Computational irreducibility is the enemy of neural net training because neural networks rely on predictive shortcuts—if a system is computationally irreducible, you cannot burn in a prediction of its outcome without explicitly computing it, making training far harder or impossible.

causalpending

Speaker

Stephen Wolfram

Evidence Quote

“computational irreducibility is the enemy of neural net training because as soon as there is computational irreducibility... you just can't do this approach... of explicitly burn in a prediction”

Source

CHATGPT won't ever be the same again after this— Machine Learning Street Talk
Created: 8/11/2026, 1:49:43 AM

Your notes

Loading notes...