As with other areas of cutting edge research, most non-experts often have little choice but to resort to woefully inadequate simplification. 'Any sufficiently advanced technology is indistinguishable from magic', and all that.

The best way I've come up with to explain the current state of AI is, "we can do tricks by hiding multiple immense databases behind the scenes and having supercomputers run complex searches among them, but we probably won't be able to fully model even a mouse brain for at least 10 years... and at that point we'll likely have learned enough new things about brain function to push it another 10 years out."

To some degree, but if somebody wants to learn about AI in 2016, they probably want to learn current techniques like deep learning. Simple rule based systems like this seems to support aren't going to be very useful to other problems.

I think the grandparent has a valid point. IMHO learning is a necessary feature of any intelligence, artificial or natural. That's why modern AI is 90% Machine Learning. All non-learning approaches (frames, symbolic AI) have failed 20 years ago.

How to make an easily approachable tutorial that combines game-like world and real learning AI (for example Q-learning or maybe evolutionary policy search) remains and open question. Maybe someone will build it.

> 'Isn't all implementations of AI a combination of "a simple xyz"?'

Not really, unless you count mathematical techniques such as gradient descent optimization or convex optimization as simple. In this case a finite state machine sounds very simple in comparison.