How Will Semantic AI Work with Math?

 

To the manner born. They are both made of exactly the samer stuf - variables, constants, operators and links in an undirected network.

Language is much more sinuous than math. How do we know that? A long time ago, we built an analytic system, with numbers, algebra and logic in an undirected network, but it did not include existential control. One or two algebraic statements are easy, but a hundred of them all interacting together is hard to think about - it was hard to use. As soon as we added language, with all its variety,  existential control became essential. With language, a physical object inherits physical attributes – a car has a mass, dimensions, moments of inertia about different axes, a colour (don’t worry, if you don’t give values, it will inherit typical values). Language is being used to describe physical and abstract objects, so it becomes trivial to connect the attributes of those objects through formulae. What Semantic AI does is turn text into a working model – especially useful when the text runs to hundreds of pages and the objects in physical or abstract motion exceed the human’s limitation of four objects in input, the rest being treated as constants. Humans are quite brilliant at reducible problems (like designing a mobile phone or an aircraft), not so good at complex strongly interrelated problems (like Climate Change or a large piece of legislation).

There is a world of difference between Semantic AI and LLMs, – one faces up to the fact that its knowledge base has to be large and detailed, the other one is trying to squeeze more out of the technology than it can possibly give, with no understanding of meaning.