EXPOSED
Lecture drafting, problem-set generation, worked solutions, and syllabus writing are already commodity AI output, and intro-physics content is the most thoroughly pre-solved material on the internet. What holds is the embodied part: running lab sections with real optics benches and oscilloscopes, diagnosing why a student's mental model of angular momentum is wrong in real time, and supervising undergraduate research. The real pressure on this occupation is structural — enrollment shifts, large-section consolidation, and adjunct staffing — more than a model replacing a professor outright.
Mixed — a routine tier and a judgment tier. At 12 rather than 6, the tasks that resist are the ones that happen with a student in front of you — watching someone set up an interferometer wrong and knowing whether to intervene, reading a confused derivation on a whiteboard mid-office-hour, advising a senior thesis on a problem with no textbook answer — but the bulk-hours work of writing lecture notes on Gauss's law, generating and grading Physics 101 problem sets, and building the syllabus is now near-instant output, which is what keeps this out of the 14+ band.
Some physical or field component. A 10 reflects that the lecture hall and grading queue are screen-substitutable, but lab sections are not: aligning optics on a rail, checking that a student wired a scope probe to ground and not to the 120V side, calibrating force sensors, and enforcing laser and radiation-source safety are hands-on in a room where students break things, and machine-shop or cryogenics access in research supervision pushes it higher than pure classroom teaching would.
No licence, no signature requirement. There is no state physics-teaching licence at the postsecondary level — the credential is a PhD and a department hiring committee, and an institution can and does staff the same PC1010 section with an adjunct, a postdoc, or a graduate TA, so the 3 reflects that nothing statutory prevents substituting who or what delivers the course.
The human relationship is the product. 14 is earned in the specific relationships that carry weight beyond the transcript: the letter of recommendation for graduate school that admissions committees actually read, the research advisor a student names when they publish, and the office-hours history that decides whether a struggling sophomore stays in the major — though the 500-person intro survey course, where most enrolled students never speak to you, keeps it out of the high teens.
Meaningful discretion. At 12, the discretionary calls are real and unappealed — deciding whether a lab report shows fabricated data, setting a curve that determines who fails and loses financial aid, judging whether an undergraduate is ready to run an experiment unsupervised — but the accreditation-driven learning outcomes, departmental common finals, and appeals processes bound those decisions in a way that keeps them below the 14+ band.
Has AI actually changed your work?