Radiologist tests Fable 5.2 vs experts: about a second-year resident's level
JHochderffer · x · 2026-11-12
A physician (@alexmd2) spent days benchmarking Fable 5.2 against expert radiology, feeding it DICOM files and having it self-grade. Verdict: "good. Very good. But not yet good enough for clinical decisions" — it sometimes under-calls grades or overstates stenosis, landing around a second-year radiology resident or strong internist. JHochderffer adds: radiologists aren't going anywhere; they'll use AI for productivity, costs will fall, and more imaging will get done. A rare hands-on clinical comparison.
More from AGI Musings
- Qualiaphobia: giving AIs justified belief in their own non-consciousness — erikphoel · 2026-11-13
- If the Platonic Representation Hypothesis Holds, Brain–ANN Alignment Research Is a Huge Opportunity — patrickmineault · 2026-11-13
- Chris Paxton: without Jensen Huang, NVIDIA will likely abandon open research and consumers — chris_j_paxton · 2026-11-13
- What should kids learn in the AI era: meta-learning and becoming a better self — yangyi · 2026-11-13
- Bicameral Labs proposes 'Qualiaphobia': AI architectures engineered to reject consciousness theories — erikphoel · 2026-11-12
- "Hard to tell a 92% model from a 99% model": the bottleneck is evals — chris_j_paxton · 2026-11-12