What’s missing from today’s AI — Part 2

… the missing link since the 1930s to usher in “Next Generation AI”

A future robot, waiting for a bus, passes the time by catching up on its reading. Here the bot worries that his hands are inadequate for the job of turning pages, as humans do, because he’s only got 2 weird plastic digits!! Photo by Andrea De Santis on Unsplash

The story so far

Part 1 of this story introduced the need to represent knowledge for AI and used the modern concept of a knowledge graph to illustrate the limitations that computer science has imposed on its design. Computer scientists that write programs don’t need to worry about how to represent meaning, since they can always write computer code to convert it from data. But a brain is unlikely to work this way since it is comparatively slow, suggesting that the model used for AI needs to be more efficient.

Next Generation AI will undo the maniacal focus on syntax (i.e. per Noam Chomsky) to restore the popular models of the early 1900s, semiotics (i.e. per C.S. Peirce) — the science of signs that link words and phrases in a language to its meaning in context.

Today’s article concludes the argument for efficiency by contrasting (a) today’s knowledge representation that aligns with modern computer programmer’s designs and (b) cognitive science as illustrated with Role and Reference Grammar (RRG), a linguistic framework that has been mapping language to meaning. It has been under constant development and improvement by linguists across the world for decades with indispensable…