ABOUT THIS EPISODE
Is AI running out of things to learn from? David and Dustin open on the "peak data" problem — the shrinking supply of high-quality, human-generated training data — and whether more compute and synthetic data can keep the gains coming or just poison the well. From there: the device that finally replaces your smartphone (Qualcomm's CEO is betting on always-on glasses within ten years; David is very much not sold), PwC's two-track AI jobs market and the entry-level rung quietly disappearing beneath it, and the first randomized clinical trial of ambient AI scribes in the exam room — modest time savings, but real relief from burnout. Plus a KCL tip that kills the "does Tuesday work?" email chain for good, and parting shots on building your own AI skills and HBO's Chernobyl.
Chapters
- 00:00 — Introduction
- 06:25 — Training Limitations on AI
- 18:13 — What Comes After Smartphones?
- 34:40 — AI and Job Market Dynamics
- 44:59 — Ambient AI in Healthcare
- 54:57 — KCL Tip: Outlook Scheduling Poll
- 57:57 — Parting Shots
Email us at decryption@mcw.edu with questions, thoughts, or topic ideas.
