Just gonna put my two cents in cuz this seems fun:
If you were to use an neural based AI opponent, the AI would have to adapt to (for example) your play style. However, all play styles don't stay stagnant due to natural adaptation. As you adapt to the AI adapting to your current play style, the AI has to follow suit. It becomes a game of who can adapt the fastest. This means, from my understanding, that the AI needs a good neural net that can keep up with human adaptations. This, although can happen, will need a lot of process power and fine tuning.
If you were to use an neural based AI opponent, the AI would have to adapt to (for example) your play style. However, all play styles don't stay stagnant due to natural adaptation. As you adapt to the AI adapting to your current play style, the AI has to follow suit. It becomes a game of who can adapt the fastest. This means, from my understanding, that the AI needs a good neural net that can keep up with human adaptations. This, although can happen, will need a lot of process power and fine tuning.
Last edited by Lotus; May 18, 2019 at 04:30 AM.
Reason: <24 hour edit/bump
¿?
thats not exactly what i meant,
this is a good way to do it only if you are going for some ultra hardcore shit,
what i meant is that some players will play agaisn't the ai until it adap to them to a certain degree, the difficulty can be controlled by stopping the games and so the evolution.
the ai is not going to learn nor adapt in real time, it doesn't have to.
i think letting the ai learn how to behave in certain situations by itself will be much easier than hardcoding the inputs for each individual situation.
correct me if i'm wrong since i don't know much about this subject.
this is a good way to do it only if you are going for some ultra hardcore shit,
what i meant is that some players will play agaisn't the ai until it adap to them to a certain degree, the difficulty can be controlled by stopping the games and so the evolution.
the ai is not going to learn nor adapt in real time, it doesn't have to.
i think letting the ai learn how to behave in certain situations by itself will be much easier than hardcoding the inputs for each individual situation.
correct me if i'm wrong since i don't know much about this subject.
Aadame:I'm very signaturable
It's just no one usues my shit .
It's just no one usues my shit .
Originally Posted by Alpha
[...]a player will fight agains't the ai, the ai's objective is to protect itself and kill the player, as the ai evolve it will learn to block and then learn to attack, with more generations it will learn to read it's opponent and do chain attacks.
ofc, the evolution need to be stopped before the ai become impossible to beat.
[...]
That's exactly why people hard code enemy AIs xD. But yeah I will try to make some projects and surely will post some updates here too.
-----
More about the fighting game example - the beautiful thing is that well developed neural net with good usage of hidden nodes is more than able to adapt to many playstyles with one evolution process.
But nah it's not easier than hard coding.
Last edited by Aadame; May 19, 2019 at 11:50 AM.
Reason: <24 hour edit/bump
