Two stories broke this week that, on the surface, have nothing to do with each other. One involves a 16th-century Spanish altarpiece. The other involves pharmaceutical molecules. Both reveal the same uncomfortable truth: AI is now better at pattern recognition than the humans who built entire careers around it.
When Machines Settle Art Historical Arguments
Artnet reports that AI has weighed in on a centuries-old dispute about an El Greco painting long thought to be partly the work of his son. The system found compositional fingerprints invisible to human eyes, essentially reassigning authorship. Meanwhile, TechCrunch covers 10x Science, a startup that raised a $4.8 million seed round to filter AI-generated drug candidates. The parallel is exact: in both cases, AI floods the field with outputs, then AI must sort the signal from the noise. TurboFund's seed-stage AI investor list tracks exactly the class of funds backing companies like 10x Science, as the life sciences AI wave accelerates.
The Connoisseurship Problem at Scale
A 2024 paper in Science Advances by Westphal et al. found that neural networks trained on brushstroke micro-texture outperform art historians at attribution tasks by statistically significant margins. The drug discovery analog is equally striking. The bottleneck is no longer generation; it is curation. This is the quiet revolution: AI does not replace experts, it makes expertise itself a filtering problem. The El Greco case is charming. The drug molecule case is urgent. The underlying epistemology is identical. Both fields built entire prestige structures around human pattern recognition, and both are now watching those structures wobble. The question is not whether AI can spot the real Greco. It is who gets to decide what counts as real, and whether that decision will always require a human at the end of the chain.