There isn't anybody working on this as their 'job.'
Nonsense. There's lots of research groups playing with this sort of stuff.
A friend worked on Semantic Webs back in the 70s. It didn't work then, either.
I know people doing interesting things with Neural Nets but there are severe problems with upscaling as soon as they run into Huble and Wiesel limitations: no such thing as the "Grandmother neuron."
The CYC Project is a joke. They've been Relationally Databasing for over 20 years now with little to show for it.
Wolfram's "Computation Knowledge Engine" is more interesting but it still requires human to tell it what to think. CAs are too limited a tool to handle n-dimensional referents and ambiguity.
I could go on.
So ... who?
