fine tune
Further training of a pretrained model on a narrower task or dataset to specialize its behavior.
As defined in Dario Amodei (Anthropic CEO) — The hidden pattern behind every AI breakthrough
Where it comes up · 2
- Dario Amodei (Anthropic CEO) — The hidden pattern behind every AI breakthrough3 claims
- Mark Zuckerberg #31 claim
Most important claims · 4 of 4Search all claims for “fine tune” →
- When a model is trained to be aligned, we don't actually know what happens inside it; current fine-tuning methods have the property that the underlying knowledge and dangerous abilities don't disappear — the model is merely taught not to output them — and it's unknown whether this is a fatal flaw or simply how things must be.Dario Amodei (Anthropic CEO) — The hidden pattern behind every AI breakthrough
- Next-word prediction provides rich structure because correctly predicting text requires solving theory-of-mind problems, math problems, and developmental-test-style reasoning embedded in the data, so scaling the objective is 'halfway to everywhere' — a single language model can then be fine-tuned with small effort to solve many disparate tasks.Dario Amodei (Anthropic CEO) — The hidden pattern behind every AI breakthrough
- Llama 2 by itself is not a consumer product but a piece of infrastructure people build with; Meta's priority is fine-tuning it and its branches into consumer products—the AIs and personas—aiming for hundreds of millions and eventually billions of users.Mark Zuckerberg #3
- There is large overlap between human skills and what models learn from internet text, but models miss some things humans have (like moving around in the physical world) while learning things humans don't (like speaking fluent Base 64), implying the skill repertoires are overlapping but not identical.Dario Amodei (Anthropic CEO) — The hidden pattern behind every AI breakthrough