As promised, following is some very insightful advice from the one and only Jeff Orkin on choosing to go for a PhD for anyone wanting to do AI in the game industry. I have certainly been struggling mightily with this decision recently. hi Don,
I'd be glad to offer you any advice that I can. Feel free to ask me questions about AI in the game industry.
I can give you brain dump on my thoughts about working as a "gun for hire", or for a game company, or pursuing a PhD...
The good news is there's lots of opportunity for AI specialists in the game industry right now. Pretty much every game company is looking for help in this area, and finding skilled/experienced people is hard. And none of the existing middleware fully does the job. Graphics guys are a dime a dozen now, and AI is differentiating one game from another, but AI guys are hard to find.
Deciding on the best career route to take depends on your priorities. Are you more interested in innovating, or making money? Do you want to travel much? Do you want to work alone, or be part of a team? Are you more interested in games, or technology?
Most game companies will require that the AI engineer works on- site, to work closely with the animators and level designers. I've only heard of one example of someone doing contract game AI from home -- after Brian Long worked for Monolith on the first No One Lives Forever, he left the company and worked from home on the navigational mesh pathfinding system for "Savage". I think it may have been hard to find additional contracts, and possibly lonely, so now he works for a dot com. It is definitely easier to find contract work with credit on recognized titles under your belt - so you might consider starting by working directly for a game company.
In general, the drawbacks of working as an external "gun for hire" are that you will not share in the profits of the game, and you might work on more isolated, less innovative stuff. I was able to innovate on the FEAR planning system because I had earned the managers' trust by working with the same team on No One Lives Forever 2.
Most games aren't very profitable for the developers anyway though, so you may not be sacrificing much in that department. It's a tough business, and publishers often rope developers in unfavorable multi-game deals before they have a hit. For example, FEAR was the end of a multi-game deal penned before the first No One Lives Forever, so Monolith saw relatively little profit on a game that sold close to a million copies for PC.
When you are "in the trenches" working for a developer, the demands of keeping up with bugs, designer requests, and the competition can consume so much of your time that it's hard to think about long-term innovation. I decided to do the PhD to get some new perspectives, explore some more far-fetched ideas, plus do some traveling and networking at conferences.
Eventually I hope to work with the game industry again, and apply ideas that I develop through research.
A PhD is definitely not needed to work in game AI, but there are a growing number of PhDs in the field now. John Hancock at Lucas Arts did a PhD at CMU. Ian Davis founded MadDoc after a PhD at CMU and then worked as CTO at Activision. Dr. Paul Krezewski founded AI.Implant. Former Media Lab Prof Bruce Blumberg now works at Blue Fang. Etc, etc...
Then there are also lots of academic opportunities for research with games, like obviously ICT, plus similar projects at Stottler-Henke's AI consulting firm.
I think my experience in industry has given me valuable perspectives to apply to my current academic research, ... but I also feel like a pretty old grad student. So, if you are going to do a PhD, you don't want to wait too many years to start -- but at least you've finished the masters.
Wow, this has gotten pretty long. Hope it's helpful!
jeff
Thanks Jeff!