Data may become the bottleneck limiting AI improvement before compute becomes limiting, but multiple alternatives exist: synthetic data generation (especially for coding tasks, where AI can generate training examples that train better models recursively), identifying efficiency (finding that ~1% of parameters in LLMs are truly useful), and metacognition (planning approaches that improve performance orders of magnitude without requiring large new datasets).
factualpending
Speaker
Erik BrynjolfssonEvidence Quote
“data is probably going to be the bottleneck... either ways of creating synthetic data other ways around it... imagine if we could find a more efficient way to find that that 1%... if you ask an LM what are the five steps you should do in writing a paper... it does a lot better than if it just starts with predicting each word”
Source
The AI Awakening: Implications for Business and the Economy – Business & Generative AI— Wharton Human-AI ResearchCreated: 8/12/2026, 6:19:17 PM
Your notes
Loading notes...