News for hackers. Stuff that matters.
Re:Competitive Self-Play by indescions_2017 on Wednesday October 11, 2017 @04:41PM (#15453272)
It seems like fighting games such as Street Fighter or Tekken are a perfect fit for Self-Play.

Anyone at OpenAI attempted to build such an agent? Are there any AI research platforms designed specifically for player vs player fighting games? As far as I know, elite human players are still massively dominant. Even though it would make for an exciting matchup. But giving the complexity of actual fighter competition, with combo attacks, power meters, time limits, etc. There is an absurdly high dimension of training variables required.

I'd actually like to try and take a step back and apply self-play to something lower dimensional. Perhaps a 2D Tron Light Cycle sim. And see if some truly unexpected strategies arise ;)

Re:Competitive Self-Play by logent on Wednesday October 11, 2017 @05:02PM (#15453466)
Given perfect information, execution, and instant reaction times, fighting games would be trivial for an AI to win at. There are already some bots that will do things like auto-block or parry any incoming attack in Tekken and they're pretty funny to watch (see TOOLASSISTED's videos like this one: https://www.youtube.com/watch?v=nNG5iRMdeg0 )

At high levels of play, where execution largely isn't an issue, it's all about reading your opponent, conditioning them to play how you want to play, and exploiting their tendencies to get in your damage. It would be pretty cool to see a bot learn to play with a human-level reaction time handicap and put them up against a pro in a long set, though!

Re:Competitive Self-Play by indescions_2017 on Wednesday October 11, 2017 @09:44PM (#15455070)
Actually your comment just triggered a sense memory for me. Of a rudimentary "classic" AI for a prototype 2D cell-shaded bitmap sprite sheet fighter game I created more than a decade ago. I think it was in adobe flash and actionscript 3...

And yeah my bot "cheated" extensively ;)

Re:Competitive Self-Play by bingojess on Wednesday October 11, 2017 @08:49PM (#15454834)
For their dota 2 bot, they included human reaction times if I recall correctly.
Re:Competitive Self-Play by dragontamer on Wednesday October 11, 2017 @05:39PM (#15453726)
> I'd actually like to try and take a step back and apply self-play to something lower dimensional. Perhaps a 2D Tron Light Cycle sim. And see if some truly unexpected strategies arise ;)

Dude needs to google "Corewars Genetic Algorithms".

http://www.corewars.org/docs/evolving_warriors.html

Corewars is just one example of an game that AI dudes have been playing for... a long time, decades at least.

-----------

Anyway, Fighting Games are mostly about reaction speed and execution. Perfect play in a fighting game is ridiculously super-human due to technique (1-frame links) and reaction speed that there's no way a human would beat any serious bot.

Most bots made aren't serious though.

Re:Competitive Self-Play by binarymax on Wednesday October 11, 2017 @04:51PM (#15453352)
The research here is more transferable to use in the real world. There is lots of other research in the video game area and it's been going on for awhile [0].

These revelations by OpenAI could potentially be adopted to a physical robot or device. Not sure we can do that with AI Ryu perfecting hadoken :)

[0] https://arstechnica.com/gaming/2013/04/this-ai-solves-super-...

Re:Competitive Self-Play by canjobear on Wednesday October 11, 2017 @05:26PM (#15453635)
> As far as I know, elite human players are still massively dominant.

I wouldn't be so sure: https://arxiv.org/abs/1702.06230