Scientists Trained WHO to Diagnose Breast Cancer? With Hannah Fry #shorts #science #hannahfry
Why It Matters
The study proves that collective, non‑human intelligence can rival expert pathologists, signaling a paradigm shift toward AI‑driven, crowd‑sourced cancer diagnostics.
Key Takeaways
- •Pigeons trained for two weeks achieved 85% accuracy diagnosing breast cancer.
- •Collective decisions of the flock raised accuracy to 99%, matching pathologists.
- •Study demonstrates non‑human agents can perform complex visual classification tasks.
- •Highlights potential of crowd‑sourcing and AI to augment medical diagnostics.
- •Challenges assumption that only humans can master specialized diagnostic skills.
Summary
The video recounts a 2015 study in which researchers taught sixteen naïve pigeons to diagnose breast cancer from pathology slides. Using a simple interface—pecking one side for malignant, the other for benign—the birds received treats for correct answers.
After just two weeks of training, the individual pigeons achieved roughly 85% accuracy, a surprisingly high figure for non‑human subjects. More strikingly, when the researchers aggregated the birds’ votes, the collective “flock‑sourcing” approach pushed accuracy to 99%, on par with seasoned pathologists.
The presenter highlights a quirky detail: one bird performed randomly, but removing it improved the group’s performance. The experiment underscores that complex visual pattern recognition can be outsourced to unconventional agents, foreshadowing similar gains from artificial intelligence and crowd‑sourced diagnostics.
The broader implication is a challenge to the belief that only humans possess the cognitive capacity for specialized medical tasks. If pigeons can be trained to spot cancer, sophisticated algorithms and collaborative platforms may soon match or exceed human expertise, reshaping diagnostic workflows and resource allocation in healthcare.
Comments
Want to join the conversation?
Loading comments...