Thinking AI

What would happen if AI developers could focus on creating AI that aims to provide fun for the player? The AI perhaps even tries to *learn* what was fun and what player did not like, instead of so that “easy AI” provides little challenge, and “difficult AI” tries to win in every possible way.

Winning or losing are just one part of the fun, sometimes even meaningless in terms of “what is fun”.

Juuso Hietalahti


  1. This is an interesting idea. I know that with me and my friends, when we play “Supreme Commander” we always agree not to use nukes and such (and usually disable them) since they really aren’t fun and just end the game instantly. When we fight AI and we forget to turn them off, or leave them on because we want to use them later (AI are great to fight because you don’t piss someone off.)

    Another good example is Minecraft. I used to play on Easy/Normal, but it just wasn’t fun for me to get mauled by Skeletons and Creepers every 10 seconds. I just wanted to build stuff, so now I mostly play on peaceful (the exception being the Nether, it is fun with damage and stuff, I guess because the enemies are rarer.)

  2. a big enough pool of beta testers?

    and then using ai to classify player types based on how they play (explorers, competite oriented, social…) and using corresponding “funAI”.

  3. Therefore you would need a non-intrusive feedback system which tells what’s fun and what not. Polling the player every minute if the game provides fun at the moment would be annoying and would kill the fun. Simply counter-productive.

    A “fun sensor” for human emotions would be great, if anything like that is out there yet. Facial expression, voice recognition and brainwave measurement may be a start point for this. Real measurement of fun seems to be as hard as defining fun itself.

    I’m wondering if anybody knows a solution for this?

Comments are closed.