IGC 04 AI Panel - Suggestions
by Josh Williams · in Technical Issues · 09/14/2004 (5:19 pm) · 9 replies
Hello,
I wanted to poll people here on the upcoming AI panel (scroll down 1/3 of the page) we're putting on for this year's IGC. I've seen some pretty detailed, knowledgable posts on AI in this part of the forum, and it's clear that there's lots of interest in the subject.
I just wanted to quickly gather thoughts on the level of expertise you all would like to see targetted in the AI panel's discussion.
Our session is themed on discussing the best techniques Indies can use to create fun, effective game AI given Indie-scale budgets and time constraints. At present, we're planning on targetting the conversation such that non-experts can easily keep up. However, there's still a few days to think about this, and I'd love to get feedback on whether to conduct a session more targetted towards experts specifically.
The panel will be about 50 minutes long. I think either way of going about it will be great fun-- an easy to pick up session would help lots of people, and an expert-level session would be great for digging into the guts of some hefty issues.
I expect that the people who post in this area of the forum will be more inclined to lean toward the hardcore session than most, so I'm not saying that we'll necessarily go with the majority opinion here or anything. But it will be interesting to hear any thoughts you all might have.
If you post below, please state whether you'll actually be attending IGC, and then let me know what kind of expertise you'd like to see the AI panel target.
Thanks! Phil, Nic, and I will appreciate the suggestions. :)
I wanted to poll people here on the upcoming AI panel (scroll down 1/3 of the page) we're putting on for this year's IGC. I've seen some pretty detailed, knowledgable posts on AI in this part of the forum, and it's clear that there's lots of interest in the subject.
I just wanted to quickly gather thoughts on the level of expertise you all would like to see targetted in the AI panel's discussion.
Our session is themed on discussing the best techniques Indies can use to create fun, effective game AI given Indie-scale budgets and time constraints. At present, we're planning on targetting the conversation such that non-experts can easily keep up. However, there's still a few days to think about this, and I'd love to get feedback on whether to conduct a session more targetted towards experts specifically.
The panel will be about 50 minutes long. I think either way of going about it will be great fun-- an easy to pick up session would help lots of people, and an expert-level session would be great for digging into the guts of some hefty issues.
I expect that the people who post in this area of the forum will be more inclined to lean toward the hardcore session than most, so I'm not saying that we'll necessarily go with the majority opinion here or anything. But it will be interesting to hear any thoughts you all might have.
If you post below, please state whether you'll actually be attending IGC, and then let me know what kind of expertise you'd like to see the AI panel target.
Thanks! Phil, Nic, and I will appreciate the suggestions. :)
#2
I would like to make a few suggestions you may have already considered.
An the AI panel, just make sure when your sharing your techniques on AI, make sure you share a few key "do's" and "Dont's" that you know can make a big impact. Good habits and Bad habits start right from the beginning so knowing what not to do can be more valuable in many cases.
About the IGC, has GG considered making some quality Videos of these panel discussions that could be sold for say 5-10 bucks each just like you sell your other products? I'd buy them for sure, and I'm pretty confident others would also. I'd pay the full $195 for a ticket if it takes that, so I sure wouldn't balk at a chance to view the discussions for 10 bucks each. It would not surprise me to hear that people that attended IGC came home and bought them as well. Beat's the hell out of taking notes and carrying around note taking material.
09/15/2004 (12:50 am)
Good God I wish I could be there. I read your FAQ and what was on the agenda and about fainted. If I can figure out a way (very doubtful) I'd be there.I would like to make a few suggestions you may have already considered.
An the AI panel, just make sure when your sharing your techniques on AI, make sure you share a few key "do's" and "Dont's" that you know can make a big impact. Good habits and Bad habits start right from the beginning so knowing what not to do can be more valuable in many cases.
About the IGC, has GG considered making some quality Videos of these panel discussions that could be sold for say 5-10 bucks each just like you sell your other products? I'd buy them for sure, and I'm pretty confident others would also. I'd pay the full $195 for a ticket if it takes that, so I sure wouldn't balk at a chance to view the discussions for 10 bucks each. It would not surprise me to hear that people that attended IGC came home and bought them as well. Beat's the hell out of taking notes and carrying around note taking material.
#3
We can cover a typical "how to get some AI in torque" overview, then move onto the actual practicalities.
I think if we did a sort of seeded "questions and answers" kind of panel thing it might be more useful.
So, in that vein, has anyone got any typical AI problems theyre facing that we can help with?
09/15/2004 (2:33 am)
Ive been thinking about it. How about we go through some typical use case scenarios? Basically run over what kind of things we would want in games, then chat around how those things would best be done (in general and using torque).We can cover a typical "how to get some AI in torque" overview, then move onto the actual practicalities.
I think if we did a sort of seeded "questions and answers" kind of panel thing it might be more useful.
So, in that vein, has anyone got any typical AI problems theyre facing that we can help with?
#4
Just ideas, as I dont think AI animation gets touched upon much really as its not usually just limited to NPC's but in some cases is.
09/15/2004 (3:37 am)
AI animations in torque? i.e skin mesh control.. possible improvments (i.e morph targets?).Just ideas, as I dont think AI animation gets touched upon much really as its not usually just limited to NPC's but in some cases is.
#5
Pros and Cons of the various techniques.
Simple script based AI for TGE.
AND!! This is is something I am very interested in: RTS AI.
-s
09/15/2004 (11:31 am)
I totally agree w/ Thomas: Creative ways of mixing the various techniques and how to not get total mud out of them. Pros and Cons of the various techniques.
Simple script based AI for TGE.
AND!! This is is something I am very interested in: RTS AI.
-s
#6
Still not sure at what level we should present the thing. Sounds like targetting it such that non-experts can easily follow is the best so far, but if there a bunch of super-hardcore, experienced people who'll be in attendance, I'd like to know. If that's the case, then it might be more useful to folks if we target the discussion differently.
Also, please post re: Phil's request above. Post about some common AI problems you've had, and if there's some that really stand out, we'll probably use them as example problem-solution pairs on the panel.
09/15/2004 (12:43 pm)
Cool beans guys. Yeah, the plan so far was to do what has been mentioned most here... discuss good techniques and tricks to get fun AI on Indie timescales and budgets. So, glad people are digging the general theme so far. Still not sure at what level we should present the thing. Sounds like targetting it such that non-experts can easily follow is the best so far, but if there a bunch of super-hardcore, experienced people who'll be in attendance, I'd like to know. If that's the case, then it might be more useful to folks if we target the discussion differently.
Also, please post re: Phil's request above. Post about some common AI problems you've had, and if there's some that really stand out, we'll probably use them as example problem-solution pairs on the panel.
#7
Phil - for me its not a matter of particular implementation problem areas, but lacking some of the experience that you have, I have a problem with figuring our how to fit the pieces together without running into problems. It is very game type specific I think - AI for a turn based strategy game is different than a FPS.
So maybe the best approach could be to disect 2-3 games that you know off. Show the game, explain what techniques were used, discuss how the different AI parts work together and then discuss pro/con about it. Pick some games where they did good things and bad things, and do a panel discussion on what could have been done different and/or in a more indie timescale way to achieve the same result.
I bet most of us AI inexperienced indies over-design the AI or dont do AI at all. So disecting a game and explaining shortcuts to get the same result could show us the "golden middle path" (not sure there is an English expression for that one)
09/16/2004 (12:42 am)
Josh - if you can have the level on par with the "Game Programming Gems" and "AI Wisdom" books, then I think that would be what to aim for. Dont spend time running through source code, and dont explain basics. A golden middle where techniques are discussed and pro/con.Phil - for me its not a matter of particular implementation problem areas, but lacking some of the experience that you have, I have a problem with figuring our how to fit the pieces together without running into problems. It is very game type specific I think - AI for a turn based strategy game is different than a FPS.
So maybe the best approach could be to disect 2-3 games that you know off. Show the game, explain what techniques were used, discuss how the different AI parts work together and then discuss pro/con about it. Pick some games where they did good things and bad things, and do a panel discussion on what could have been done different and/or in a more indie timescale way to achieve the same result.
I bet most of us AI inexperienced indies over-design the AI or dont do AI at all. So disecting a game and explaining shortcuts to get the same result could show us the "golden middle path" (not sure there is an English expression for that one)
#8
I'd rather we looked at ways of improving torque-users games than an in depth review of commercial games techniques, but some middle ground is definitely good.
I guess maybe we should choose a few genre's to discuss the potential AI of, then how we would apply the typical AI techniques to the specific genre.
So that begs the question, which genres?
RTS - seems reasonable (Josh?)
God/Sim game (Phil)
CTF FPS game (Nick?)
So ideally we want an overview of the problem area's, then a brief discussion about what techniques are commonly seen in those area's, followed by a walkthrough of the problems of a specific genre and how the techniques mentioned apply for each genre?
Seem reasonable?
09/16/2004 (1:09 am)
Thomas: like it.. Josh, whatcha think? we could certainly have a look at some typical game examples out there. I'd rather we looked at ways of improving torque-users games than an in depth review of commercial games techniques, but some middle ground is definitely good.
I guess maybe we should choose a few genre's to discuss the potential AI of, then how we would apply the typical AI techniques to the specific genre.
So that begs the question, which genres?
RTS - seems reasonable (Josh?)
God/Sim game (Phil)
CTF FPS game (Nick?)
So ideally we want an overview of the problem area's, then a brief discussion about what techniques are commonly seen in those area's, followed by a walkthrough of the problems of a specific genre and how the techniques mentioned apply for each genre?
Seem reasonable?
#9
Those genres sound good. Might be tough to cover so much in our timeframe, but we can plan that out. I don't know if we need to just have one person discuss each genre. I figure each of us will have interesting things to say in each.
This is sounding great.
To note: preparing vids or demos of various games is probably too-high overhead for me at this point. Unless one of you guys wants to charge ahead on such an effort, I think we'll probably stick with screenshots or some such (though it would be cool to show games in action).
09/16/2004 (2:26 pm)
Sounds just about perfect. I'm glad to hear that folks are interested in an AI Prog Wis-level discussion as well. Good stuff (and it just so happens that Phil is an excellent contributor of articles to the AI Prog Wis series, so that part should be easy for him ;)Those genres sound good. Might be tough to cover so much in our timeframe, but we can plan that out. I don't know if we need to just have one person discuss each genre. I figure each of us will have interesting things to say in each.
This is sounding great.
To note: preparing vids or demos of various games is probably too-high overhead for me at this point. Unless one of you guys wants to charge ahead on such an effort, I think we'll probably stick with screenshots or some such (though it would be cool to show games in action).
Torque Owner Thomas \"Man of Ice\" Lund
Personally I would like the panel not to run through basics like "What is a FSM, what is a blackboard, how to draw state diagrams" and similar. Basics is all right, but please keep it practically related to how to create fun games.
I would personally love if the panel could share some wisdom on how to mix and match various techniques for creating the "magic AI illusion". Noe of us indies can adfford to code AI for 2 years, so how can I use e.g. a blackboard coupled with very basic pathfinding tricks to create an illusion of a fun and challenging AI - if yoy know what I mean. The AI Wisdom 2 books has a perfect example of this, where a NOFL2 developer wrote about the cheap tricks they used to fake squad level AI without actually implementing any.
Its tips and tricks like that, that I personally can use in getting a more fun game with less effort.