How To Make Robots Talk. Stage 3 (possibly): Multiply
So now we can reproduce scripts word by word — but only if we select a script beforehand for each interaction. Surely we want an auto-selection of scripts at runtime, so the robot can react to any known input any time.
Let’s do that next.
What do we do? We combine multiple “super precise” engines at once, each acting on one allowed input. We match their output together and see what we got there. Do all sub bots offer the same opinion? Fine, print it out. One disagrees? Fall back to outputting “????”, in the spirit of “super preciseness”.
This engine shouldn’t be hard to make. We need to turn the “super precise” engine into a Java(X) class so it can be reused. Then make a couple engines and work them together.
Huh. I thought about it some more and found out another thing is needed first to make things much clearer: a prediction architecture.