AI4S Commentary

ArgumentStandard lab interfaces make instrument control easier, but API access does not give an agent a model of physical state or safe operation.

Why readRead it for the Genentech foam failure: the software call was correct while the experiment was physically wrong. It makes the next bottleneck unusually concrete.

ArgumentScience optimized for machine-readable output may accelerate discovery while making its explanations and control less accessible to people.

Why readUseful because it treats interpretability, corporate control and biological black-box data as one institutional problem rather than three technical side issues.

ArgumentA useful AI scientist needs non-obvious analogy, productive disagreement and a model of environmental dynamics, not only fluent reasoning in language.

Why readZou shifts the question from whether agents can execute a workflow to whether they can generate the kind of disagreement from which new science begins.

ArgumentAI scientists should be evaluated in hidden, interactive worlds where evidence can change hypotheses, not with static question-answer benchmarks.

Why readOne of the clearest proposals for what to build next: limited budgets, real feedback and traces that reveal whether an agent actually updates.

ArgumentAI may compress target and molecule discovery, but toxicity, pharmacokinetics, trial design, recruitment and regulation still return evidence on year-long clocks.

Why readIt turns the promise of faster biomedicine into a sequence of named stages. That makes clear which clock a model can shorten and which clocks it cannot yet touch.

ArgumentPlausibility is not enough in experimental science; automation needs inspectable methods, constraints and provenance because mistakes enter the physical world.

Why readRickerby actually let an AI choose work down to reagent and well. That makes this a report from the boundary, not a forecast about it.

ArgumentStrong benchmark scores do not yet transfer to open-ended research where problem choice, judgment and changing direction are part of the task.

Why readTheir method matters more than the headline: inspect how the number was produced, how much scaffolding contributed and what the agent chose without help.

ArgumentClaims that AI is transforming productivity remain dominated by anecdotes, while rigorous evidence is sparse and unusually hard to interpret.

Why readAfter the withdrawn MIT preprint, this becomes an editorial standard for every spectacular productivity number: ask what was actually measured.

ArgumentIf agents make hypotheses cheap, validation, agent-ready data and peer review become the new scientific bottlenecks.

Why readA rare piece that asks what funders and institutions should build once idea generation is no longer scarce.

ArgumentPublic commitments to responsible AI can become ineffective when internal incentives make dissent costly and external accountability weak.

Why readIts value is specificity, not neutrality. Turner separates public evidence from private exchanges that readers cannot independently verify.

ArgumentModels can solve well-specified problems without possessing the tacit knowledge, experience of failure and judgment needed to identify which problems are worth specifying.

Why readA structural challenge to today’s science benchmarks: they score answers to questions that have already survived the hardest act of scientific judgment.

ArgumentRLVR works in mathematics and code because verification is cheap and reliable; most important scientific questions lack that condition.

Why readThe most direct rebuttal to the claim that success in math and code should transfer to science. The verifier, not only scale, may decide the outcome.

ArgumentA correct result is not evidence of a self-correcting scientific process if the agent gathers evidence without updating its beliefs.

Why readTrace-level evaluation exposes epistemic failure that outcome scores hide. It is a compact case for judging the process, not only the answer.

ArgumentGovernments cannot plan AI for science without first mapping which datasets exist, who controls them and whether machines can actually use them.

Why readLess glamorous than a model launch, but far more concrete about the missing public infrastructure on which many launch claims depend.

ArgumentTransforming ideas into scoped projects, funded teams and executable institutions is itself a piece of research infrastructure.

Why readIt treats metascience as something to build and test, making it a useful answer to the question of where the bottleneck moves when hypotheses get cheaper.

ArgumentUseful BioML work depends more on strong foundations, comfort with messy data and rapid question testing than on memorizing large amounts of biology.

Why readA practical map of the hybrid scientific and engineering talent AI4S labs need, written by someone hiring and building in the field.

ArgumentBetter structures and molecular suggestions do not remove the slow biological, pharmacokinetic, toxicological and clinical failure modes that dominate drug development.

Why readA representative dispatch from medicinal chemistry rather than a generic blog index. Lowe’s durable point is that structure can be difficult without being the rate-limiting step.

ArgumentPowerful AI could compress fifty to one hundred years of biological and medical progress into five to ten years.

Why readThe most cited optimistic claim belongs here because it places the bet somewhere testable: is the bottleneck ideas, or experimental throughput, trials and regulation?

ArgumentA high-impact methods paper without runnable code cannot meet the reproducibility standard expected of scientific publication, especially when the result comes from a corporate lab.

Why readThe letter matters because it tested whether journal norms still bind frontier corporate research. AlphaFold 3’s later code release made the dispute consequential rather than symbolic.

ArgumentGNoME’s proposed structures offer scant evidence of compounds that are simultaneously novel, credible and useful without synthesis expertise and validation.

Why readThe key distinction is between computational discovery and experimental discovery. Later corrections in autonomous labs return to exactly this boundary.

ArgumentReanalysis of A-Lab’s reported products found systematic diffraction and novelty problems, so the original evidence did not establish 43 scientifically new materials.

Why readThe archive’s hardest external check: it uses the authors’ diffraction data rather than a rival forecast. The later correction adopts the same distinction between new to a platform and new to science.

ArgumentData leakage is a cross-disciplinary failure mode that can make machine-learning results look dramatically more reliable than they are.

Why readIf you read one reliability piece, choose this one. It audits the foundation of ML-based science rather than another isolated model.

ArgumentAI changes discovery through a recurring set of roles, but its value depends on the data, representations, experiments and institutions of each field.

Why readIf you read one overview, choose this. The comparison across fields is more informative than any single method it surveys.

ArgumentAcross disciplines, papers and patents have become less likely to break with prior work over several decades.

Why readEvery claim that AI is accelerating science needs this baseline. The comparison is not zero progress, but a long-running decline in disruption.

ArgumentScientific institutions can be deliberately redesigned and tested so the discovery ecosystem learns how to improve its own social processes.

Why readIt answers the question raised by cheap hypotheses: attention, funding and experimental capacity become allocation problems that models alone cannot solve.

ArgumentAccurate structure prediction is not the same as understanding the physical and chemical principles that govern protein folding.

Why readFor readers past the popular explanation. The distinction between prediction and understanding will recur in every field that follows.

ArgumentMaintaining the same rate of progress across several fields has required rapidly growing research effort.

Why readThe denominator for the whole debate: does AI change the slope of declining research productivity, or merely add another large cohort of researchers?

ArgumentMethods that scale with computation eventually outperform approaches built around human knowledge and handcrafted structure.

Why readAI4S architecture debates repeatedly return here. Keep its boundary visible: the historical evidence comes from domains rich in data, while science is often data-poor.