Skip to main content

"Bigger Is Always Better"

This video presents the same text shown beside it, spoken and on screen. It adds nothing the text does not say.

State

Bigger models without more data feed the memorization appetite; bigger data of the same skew polishes the skew; bigger both buys capability on the training distribution and nothing, by itself, beyond it.

Show

Scale is a strategy with costs and preconditions — computation, energy, opacity — not a law of nature.

Watch for

Each precondition is checkable with questions this course has already issued.

Builds on

Unlocks

  • Nothing yet depends on this.