Rely more on real data and less on hand-crafted feature extraction
Yann LeCun et al., November 1998 (Proceedings of the IEEE). Yann LeCun and colleagues at AT&T Labs explaining why learning from data was overtaking hand-built pattern recognition, in their November 1998 paper.
Gradient-Based Learning Applied to Document Recognition
More from Yann LeCun et al. · AI and science · Before ChatGPT
Who else?
What it refers to: Large databases for widely used tasks such as handwriting recognition, which let designers build recognition systems without relying on hand-designed feature extractors.
Other voices
- “Data is the moat for ML systems right now” Nathan Lambert, Writer,
- “Creativity depends on theories that challenge data, not algorithms” Teppo Felin, Researcher,
- “Solely measuring skill at any given task falls short of measuring intelligence” François Chollet,
- “These models somehow just generalize dramatically worse than people” Ilya Sutskever, Builder,
- “Learning meaning requires more than form” Bender et al.,










