causal

Training Data Quality Beats Training Steps

In AI fine-tuning of faces, how good your training data is matters more in many ways than how many steps you train for; the photos should be diverse in everything (lighting, clothing, location) except the face, because whatever stays constant across photos becomes the trained concept.

causalpending

Speaker

Pieter Levels

Evidence Quote

“the things that are the same every photo are the concepts that are trained. What you want is your face to be the concept that's trained and everything else to be diverse different.”

Source

Pieter Levels— Lex Fridman Podcast
Created: 6/13/2026, 3:32:34 AM

Your notes

Loading notes...