Question 2 of 20 · Change and evidence
Understanding exponential and discontinuous change
How fast is this moving, and can anyone chart it?
1 min read · Well measured today
AI capability has advanced unusually fast in several measurable areas, but no single growth curve predicts what happens in clinics. Progress can jump, plateau or stall, and adoption runs on a different clock from capability.
What we can measure today
- Established now
A multi-agent system (Microsoft's AI Diagnostic Orchestrator with OpenAI's o3) scored 85.5% on diagnostically challenging New England Journal of Medicine cases, against about 20% for 21 practicing physicians.
What this does not prove: One test design. The physicians worked without their usual tools, and these were published puzzle cases, not real patients.
Stanford HAI, 2026
The 2026 AI Index Report, Chapter 6: Medicine
Scope: Global research, with U.S. data on FDA authorizations.
How it was done: An annual review that summarizes studies by other researchers.
Limits: A summary of studies with different designs. Each headline figure comes from one setup and does not generalize on its own.
Checked against the original on November 6, 2026.
Open the source (opens in a new tab) - Established now
Stanford's overview says strong model performance has not consistently translated into real-world clinical impact, and nearly half of clinical AI studies still rely on simulated scenarios rather than real patient data.
What this does not prove: This gap is real today. It does not prove progress will stay slow.
Stanford HAI, 2026
The 2026 AI Index Report, Chapter 6: Medicine
Scope: Global research, with U.S. data on FDA authorizations.
How it was done: An annual review that summarizes studies by other researchers.
Limits: A summary of studies with different designs. Each headline figure comes from one setup and does not generalize on its own.
Checked against the original on November 6, 2026.
Open the source (opens in a new tab) - Established now
The FDA authorized 258 AI medical devices in 2025. Among devices with clinical studies, only 2.4% were supported by randomized trial data, in an analysis of authorizations through 2024.
What this does not prove: Authorizations can grow far faster than proof of benefit.
Stanford HAI, 2026
The 2026 AI Index Report, Chapter 6: Medicine
Scope: Global research, with U.S. data on FDA authorizations.
How it was done: An annual review that summarizes studies by other researchers.
Limits: A summary of studies with different designs. Each headline figure comes from one setup and does not generalize on its own.
Checked against the original on November 6, 2026.
Open the source (opens in a new tab)
What may change
- Capability, cost and autonomy can compound on each other.
- Adoption can accelerate suddenly once payment or regulation shifts.
What may endure
- Physical, institutional and ethical constraints, which slow things down in ways a benchmark never shows.
The hard questions
- What if change arrives before I graduate?
- What if adoption stalls for a decade after I've rearranged my life for it?
Evidence worth watching
Measured performance · The cost of running models · Reliability · Regulatory milestones · Evidence from real deployment
Sources last checked November 6, 2026
Next review by February 6, 2027.
What changed, and how we decideCareer guidance, not medical, legal or financial advice. For a major loan or licensing decision, talk with a qualified professional, and bring your questions from here.
Do you disagree?
If you have counterevidence, or we got something wrong, we want to see it. Don't take our word for any of this.
Send us the evidence