Levels of Abstraction
François discusses the evolution of abstraction in machine learning, emphasizing the distinction between organized knowledge and strong generalizable models. He highlights that current LLMs, while knowledgeable, still fall short of true intelligence, which requires the ability to autonomously generate abstractions. The conversation delves into two forms of abstraction—value-centric and program-centric—illustrating how analogy-making drives cognitive processes in both humans and machines.In this clip
From this podcast
Machine Learning Street Talk (MLST)
It's Not About Scale, It's About Abstraction - Francois Chollet
Related Questions
Give an example of the steps in the intelligence definition from the episode "It's Not About Scale, It's About Abstraction - Francois Chollet": "A big part of intelligence is the process of mining your experience of the world, identifying bits that are repeated, and extracting these abstractions. We can use these abstractions to navigate novel situations."
Give an example of the steps in the intelligence definition from the episode It's Not About Scale, It's About Abstraction - Francois Chollet: "A big part of intelligence is the process of mining your experience of the world, identifying bits that are repeated, and extracting these abstractions. We can use these abstractions to navigate novel situations."
Give an example of the steps in this intelligence definition from the episode It's Not About Scale, It's About Abstraction - Francois Chollet: "A big part of intelligence is the process of mining your experience of the world, identifying bits that are repeated, and extracting these abstractions. We can use these abstractions to navigate novel situations."