Role Definition
| Field | Value |
|---|---|
| Job Title | Physical Scientists, All Other (BLS SOC 19-2099) |
| Seniority Level | Mid-Level (5-10 years, independent research capability) |
| Primary Function | Catch-all for physical scientists not classified elsewhere — includes acousticians, plasma physicists, spectroscopists, rheologists, metrologists, optical scientists, and niche geophysicists. Conducts original research using a mix of computational modelling, laboratory experimentation, and field measurement. Designs experiments, builds and validates theoretical models, interprets complex datasets, and publishes findings. Works in national laboratories, defence contractors, universities, semiconductor firms, energy companies, and government agencies. |
| What This Role Is NOT | NOT a Materials Scientist (SOC 19-2032, scored 33.0 — assessed separately). NOT a Physicist proper (SOC 19-2012, separate BLS code). NOT an Astronomer (SOC 19-2011, scored 45.2). NOT an Atmospheric/Space Scientist (SOC 19-2021, scored 30.6). NOT a Chemist (SOC 19-2031, scored 38.4). This is the residual category for physical scientists who fall outside those specific codes. |
| Typical Experience | 5-10 years. PhD in physics, applied physics, acoustics, plasma science, optics, or related field. Independent researcher with publication record and equipment proficiency. Proficient in domain-specific simulation tools (COMSOL, ANSYS, EPOCH, OSIRIS, custom codes) and increasingly in Python/ML frameworks. |
Seniority note: Entry-level researchers (postdocs executing defined experiments, 0-3 years) would score deeper Yellow or borderline Red due to higher proportion of routine computation and data processing. Senior principal investigators with strategic research direction and grant leadership would score Green (Transforming) ~50-55.
Protective Principles + AI Growth Correlation
| Principle | Score (0-3) | Rationale |
|---|---|---|
| Embodied Physicality | 1 | Many sub-specialisms require lab or field presence — operating plasma chambers, configuring acoustic measurement rigs, aligning optical systems, conducting field measurements. But a significant fraction of daily work (modelling, data analysis, writing) is desk-based. Structured lab/field environments, not unstructured trades work. |
| Deep Interpersonal Connection | 1 | Collaborates across interdisciplinary teams, mentors junior researchers, presents at conferences, coordinates with engineers and technicians. Professional relationships matter but are not the core deliverable. |
| Goal-Setting & Moral Judgment | 2 | Generates novel hypotheses, decides which experiments are worth pursuing, interprets ambiguous or contradictory results, defines research directions within a programme. Mid-level scientists exercise meaningful scientific judgment but typically work within research agendas set by senior PIs or programme managers. |
| Protective Total | 4/9 | |
| AI Growth Correlation | 0 | Demand for these niche physical scientists is driven by fusion energy, semiconductor manufacturing, defence/sensing, medical imaging, and environmental monitoring — not AI adoption directly. AI makes these scientists more productive (neural surrogates replace slow simulations, ML classifies sensor data) but it is unclear whether this reduces headcount or expands research scope. Net neutral. |
Quick screen result: Protective 4/9 with neutral growth — Likely Yellow Zone. Proceed to quantify.
Task Decomposition (Agentic AI Scoring)
| Task | Time % | Score (1-5) | Weighted | Aug/Disp | Rationale |
|---|---|---|---|---|---|
| Experimental design & hypothesis generation | 25% | 2 | 0.50 | AUGMENTATION | Designing experiments to test novel physical phenomena — configuring plasma parameters, designing acoustic measurement protocols, selecting optical configurations. Requires deep domain expertise and creative problem-solving. AI suggests experimental parameters but the scientist decides what questions are worth asking and how to test them. |
| Computational modelling & simulation | 20% | 4 | 0.80 | DISPLACEMENT | Running simulations — plasma MHD codes, acoustic wave propagation, optical ray tracing, Monte Carlo methods. Neural surrogate models increasingly replace full-fidelity simulations for parameter sweeps. DeepMind's deep RL controls tokamak plasmas autonomously. Standard simulation workflows are highly automatable; the scientist reviews outputs but AI executes end-to-end. |
| Data analysis & scientific interpretation | 15% | 3 | 0.45 | AUGMENTATION | Analysing experimental data — spectroscopic signals, acoustic waveforms, sensor arrays, imaging data. ML excels at pattern recognition and classification. But interpreting anomalies, reconciling theory with observation, and generating scientific insight from unexpected results remains human-led. AI accelerates; scientist validates and contextualises. |
| Laboratory experimentation & measurement | 15% | 2 | 0.30 | AUGMENTATION | Operating plasma chambers, acoustic test facilities, optical benches, cryogenic systems, particle detectors. Physical handling of equipment in semi-structured environments. Self-driving labs exist for some domains but most niche physical science labs lack standardised automation. Scientist uses AI diagnostics but performs the physical work. |
| Technical documentation & publication | 10% | 4 | 0.40 | DISPLACEMENT | Research papers, grant proposals, technical reports, lab notebooks. AI agents draft literature reviews, generate methods sections, create figures from raw data. Human refines arguments and owns scientific claims, but AI handles 50-70% of initial drafting. |
| Literature review & knowledge synthesis | 5% | 4 | 0.20 | DISPLACEMENT | Surveying published research, identifying gaps, synthesising findings. AI tools (Semantic Scholar, Elicit, Consensus) already perform systematic literature synthesis at scale. Scientist directs the search and evaluates relevance, but AI executes the retrieval and summarisation. |
| Cross-functional collaboration & mentoring | 10% | 1 | 0.10 | NOT INVOLVED | Interdisciplinary meetings, conference presentations, mentoring students, coordinating with engineers and technicians. Human relationships and scientific communication. |
| Total | 100% | 2.75 |
Task Resistance Score: 6.00 - 2.75 = 3.25/5.0
Displacement/Augmentation split: 35% displacement, 55% augmentation, 10% not involved.
Reinstatement check (Acemoglu): Meaningful reinstatement. AI creates new tasks: validating AI-generated surrogate models against physical experiments, curating domain-specific training datasets, interpreting why ML predictions fail for edge-case physics, designing experiments that stress-test AI predictions, and bridging computational outputs with real-world measurement constraints. Scientists who master AI-augmented workflows become dramatically more productive — the role transforms toward AI-directed experimentation rather than manual computation.
Evidence Score
| Dimension | Score (-2 to 2) | Evidence |
|---|---|---|
| Job Posting Trends | 0 | BLS projects 0.6% total growth 2024-2034 for SOC 19-2099 (31,900 to 32,100) — essentially flat. This is a tiny, niche occupation. Demand varies heavily by sub-specialism: plasma/fusion physicists see growing demand as ITER and private fusion companies (Commonwealth Fusion, TAE Technologies, Helion) scale; acousticians see steady demand in medical devices and defence; optical scientists see semiconductor-driven demand. Aggregate is stable. |
| Company Actions | 0 | No companies cutting niche physical scientists citing AI. National labs (LLNL, ORNL, Sandia, DIII-D) continue hiring. Private fusion companies are expanding. AI is deployed as a productivity tool, not a headcount replacement. No restructuring signals. |
| Wage Trends | 0 | BLS median $123,070 (May 2024) for SOC 19-2099. Mean $132,110. Growing modestly, tracking inflation. ML-proficient physical scientists command premiums, particularly in fusion/plasma and semiconductor domains. No wage compression signals. |
| AI Tool Maturity | -1 | Strong AI tools in pilot/early adoption for core tasks: DeepMind deep RL for tokamak plasma control (Nature 2024, 183 citations), neural surrogate models replacing full simulations, ML classifiers for sensor data, AI-assisted experimental design. COMSOL and ANSYS integrating ML modules. Tools perform 30-50% of computational workflows with human oversight. Not yet at Materials Science levels of autonomous discovery (no GNoME equivalent for general physical science), but advancing steadily. |
| Expert Consensus | 0 | Mixed. OpenAI's "AI as Scientific Collaborator" white paper (Jan 2026) emphasises collaboration over replacement. Sabine Hossenfelder (Jan 2026): "AI is coming for scientists' jobs — seriously" — but argues transformation, not elimination. Wharton (Sept 2025): ~42% of jobs at physicist income level have work "susceptible to automation." No consensus on displacement vs augmentation for this specific residual category. The diversity of sub-specialisms makes generalisation difficult. |
| Total | -1 |
Barrier Assessment
Reframed question: What prevents AI execution even when programmatically possible?
| Barrier | Score (0-2) | Rationale |
|---|---|---|
| Regulatory/Licensing | 1 | No formal licensure for most physical scientists. But grant-funded research mandates qualified PIs (PhD + track record). National lab work may require security clearances (particularly plasma/fusion and defence-related acoustics/optics). Peer review enforces human accountability for published claims. Moderate institutional friction. |
| Physical Presence | 1 | Lab-based experiments — operating plasma devices, acoustic chambers, optical benches, cryogenic systems — require physical presence. Field measurements (environmental acoustics, geophysics) require on-site work. But computational and analytical work (growing proportion) is desk-based. Self-driving labs remain rare in these niche domains. |
| Union/Collective Bargaining | 0 | Physical scientists in academia and national labs are generally not unionised. Some government lab researchers have limited union representation, but collective bargaining is not a meaningful barrier. |
| Liability/Accountability | 1 | Published scientific claims carry reputational consequences. Research integrity violations lead to retractions and career damage. Grant agencies require human PIs accountable for outcomes. Safety-critical applications (fusion reactors, medical acoustics) involve oversight requirements. But liability is professional/reputational, not criminal. |
| Cultural/Ethical | 0 | Physical science communities actively embrace computational and AI tools. ML is celebrated as accelerating discovery. No cultural resistance — the field has been computationally intensive for decades. |
| Total | 3/10 |
AI Growth Correlation Check
Confirmed at 0 (Neutral). Fusion energy, semiconductor manufacturing, defence sensing, medical imaging, and environmental monitoring drive demand for niche physical scientists — not AI adoption. AI makes individual scientists more productive but the net headcount effect is ambiguous. Fusion is the strongest growth driver: ITER, Commonwealth Fusion Systems, TAE Technologies, and Helion are all hiring plasma physicists and related specialists. But this is fusion-demand-driven, not AI-demand-driven. The role does not exist because of AI, nor does AI adoption directly reduce or increase demand.
JobZone Composite Score (AIJRI)
| Input | Value |
|---|---|
| Task Resistance Score | 3.25/5.0 |
| Evidence Modifier | 1.0 + (-1 x 0.04) = 0.96 |
| Barrier Modifier | 1.0 + (3 x 0.02) = 1.06 |
| Growth Modifier | 1.0 + (0 x 0.05) = 1.00 |
Raw: 3.25 x 0.96 x 1.06 x 1.00 = 3.3072
JobZone Score: (3.3072 - 0.54) / 7.93 x 100 = 34.9/100
Zone: YELLOW (Green >=48, Yellow 25-47, Red <25)
Sub-Label Determination
| Metric | Value |
|---|---|
| % of task time scoring 3+ | 50% |
| AI Growth Correlation | 0 |
| Sub-label | Yellow (Urgent) — 50% >= 40% threshold |
Assessor override: None — formula score accepted. At 34.9, this sits 1.9 points above Materials Scientist (33.0) and 3.5 points below Chemist (38.4). The slightly higher task resistance (3.25 vs 3.10) reflects the broader diversity of sub-specialisms within this catch-all — many involve physical experimentation and measurement in niche domains where AI tool maturity is lower than in materials discovery (no GNoME equivalent for acoustics, plasma physics, or optical science). Evidence is slightly less negative (-1 vs -1) because AI tools target core tasks less directly than in materials science. The positioning between Materials Scientist and Chemist is honest for a residual category that averages across more-exposed computational subspecialties and less-exposed experimental ones.
Assessor Commentary
Score vs Reality Check
The Yellow (Urgent) classification at 34.9 is honest but masks enormous internal variation. This is a catch-all BLS code covering subspecialties with wildly different AI exposure profiles: a computational plasma physicist running MHD simulations faces near-term displacement of core computational tasks, while an experimental acoustician designing novel measurement protocols in anechoic chambers faces minimal AI threat. The 3.25 task resistance is a defensible average, but no individual sub-specialism actually scores exactly 3.25. The score is not borderline — at 13.1 points from the Green threshold and 9.9 from the Red threshold, it sits firmly in mid-Yellow.
What the Numbers Don't Capture
- Extreme bimodal distribution — More than any other role assessed, this catch-all category spans subspecialties with genuinely different automation profiles. Computational physical scientists (plasma simulation, optical modelling, acoustic wave propagation) face task resistance closer to 2.5-2.8. Experimental physical scientists (lab-based measurement, field work, instrument development) sit closer to 3.5-3.8. The 3.25 average is technically correct but not descriptively useful for any individual.
- Fusion-sector demand surge — Private fusion investment exceeded $6B cumulative by 2025. Commonwealth Fusion, TAE Technologies, Helion, and Zap Energy are hiring plasma physicists aggressively. This sub-population within SOC 19-2099 likely scores Yellow (Moderate) or even borderline Green due to acute talent shortage. The aggregate BLS projection (0.6% growth) does not capture sector-specific surges.
- National lab vs industry divergence — National lab scientists operate under institutional structures (clearances, peer review, safety oversight, government employment) that function as de facto barriers beyond the 3/10 score. Industry researchers at startups face fewer protections. The barrier score is an average across both contexts.
- Self-driving lab trajectory — Autonomous experimental platforms are advancing fastest in chemistry and materials science. Most niche physical science domains (plasma, acoustics, optics) lack equivalent autonomous lab capabilities. This gives experimental physical scientists more runway than materials scientists — but the gap may narrow as general-purpose lab automation matures.
Who Should Worry (and Who Shouldn't)
Experimental physical scientists who spend most of their time in the lab or field should not worry as much as the label suggests. If you operate plasma chambers, configure acoustic measurement rigs, build optical systems, or conduct field measurements — your value comes from physical-world expertise and instrument mastery that autonomous systems cannot yet replicate in niche domains. Most protected: specialists working with novel or custom instrumentation, extreme environments (cryogenic, high-vacuum, high-power), or one-of-a-kind experimental facilities. More exposed: computational physical scientists whose daily work is primarily simulation, parameter sweeps, and numerical modelling — these workflows are directly targeted by neural surrogate models, ML-accelerated solvers, and AI-driven experimental design tools. The single biggest factor: whether you work primarily with physical instruments and materials (protected) or primarily with computational models and code (exposed). Plasma physicists working on experimental tokamaks at national labs are meaningfully safer than computational plasma physicists running EPOCH simulations from their desks.
What This Means
The role in 2028: Mid-level physical scientists spend less time on routine simulation runs and parameter sweeps as neural surrogate models and ML-accelerated codes handle computational workflows. More time shifts toward experimental validation, instrument development, interpreting anomalies between AI predictions and physical reality, and designing experiments that probe regimes where AI models break down. The scientist who masters AI tools becomes 5-10x more productive — running hundreds of computational experiments in the time it once took to run ten. But the productivity multiplier compresses computational teams: groups that needed 4 simulation researchers may need 1-2 plus AI tooling.
Survival strategy:
- Deepen hands-on experimental and instrumentation expertise. Physical-world skills — operating plasma devices, configuring acoustic measurement rigs, aligning optical systems, building custom instruments — are the AI-resistant core. Seek time on unique, specialised facilities.
- Master AI/ML tools for your specific domain. Neural surrogate models, physics-informed neural networks (PINNs), deep RL for experimental control, ML-assisted data analysis. Python and ML fluency are becoming non-negotiable even for experimentalists. Learn to critically evaluate AI-generated results against physical intuition.
- Specialise in regimes where AI predictions fail. Extreme conditions, multi-physics coupling, novel phenomena without training data, turbulent and chaotic systems — these are the domains where computational tools are weakest and human scientific judgment is most valuable.
Where to look next. If you're considering a career shift, these Green Zone roles share transferable skills with physical science research:
- Medical Scientist (Mid-Level) (AIJRI 54.5) — Your experimental design, data analysis, and hypothesis testing skills transfer directly to biomedical research, with stronger structural barriers.
- Computer and Information Research Scientist (Mid-to-Senior) (AIJRI 57.5) — Your computational modelling and ML experience positions you for applied AI/ML research roles, where demand is surging.
- Natural Sciences Manager (Mid-to-Senior) (AIJRI 51.6) — Your technical expertise plus any team leadership or grant management experience positions you for R&D management where strategic judgment is the core value.
Browse all scored roles at jobzonerisk.com to find the right fit for your skills and interests.
Timeline: 3-5 years for significant transformation of computational and simulation-heavy subspecialties. Experimental and instrumentation-heavy subspecialties persist longer but face gradual automation pressure as domain-specific self-driving labs mature. The diversity of this catch-all category means the timeline varies enormously by sub-specialism.
