Resources

AI 101 Learning Tracks

The guide’s hands-on tracks. Each is self-paced, assumes no programming background, and can be started and stopped without losing your place.

Key Frameworks

These frameworks are referenced throughout the course and serve as practical tools for evaluating and governing clinical AI.

AI Tools for Clinicians

Tools referenced in the course, current as of July 2026. Model names move fast; the categories and the cautions do not. Check the vendor’s own page before relying on any capability claim below.

General-Purpose Assistants

Evidence and Literature Search

Documentation and Administrative Work

Professional Guidelines

Complete Reading List

All required and optional readings from the course, organized by topic. Every entry links to the publisher’s page. Some are paywalled; abstracts are free, and most institutions have access.

Foundations

• Topol, E. “High-performance medicine: the convergence of human and artificial intelligence” Nature Medicine, 2019 — the paper that framed the field for clinicians, and still the best single place to start.
• Rajpurkar, P. et al. “AI in health and medicine” Nature Medicine, 2022 — a survey of where clinical AI actually works, specialty by specialty, and what keeps blocking deployment.
• Obermeyer, Z. et al. “Dissecting racial bias in an algorithm used to manage the health of populations” Science, 2019 — how using cost as a proxy for illness built racial bias into a system covering millions of patients.
• Omiye, J.A. et al. “Large language models propagate race-based medicine” npj Digital Medicine, 2023 — the LLM-era companion to Obermeyer: commercial chatbots reproduced debunked race-based claims when asked directly.
• Wong, A. et al. “External Validation of a Widely Implemented Proprietary Sepsis Prediction Model in Hospitalized Patients” JAMA Internal Medicine, 2021 — a model deployed at hundreds of hospitals performed far worse in practice than the vendor reported.
• Mongan, J. et al. “Checklist for Artificial Intelligence in Medical Imaging (CLAIM)” Radiology: Artificial Intelligence, 2020 — the checklist itself, with the reasoning behind each item.
• Budzyń, K. et al. “Endoscopist deskilling risk after exposure to artificial intelligence in colonoscopy” Lancet Gastroenterology & Hepatology, 2025 — unassisted adenoma detection fell after endoscopists had been working with AI. Observational and before-and-after, so a signal rather than proof, but the first of its kind.
• Ke, Y. et al. “AI-induced never-skilling in medical education” Nature Medicine, 2026 — names the risk that trainees never build reasoning skills they would otherwise have developed. A Perspective; the authors are explicit that direct evidence from medical training does not yet exist.
• Correa Soto, R.A. et al. “Parallel pressures: the common roots of doctor bullshit and large language model hallucinations” BMJ, 2025 — why clinicians and language models both produce confident filler under the same incentives.

Generative AI

• Singhal, K. et al. “Large language models encode clinical knowledge” Nature, 2023 — the Med-PaLM paper; introduced MultiMedQA and showed where model answers still fell short of clinician answers.
• Nori, H. et al. “Capabilities of GPT-4 on Medical Challenge Problems” Microsoft Research (arXiv), 2023 — the source of most “AI passed the boards” coverage. Read it to see exactly what was and was not tested.
• Lee, P. et al. “Benefits, Limits, and Risks of GPT-4 as an AI Chatbot for Medicine” NEJM, 2023 — a readable tour of what these models do well and badly in clinical work.
• Wei, J. et al. “Chain-of-Thought Prompting Elicits Reasoning in Large Language Models” NeurIPS, 2022 — the result behind “show your reasoning,” and why that instruction changes output quality.
• Goh, E. et al. “Large Language Model Influence on Diagnostic Reasoning: A Randomized Clinical Trial” JAMA Network Open, 2024 — giving 50 physicians an LLM did not improve their diagnostic reasoning scores, while the LLM working alone scored 16 points higher than the conventional-resources group. The uncomfortable finding is about how clinicians use the tool.
• Goh, E. et al. “GPT-4 assistance for improvement of physician performance on patient care tasks: a randomized controlled trial” Nature Medicine, 2025 — the follow-up, on management rather than diagnostic reasoning, where assistance did improve physician performance. Read it alongside the 2024 trial.
• Tu, T. et al. “Towards conversational diagnostic artificial intelligence” Nature, 2025 — Google’s AMIE outperformed 20 primary care physicians across most rated axes in text consultations with patient-actors. The authors are candid that synchronous text chat is not how medicine is practiced.
• Bedi, S. et al. “Testing and Evaluation of Health Care Applications of Large Language Models: A Systematic Review” JAMA, 2025 — how these tools are actually being evaluated, and why exam-style benchmarks keep standing in for clinical outcomes.
• Chen, S.F. et al. “LLM-assisted systematic review of large language models in clinical medicine” Nature Medicine, 2026 — 4,609 studies published between 2022 and 2025, of which 19 were prospective randomized trials. The clearest answer available to “what is the evidence base?”

Governance

• Price, W.N. et al. “Potential Liability for Physicians Using Artificial Intelligence” JAMA, 2019 — short, and the clearest explanation of why following an AI recommendation and ignoring one carry different legal exposure.
• Liu, X. et al. “Reporting guidelines for clinical trial reports for interventions involving artificial intelligence: the CONSORT-AI extension” Lancet Digital Health, 2020 — what an AI trial report has to disclose that a conventional one does not.
• Cohen, I.G. “Informed Consent and Medical Artificial Intelligence: What to Tell the Patient?” Georgetown Law Journal, 2020 — the first sustained treatment of whether, and how, AI use has to be disclosed to patients.
• Reddy, S. et al. “A governance model for the application of AI in health care” JAMIA, 2020 — a practical governance structure for a health system rather than a set of principles.
• Collins, G.S. et al. “TRIPOD+AI statement: updated guidance for reporting clinical prediction models that use regression or machine learning methods” BMJ, 2024 — the reporting standard for prediction models. Use it as a reading checklist for any risk score being sold to your institution.
• Lekadir, K. et al. “FUTURE-AI: international consensus guideline for trustworthy and deployable artificial intelligence in healthcare” BMJ, 2025 — consensus guidance covering the whole lifecycle, including what happens after go-live. The natural next read after CLAIM and TRIPOD+AI.

Ongoing Learning

Stay current with these resources:

• Nature Medicine — publishes much of the primary clinical AI literature, including several papers above.
• NEJM AI — the NEJM Group journal dedicated to AI in medicine.
• Ground Truths (Eric Topol) — the author of the 2019 paper above, writing regularly on new clinical AI results.
• The Medical Futurist — accessible analysis of healthcare technology.
• STAT News health tech — reporting on the industry and policy side, including the parts vendors would rather not discuss.
• AI in Healthcare — trade coverage of deployments, vendors, and regulation, with free newsletters.
• This Week Health — podcast network on health IT, formerly This Week in Health IT.

Glossary

AI Triad

The three fundamental components of any AI system: data (what it learns from), algorithms (how it learns), and compute (the processing power that enables learning).

Automation Bias

The tendency to over-rely on automated systems and accept their outputs uncritically, even when contradicted by other information.

Centaur

A human-AI collaboration model with clear division of labor—human handles strategy and judgment, AI handles calculation and data processing.

Chain-of-Thought Prompting

A prompting technique that asks AI to show its reasoning step-by-step before providing a final answer, making its logic visible for evaluation.

Cyborg

A human-AI collaboration model with seamless integration, where the boundary between human and machine cognition is blurred.

External Validation

Testing an AI model on data from a different institution or population than where it was developed.

Few-Shot Prompting

Providing examples of desired input-output pairs before asking the AI to perform a task.

Hallucination

When an AI generates plausible but factually incorrect information.

Label Bias

Bias introduced through the outcome variable being predicted—when the “ground truth” itself encodes discrimination.

Proxy Variable

A measurable variable used as a stand-in for something difficult to measure directly—can encode hidden biases.

Shadow AI

Use of AI tools outside institutional oversight and governance, often consumer products used for clinical work.

Temporal Validation

Testing an AI model on data from a different time period than the training data.

Zero-Shot Prompting

Asking an AI to perform a task without providing examples—relying on its pre-training.