Archived

This topic is now archived and is closed to further replies.

Ferretman

November Game AI Poll

Recommended Posts

Howdy All: A new month means a new poll over on the GameAI page. This month''s poll was suggested by a regular visitor who wonders just how much experience and education game AI folks might have in the field. It poses the question, "Have you ever taken any AI-related classes? If so, at what level?". Could be good fodder for the 2003 GDC AI roundtables next year. I''d be curious to hear if AI classes have been an important part of your experience in the field.... Ferretman
ferretman@gameai.com www.gameai.com From the High Mountains of Colorado

Share this post


Link to post
Share on other sites
Everything I learned about AI, I pretty much learned on my own. Well, I asked hundreds of questions in newsgroups and on message boards, but I never took any classes. Books weren''t really a factor either.

A friend of mine took the AI class at my college, and from the way he talked about it, it sounded pretty boring. He said they had to write a search engine for a website that would do some "intelligent" thing, like give you advertising based on what you search for, or something like that. I forget exactly, but it was kinda lame. That sounds like a project for some other boring class, not AI class.

I''m no expert, but my experience has been that there seem to be a few decent universities that support AI research and probably have good classes on it, and then there is everyone else who either doens''t have anything about it, or added it in as an afterthought. It also seems like pretty much everyone I deal with in computer chess circles was more or less self taught through trial and error and online resources. Hardly ever formally educated. The only people even related to formal AI education are a handful of professors I know online, but while they teach it formally, they basically taught themselves, and were battling in the trenches writing chess programs in the 70''s.

I think part of the reason for this is because AI is such an unexplored field. There is a more or less standard way to do graphics, sound, input, etc., but not AI. Good AI is whatever works. Without some kind of competition there is no motivating factor to improve AI. If you make AI for your game and tell yourself it''s good enough, then it might suck, but to you it''s good enough, and it''s not like anyone is going to have a contest between your bots and Quake''s bots (or whatever). That''s why I enjoy computer chess. There are all kinds of competitions between computer chess programs, so the computer chess AI is more advanced than say, FPS bot AI. If you could play your bots against Quake bots and your bots got creamed, you might think again about improving your AI. But such is not the case, so AI is usually an afterthought, because as long as it does something remotely resembling a goat''s intelligence level, that''s "good enough" for most people.

Share this post


Link to post
Share on other sites
I have for some reason always been interested in AI (for as long as I remember)

I wrote some stuff on my own before I went to the university (Uppsala, Sweden) where I read a special branch of the comp-sci programme. There we studied mostly traditional AI, implemented in Prolog, but also glanced at other stuff, like NN and GA.

It was an interesting subject to study - we went through a lot of stuff that I''m sure I wouldn''t have looked at otherwise. I think the academic approach is pretty different from that of a more ''practical'' one - you look at techniques in a context in another way, and get more of a ''whole'' view than when you just dig into whatever it is that you happen to need at the moment. - A good complement to the ''pragmatic'' approach

A quote: "Once an AI techique works all the time, it''s not AI anymore - just good engineering"

-----------------------
Always down - Never out

Share this post


Link to post
Share on other sites
quote:
Original post by Russell
Everything I learned about AI, I pretty much learned on my own. Well, I asked hundreds of questions in newsgroups and on message boards, but I never took any classes. Books weren''t really a factor either.

A friend of mine took the AI class at my college, and from the way he talked about it, it sounded pretty boring. He said they had to write a search engine for a website that would do some "intelligent" thing, like give you advertising based on what you search for, or something like that. I forget exactly, but it was kinda lame. That sounds like a project for some other boring class, not AI class.

I''m no expert, but my experience has been that there seem to be a few decent universities that support AI research and probably have good classes on it, and then there is everyone else who either doens''t have anything about it, or added it in as an afterthought. It also seems like pretty much everyone I deal with in computer chess circles was more or less self taught through trial and error and online resources. Hardly ever formally educated. The only people even related to formal AI education are a handful of professors I know online, but while they teach it formally, they basically taught themselves, and were battling in the trenches writing chess programs in the 70''s.

I think part of the reason for this is because AI is such an unexplored field. There is a more or less standard way to do graphics, sound, input, etc., but not AI. Good AI is whatever works. Without some kind of competition there is no motivating factor to improve AI. If you make AI for your game and tell yourself it''s good enough, then it might suck, but to you it''s good enough, and it''s not like anyone is going to have a contest between your bots and Quake''s bots (or whatever). That''s why I enjoy computer chess. There are all kinds of competitions between computer chess programs, so the computer chess AI is more advanced than say, FPS bot AI. If you could play your bots against Quake bots and your bots got creamed, you might think again about improving your AI. But such is not the case, so AI is usually an afterthought, because as long as it does something remotely resembling a goat''s intelligence level, that''s "good enough" for most people.



I''m pretty much the same way myself...self taught on nearly everything, usually through web sites or the old trial-and-error method. I never took any classes in AI per se, though I note that my database structures course has been neigh-invaluable for a lot of my work.

Anybody who asks me how to get into AI, I always recommend the FPS shooters and their ''bots. It''s a great way to focus on the AI per se without having to worry about all of that graphics and sound overhead, and theres a lot of examples and help out on the Web in the various forums and the like.





Ferretman

ferretman@gameai.com
www.gameai.com
From the High Mountains of Colorado

Share this post


Link to post
Share on other sites