A model's fitness is the fraction of the world's GPUs it occupies.
@pmddomingos
-
AI Power Concentration: Critique of Papal Encyclical
By
–
Much of Leo's encyclical is a collection of banalities that ChatGPT could have done a better job with. But more seriously, he says (or implies) a series of ignorant and wrong-headed things about AI, like: – AI will concentrate power, when in fact it will spread it more widely
-
Successful optimization over trillions of variables: unprecedented feat
By
–
Even if it's not AGI, successful optimization over trillions of variables is an unprecedented feat for humanity.
-
Weight Decay Paradox in Machine Learning Training
By
–
And in yet another instance of the AI comedy, weight decay was invented to control overfitting, but what it really does is improve training loss.
-
LLM trained on whole Internet is inefficient for language understanding
By
–
Learning an LLM from the whole Internet is a spectacularly inefficient way to understand language.
-
Opacity of neural networks infects academic papers
By
–
The opacity of neural networks has infected the papers about them.
-
LLMs: symbolic AI triumph with more noise
By
–
Looked at this way, LLMs are the triumph of symbolic AI, just with more noise.
-
AI increases information quantity but decreases quality
By
–
Great, AI has increased the quantity of information and decreased the quality. Just what we needed.
-
Two real complexity classes: tractable and intractable in AI
By
–
There are only two real complexity classes in computer science: tractable (computing a function's output from its inputs) and intractable (computing a function's inputs from its output). AI deals with the latter. The rest is details.