A very good piece! It seems that we need a new breed of AI - different from the generative one, different from statistical one. What about the one based on differences and differentiation, comparisons and filtering, as a new computational paradigm? Think about the game "20 Questions" or Venn diagrams - they narrow down on the most fittin…
A very good piece! It seems that we need a new breed of AI - different from the generative one, different from statistical one. What about the one based on differences and differentiation, comparisons and filtering, as a new computational paradigm? Think about the game "20 Questions" or Venn diagrams - they narrow down on the most fitting candidate rather quickly.
You have influence, you are familiar with researchers, investors and policy-makers, why don't you step in and control the whole process? AGI potential is there but to implement it properly it will take efforts of more people. We don't need hype, we need a working thing. You will make it work. But it will be different AI, not generative one.
A very good piece! It seems that we need a new breed of AI - different from the generative one, different from statistical one. What about the one based on differences and differentiation, comparisons and filtering, as a new computational paradigm? Think about the game "20 Questions" or Venn diagrams - they narrow down on the most fitting candidate rather quickly.
There are two ingredients to the solution - my approach discussed here https://alexandernaumenko.substack.com/ and sensorimotor primitives discussed here https://dileeplearning.substack.com/p/ingredients-of-understanding
You have influence, you are familiar with researchers, investors and policy-makers, why don't you step in and control the whole process? AGI potential is there but to implement it properly it will take efforts of more people. We don't need hype, we need a working thing. You will make it work. But it will be different AI, not generative one.
What you are looking for seems to be... decision trees.
Pretty much ... more. I add comparable properties and ranges. I describe how that approach handles natural languages naturally.
Read Bateson 1972 - he is in love with differences.
Generalization out of the box. Even with flavors.
Applicable to all modalities.
Decision trees were not explored to the maximum.