Differences between my approach to AI and the traditional symbolic approach:
- My approach produces reasonings with an arbitrary number of assumptions and end goals
- Lisp and Prolog have to be coded by you to work as you intend, but my approach organizes the information and the rules by itself
- Lisp and Prolog use a linear navigation in the solution space (“backtracking”) while my approach is non linear
- My approach is a geometric system
In fact, my approach is so non linear that it works in multithreading already natively. That is, the reasoning is constructed in parallel at the same time and consolidated in real time.
That is why it is scalable to much complex problems, unlike Prolog or Lisp, and that is something you need to get to human level intelligence and beyond.
As for the comparison between my approach and statistical approaches such as LLMs, the engineers’ favourite toy, well… I started using geometry for AI when everybody else was masturbating with LLMs. Today every new paper claims to improve LLMs with geometric approaches.
Draw your conclusions.
Compliments of the Author 