2020 Air Street Capital The State of AI Report 2020 · page 16 of 177
The charts demonstrate power-law relationships between model performance (test loss) and three key scaling factors: compute, dataset size, and parameter count.
Consulting deck · industry_trends · analyze_data · dense density
Slide locked Sign in to view
Slide schematic 5/7
callout 0/1 chart 3/3 metric 0/1 paragraph 1/1 title 1/1
2 without position callout 1 · metric 1
calloutmetric · primary
Components 7
callout line line line primary paragraph headline
Tools 3
Action Titles slide · 90%
Title states the empirical scaling law conclusion.
Causal Chain slide · 70%The description of how bigger models, datasets, and compute budgets drive performance implies a causal relationship.
Chart Selection Guide slide · 70%Power-law scaling chart appropriate for log relationship.
Metrics 0
∅
No metrics
No metric insight is anchored to this slide.
Frameworks 2