1.) Do you see AI constructs such as neural networks and genetic algorithms becoming a part of most new games in the near future? Or do you envision another approach? Yes I do. I think that NNs and GAs are the missing pieces that we have been lacking in a lot of game AI, search, and learning algorithms. I think that when everyone learns how/when/ to use them, then you are going to see many more games that aren't dumb. I believe that we are looking at 3 years max before all games are "smart". 2.) What approaches do you take when programming AI in a game. I use whatever is appropriate. Usually for the "high level" objects in the game I use finite state machines coupled with probabilistic decision trees along with fuzzy decision making. I perform many searching algorithms using GAs and I have tried using NNs for pattern recognition etc., but the design time of AI based on NNs is fairly high, but it has good payback when doing games that you want to "teach" rather than program. A good example is using a NN to learn responses to fighting moves of another player. The position of the opponent can be thought of an input vector and then used as an input to a NN. 3.) What's your favourite game in terms of AI. Why? I would have to say Half Life, it has a gamut of AI techniques, and I have never played a game that seemed like it was really planning (other than strategy turn based games). Very impressive. 4.) Where do you think the artificial intelligence in gaming will be in 5, 10, 20 years? In 5, all games will be smart. The tools to make games that can outplay a human always will be available. Also, artificial life will be much more prevalent and autonomous, learning agents will exist in a number of areas. In 10 years, I believe that we will have the hardware and software to make a machine problem solve at least within a specified domain as well as we can. I think that massive "brain" models composed of billions of neurodes and other virtual circuitry and algorithms will exist. In addition, I think that the first bipedal, intelligent robots that can navigate their environment as well as communicate will exist, maybe not commercially in large numbers, but still. In 20 years I believe that without a doubt we will be able to link to a machine with some kind of electromagnetic interface. I believe that our brain will be sufficiently fooled by the simulation and the AI of the virtual entities inhabiting the space will be on par with our own. I mean, a toy -- the playstation II is absolutely photorealistic. At our current rate of technology in 20 years, our computing power should be a minimum of 220 greater, since we roughly double are computing speed and storage every year. And when we have reached theoretical limits of electrons and materials like Gallium Arsenide etc. then I'm sure that photonics and massively parallel machines will help with the computing speed. Not to mention supercondutivity at room temperature (not that far off) and generation of the theoretical particles Tacheons used for electronics might be the key to faster than light computations. There is no such thing as science fiction anymore — the technological discoveries we will make on our own along with those that computers help us make, will make my wild claims — seem like 8086 stuff :) 5.) The gaming field of computing is probably one of the most quickly advancing areas. It forces new software and hardware to be developed, and is continually on the edge of computing technology (especially in terms of graphical technology). Do you see gaming having the same effect on artificial intelligence at some point too? Absolutely. I used to work on top secret projects at NASA and many game programmers are X weapons designers, AI modelers for missiles etc. Basically, game programmers want the games to be real and once they really get their teeth into AI then the accedemic community won't be the only people working in the area. 6.) Are you interested in any other fields of AI besides that of gaming? Why? Yes all of them. Physics, math, electrical engineering, biology, mechanics,medicine, it's all part of the same thing — I want to make a machine think, a virtual universe, and then dump myself in there along with 5 billion other people and have a real DEATHMATCH. All of these sciences and more are pieces to the grand puzzle. I think that game programmers are the most "religious" people in the universe, we are trying to model the universe, and imitation is the highest level of flatery. Not only that, but what interests me is this: "I am a conglomerate of atoms 20 billion years old. These atom have formed into computing structures. These computing structures are aware of their own existance and now are trying to replicate themselves. Crazy when you think about it!" |
Andre LaMothe is CEO of Xtreme Games LLC, and has authored many books on game programming. He has degrees in Mathematics, Computer Science, and Electrical Engineering. He has been in computing industry for 20 years and has worked on 2D/3D graphics, Artificial Intelligence, Compiler Design, Robotics Engineering, Virtual Reality, and Telecommunications. He was also the inventor of one of the worlds first virtual reality games.