Hardt's Historical Critique of AI Benchmarking

Computer scientist Moritz Hardt critiques how historical AI benchmarking mechanisms have shaped research cycles without current abundance implications

Computer scientist Moritz Hardt, director at the Max Planck Institute for Intelligent Systems, published 'The Emerging Science of Machine Learning Benchmarks' in October 2026. His analysis traces how past benchmarking practices—like DARPA’s 1987 speech-to-text test—triggered specific AI research winters, while the 2012 ImageNet benchmark accelerated the deep learning revolution. Hardt identifies these historical mechanisms as the origin of flawed AI development pathways, arguing benchmarks have driven progress despite long-standing inaccuracies. The source explicitly states this critique has no direct consequence for abundance, as it focuses on historical patterns without current practical fixes or cost reductions. This historical perspective does not indicate present friction for knowledge goods or affordability. The source describes this as a theoretical analysis with no direct consequence for current abundance.

Source: Quanta