Hey I have worked on minimax and pathfinding in the past, some of A* as well. What should I try working on next?

Why don't you build a game and try to put it all together.

Neural nets, Fuzzy Logic, Hierarchical Task Networks, Behavior Trees, Rule-Based Systems, Goal-Oriented Action Planning...
The sky's the limit.
AI is an ever evolving field with alot of potential for new applications and discoveries.
I guess the kinds of games you want to make should help to narrow down your choices for what to focus on.
Good luck!

I know... I just want some context to start...

Cause if I don't have some AI in mind I might make a game and do the AI myself without studying something first...

I'll check the ones delarosa included one by one and see which I like best and try it I suppose :)

Quote: Original post by Tipotas688
Cause if I don't have some AI in mind I might make a game and do the AI myself without studying something first...

That's arguably the better route. It'll certainly be easier to keep motivated and focus your study to only those things that you need for game AI (which is basically mostly A* and state machines) :)

-me

yeah that's why :P

I want to learn new stuff and do them "properly".

Last time I tried a pacman with wide rooms and decided A* is
too expensive so instead of modifying A* I tried some linear
if statements which worked as fine... Didn't always find the
correct path but I didn't want it to anyway...

I think it'd be better to first know the ways I can do something
and then decide if I want to do it my way else my judgement is
flawed, as for motivation, if I was to motivate myself every time
I had to learn something, I'd never get to get my MSc :P

Quote: Original post by Tipotas688
Last time I tried a pacman with wide rooms and decided A* is
too expensive so instead of modifying A* I tried some linear
if statements which worked as fine... Didn't always find the
correct path but I didn't want it to anyway...

Which is basically how pacman was actually programmed [smile]. The "AI" for pacman was simply about which way to turn at a given intersection, there was no proper pathfinding.

But yeah, there are definitely things that you could learn that would be useful. Just depends on what your goal is. If it's just learning things, then your learn things is definitely the correct approach for you :)

-me

hehe it is! that's why I keep asking people to tell me what to learn cause there are so many things out there, I dunno which is relevant and which is not...

In my pacman it had seeing, hearing, memory and pathfinding was quite fun :)

Try just developing simple context-dependent decisions. i.e. FSM state transitions. That is far more "AI-like" than Minmax or A*.

How about time-series analysis?