Ever.
Whatever system will be able to do this will not be an Auto-Regressive LLM.
@ylecun
-
Beyond Auto-Regressive: Future AI Architecture Requirements
By
–
-
Rational Skepticism vs Systematic Negativism in AI Discourse
By
–
There is a big difference between a Shermer-style rational skeptic and a systematic negativist.
-
LLM Size vs Genome: Knowledge Encoded vs Learned
By
–
A small LLM is 14GB.
Your entire genome is 800MB.
Most of it encodes basic biological functions.
A small part encodes brain circuitry.
The difference between your genome and a chimp's genome is 8MB.
Not enough to encode the knowledge you mention.
Much of that must be learned. -
Genome vs LLM: Information Density Comparison
By
–
That's very little information though.
The whole human genome is 3 billion base pairs.
That's 800MB uncompressed.
The difference between chimps and humans is 1% of that, or a mere 8MB.
Even a small 7B LLM is 14GB…. -
Why Hard Tech Breakthroughs Always Hit Walls First
By
–
The phrase "whole career" is particularly accurate. 1. Pick a difficult technology area. Anything: Flying cars, quantum computing, interstellar travel, nuclear fusion, hypersonic flight, AI, whatever.
2. Complain that "current approaches are hitting a wall and will never work." -
Uncommon AI Applications Requiring Physical Intuition
By
–
Almost everything that is uncommon and requires physical intuition.
-
Tacit Knowledge: The Unwritten Foundation of Human Animal Learning
By
–
Everything we learn through apprenticeship or experience is essentially not written down.
That's most of human knowledge and all of animal knowledge. -
Human and Animal Knowledge Beyond Written Language
By
–
I have argued that most of human knowledge and essentially all of animal knowledge isn't written down anywhere, nor even expressible in human language.