logoalt Hacker News

drugstorecowboytoday at 8:18 PM1 replyview on HN

In my completely unqualified and baseless opinion, I think the reason machine learning in games isn't more popular, is that ruthlessly difficult enemies that learn and adapt to the player just aren't.. fun. Level playing fields between characters and enemies just don't work. For example, the Dark Souls series is generally known for being "difficult but fair", however enemies are not smart by any means, they ultimately have a very limited number of attacks and very basic AI that consists of "Wait until player is in range, Run in a straight line toward the player until in attack range, attack". Even with that simplistic AI its pretty difficult and enjoyable.The player is meant to triumph over hundreds of enemies, its the player that learns and adapts to a static set of obstacles. If enemies in Dark Souls learned and adapted to the player, it would go from difficult to impossible very quickly and would not be enjoyable.

The purpose of AI in games isn't making something thats impossible to beat, its making something that is ALWAYS possible to beat while remaining fun and engaging.


Replies

luke5441today at 9:58 PM

There are enough people playing multiplayer games where this is not the case. Those are actually kind of annoying because ELO systems cause you to always loose about half the games (except if you are the top player of all the players) no matter how much better you get.

You could tune this to handicap the LLM such that the player wins 80% of the time or so and it should be more fun (except for the competitive aspect).