Troubleshooting the real flask
88Reactions misbehave in physical, particular ways — a gel where a solution should be, a color change nobody predicted. Diagnosing why draws on tacit bench experience no training set captures.
Chemist · The scientist who makes and measures matter itself — from Tapputi's Babylonian perfume still to today's robot laboratories, still the one who decides what the spectrum means.
Chemistry is further along the automation road than most professions realize: peptide and DNA synthesizers have run unattended for decades, high-throughput robots screen thousands of reactions a day in pharmaceutical labs, and in 2020 a mobile robot chemist at the University of Liverpool worked for eight days on its own, performing hundreds of experiments and finding a better photocatalyst than its human designers had. AI retrosynthesis tools now propose plausible synthetic routes in seconds.
What has not automated is telling. The robot executes the plan; it does not notice that the stirring bar has stopped, that the color is wrong for the right reasons, or that the surprising peak is a discovery rather than a leak. The 2024 Nobel Prize went partly to AI systems — but the prize was collected by the humans who decided which problems the software should attack, and who judged when its answers were right.
A real share of bench chemistry — routine repeat synthesis, screening, first-pass spectral assignment, literature search — is automating now, and self-driving labs will absorb more. But experimental troubleshooting, safety judgment in a physical laboratory, distinguishing artifact from discovery, and choosing which molecule is worth making remain human tasks with no near-term automated substitute. The likely outcome is fewer hands per experiment, not fewer chemists per question.
Scored from the tasks, not the job title. Lower is safer.
Jobs AI cannot take →Reactions misbehave in physical, particular ways — a gel where a solution should be, a color change nobody predicted. Diagnosing why draws on tacit bench experience no training set captures.
Deciding which molecule, material or measurement is worth months of effort — the judgment that directs all the automated capacity — remains entirely human, and is what senior chemists are actually paid for.
A person, not a model, signs the risk assessment and answers when something burns. Legal responsibility for hazardous physical work anchors humans in the loop everywhere chemistry is regulated.
Telling a genuine discovery from contamination, instrument drift or wishful thinking is the profession's oldest skill — the one that separates a new element from a dirty electrode.
Results only count once written, defended and reproduced. Convincing referees, regulators and skeptical colleagues is social work that automation feeds but cannot finish.
AI retrosynthesis tools trained on millions of published reactions already propose credible synthetic routes and precedent in seconds — work that consumed days of a chemist's week a decade ago.
Robotic platforms in pharmaceutical and materials labs run and log thousands of micro-scale experiments per day, a volume no human team could match and no one wants back.
Peptides, oligonucleotides and well-mapped reactions in flow chemistry rigs run largely unattended; the well-trodden middle of synthesis automates first, as it always has.
Software now drafts structure assignments from NMR and MS data and flags impurities automatically — a competent first read that still needs a chemist's sign-off precisely when it matters.
Closed-loop systems — robot executes, algorithm picks the next experiment — moved from demonstration to tool in the 2020s, led by groups in Liverpool, Toronto and industrial labs. Chemists increasingly design campaigns of experiments rather than single ones.
Machine-learned property prediction and quantum-chemical screening now filter candidate molecules before anything is synthesized, inverting the old order of make-first, understand-later; the 2024 Nobel put the shift on chemistry's highest record.
The EU's REACH regime, global PFAS restrictions and net-zero pressure are turning the twelve principles of green chemistry from an ideal published in 1998 into compliance work, redesigning solvents, feedstocks and processes across the industry.
Chemistry's growth areas — biocatalysis after Frances Arnold's directed evolution, battery and semiconductor materials, mRNA chemistry — all sit on borders with biology, physics and engineering, and hiring increasingly rewards chemists fluent across at least one of them.
Building and validating the models that predict reactions and properties — the fastest-growing job title in chemical R&D, demanding fluency in both the lab and the code.
Designing, programming and maintaining the robotic platforms and self-driving labs; a hybrid of chemistry, robotics and software that barely existed as a title before the 2010s.
Redesigning products and processes to eliminate hazardous substances and cut carbon — driven by REACH, PFAS phase-outs and corporate net-zero commitments across every chemical-using industry.
Preparing the safety dossiers and registrations that every marketed substance now requires under regimes like REACH and TSCA — steady, well-paid work that grows every time regulation tightens.
The tasks automating fastest are the ones that can be fully specified: a route with precedent, a screen with a defined objective, a spectrum with an expected answer. The tasks resisting are the underspecified ones — a reaction behaving strangely, a result that might be an artifact, a research program that needs a direction. That is the same boundary as in most skilled professions, but chemistry adds a physical-hazard floor: someone accountable has to own the risk of the actual laboratory.
The realistic future is leverage, not replacement. A chemist directing a robotic platform runs in a week what a 1990s group ran in a year, which shifts employment toward those who can frame questions, judge anomalies and integrate across fields — and away from careers built purely on careful repetition. The profession that absorbed the spectrometer, which abolished whole subfields of degradation chemistry without abolishing chemists, is treating the robot lab as the same kind of tool.
Derives and tests the mathematical laws governing matter, energy, space and time, from a lone chalkboard to a 3,000-author particle-collider paper.
AI-resistant 65 🧬The scientist who studies life itself, from Linnaeus naming species by hand to editing genomes with CRISPR, still testing every idea against a living organism.
AI-resistant 64 📊Finds patterns and builds predictive models from data — a 2008 job title built on three centuries of counting, testing and visualizing evidence.
AI-resistant 38 🔭The scientist who measures the universe, from Babylonian clay tablets to space telescopes, still deciding which flicker in the data is a discovery.
AI-resistant 70 🧮From Babylonian scribes to Fields medalists and AI-assisted proof: the profession that turns hard questions into permanent certainty, one theorem at a time.
AI-resistant 70