 12:45
  • $\begingroup$ Actually I didn't understand, can you edit? $\endgroup$ May 1 '18 at 18:37

The general algorithms would be Deep Reinforcement Learning.

One possible approach would be to collect a large number of games. Then train the agent on those games.

Another possible approach would be the agent play against itself, aka self play.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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