Agents in an Environment

The environment is really important to the idea of CABot.
I also think that it's really important to developing real AI.
People learn things from the environment.
We need to have our systems in relatively complex environments, and learning from those environments.
This seems like a strong requiremtn to get real AI.
This is intimately related to the symbol grounding problem.
In the shorter term, it may also make for a really good game.
Instead of being programmed, the system learns from the environment. This should make agents more fun to play with and against.
It also has other obvious advantages with intelligent assistants.
Big world small brain. People (and agents) have small brains but need to cope with a big world.
We're hoping to (say in 5 years) have a game that people will play. The agents in the game can then learn semantics and language by interacting with people.
This will give us an entry into universal grammar, but also into semantics.