A Brief History of AI in Healthcare W/ Lekan Wang, Partner, JSL Health Capital

On CareTalk, Lekan Wang traces his journey from a genomics‑obsessed student to a Palantir engineer who first tackled healthcare data problems. At Palantir, he helped the CDC stitch together fragmented state‑level food‑borne outbreak reports, proving that integrating disparate datasets could yield actionable public‑health insights. He describes how early work relied on classic statistical models and large databases, but the emergence of machine‑learning loops gradually turned those tools into what we now call artificial intelligence. This historical lens shows why data integration remains the foundation for every AI breakthrough in medicine.

The current AI surge is powered by three forces: unprecedented data volumes, exploding compute power, and the transformer model introduced in the 2017 “Attention Is All You Need” paper. GPUs, TPUs and custom ASICs now train models on internet‑scale and synthetic datasets, while transformers enable scalable attention across billions of parameters. In healthcare, these advances promise to cut the massive documentation load that clinicians face, streamline patient triage, and close the access gap that leaves many Americans without timely care. Early clinical studies already show AI reducing avoidable medical errors and accelerating diagnostic pathways.

JSL Health Capital embodies this optimism by backing AI‑first companies that own or generate proprietary clinical data. Their portfolio includes Symphony, which uses generative chemistry models and robotic flow‑chemistry to design and synthesize drug candidates on demand, and General Medicine, which leverages AI to make care navigation transparent and to automate scheduling and triage. Wang stresses that the real value lies in models trained on rich, domain‑specific data rather than generic internet text, addressing the “stochastic parrot” criticism. While questions about true generalization remain, investors see a target‑rich environment—high administrative overhead, diagnostic delays, and preventable errors—where AI can deliver measurable improvements.