Thomas Burger Warns 10 GenAI Uses Could Distort Biological Evidence
Updated
Updated · ScienceAlert · Aug 8
Thomas Burger Warns 10 GenAI Uses Could Distort Biological Evidence
2 articles · Updated · ScienceAlert · Aug 8
Summary
An opinion article in Patterns maps hallucination risk across 10 biological uses of generative AI, arguing the biggest danger comes when synthetic outputs are treated as evidence rather than hypotheses.
Lower-risk uses include screening drugs or proteins, where AI can narrow candidates but lab experiments still decide whether a finding is real.
Risk rises when models fill gaps in omics datasets or generate synthetic biological data, because subtle distortions can amplify, erase or invent signals that researchers may struggle to detect.
AlphaFold 3 offered a real example in 2024: its developers reported hallucinated structures in disordered protein regions, though low-confidence scores can flag some errors.
Burger argues the key safeguard is independent experimental validation, since AI may occasionally inspire real discoveries but does not turn plausible output into biological fact.