We recently enabled some new technology at Microsoft that automatically translates voice messages into text and then sends you the text as email. The audio is included as an attachement. Now whenever somebody leaves me a phone message, it’s the same as if they typed in the message and sent it as email. At least that’s the theory.
My daughter recently left me a phone message. The resulting email I received started: "Hi Dad, this is Willie Sharp," although the voice attachment clearly said "Hi Dad, this is Alisha." I joked about it when I called her back. The next day Alisha called and said, for fun, "Hi Dad, it’s Willie Sharp." The resulting email began: "Hi Dad, it’s 20 Shot."
Humorous, yes. The technology isn’t quite ready for prime time. Will it ever be? Maybe. What the technology is missing is context. There’s no database that tells the software that Pete’s daughter is Alisha, spelled s-h-a, not c-i-a. There’s no algorithm in place that says a person in a phone conversation usually starts out by saying their name, and that "20 shot" does not remotely resemble a name.
With game programming, we run into the same types of problems with technology — no context, or not enough context. The result: humorous, if not very lifelike, results. NPCs that say off the wall things. Irrational enemy behaviors. Willie Sharp instead of Alisha.
Do games need context databases? Better algorithms? Should we create games that avoid emulating the real world altogether? Or should we just focus on multiplayer games so users are always dealing with human intelligence? There’s no right answer. We may just have to put up with Willie Sharp for a while.