Who Makes the Discovery: The Scientist or Artificial Intelligence?
Al-Anoud D. Al-Otaibi ¹
¹ LearTechX – Science & Technology
Correspondence: Info@leartechx.com
Received: 31 July 2026 | Published: dd/mm/2026
1. INTRODUCTION
Who makes a discovery: scientist or artificial intelligence? In today’s laboratories, the answer is no longer simple. Scientists still ask questions, shape the goals, and decide what counts as meaningful. But AI can now scan massive datasets, generate hypotheses, and uncover patterns faster than any human team. The results are a new kind of discovery process, where human imagination and machine speed work together. The real debate is not who wins, but how science changes when both become partners.
2. The Scientist as the Mind of Discovery
Human curiosity gives science meaning. Science begins with curiosity, and curiosity is deeply human. A researcher notices a gap, asks why a pattern exists, and decides which mystery deserves attention. AI can suggest options, but it does not feel wonder, doubt, or responsibility. That matters because discovery is not only about finding an answer; it is about choosing the right question. Recent research shows that while AI can improve productivity, it may also narrow the range of topics scientists explore, which means human judgment remains essential for keeping science bold and diverse [1, 2].
3. Artificial Intelligence as the Engine of Speed
Machines extend the reach of discovery. Artificial intelligence is changing scientific work by acting like a powerful accelerator. It can analyze huge data sets, propose likely hypotheses, and support experiment design in ways that save time and effort. Recent Nature coverage reports that AI assistants can help throughout the research process, from idea generation to interpretation [3]. In protein science, AlphaFold showed how machine intelligence could solve a long-standing problem and transform an entire field [4]. Still, AI does not “understand” science the way humans do; it expands the search space, but scientists must decide what the results mean and whether they matter [5].
4. Evidence from Recent AI-Agent Systems
The division of labor argued above is not only theoretical; it shows up directly in how recent AI-agent systems have been used in practice. A 2023 review already framed AI as a partner across every stage of discovery, from forming hypotheses to designing experiments and analyzing results, while stressing that each stage still needs a human check [6]. Two 2026 case studies illustrate this partnership concretely. Robin, an AI-agent system built by FutureHouse, proposed a candidate treatment strategy for dry age-related macular degeneration; researchers still had to run the wet-lab experiments Robin suggested and interpret what the results meant, but the overall cycle moved roughly 200 times faster than a typical human-only workflow [8]. Google's Co-Scientist system, built on a similar principle, generated a hypothesis about antibiotic-resistance gene transfer that matched what a human research team had taken about a decade to establish — but only after scientists had framed the problem and kept prioritizing which hypotheses to pursue [3].
Both cases point to the same conclusion: AI can compress years of work into days, but a Nature editorial accompanying these studies warned that this speed only helps science if humans keep verifying it, since AI agents can still fabricate results or misread data without anyone noticing [7]. A related 2026 study on AI adoption in laboratories found that scientists, rather than stepping back as AI tools spread, have been reasserting their oversight role — checking, correcting, and taking responsibility for what the AI produces [9].
5. Research Ethics in AI-Assisted Discovery
Faster discovery is only valuable if it stays trustworthy, and that trust rests on four reinforcing commitments. The first is human oversight: as the case studies above show, scientists still had to run experiments, catch fabricated or misinterpreted data, and take responsibility for conclusions, rather than accepting AI output at face value [7–9].
The second is transparency — disclosing clearly when and how AI contributed to a hypothesis, an analysis, or a written result, so that readers and reviewers can judge the finding accordingly rather than assume it was reached by traditional means alone. The third is accountability: a named researcher, not the AI system, must remain answerable for a study's findings and their limitations, since AI tools cannot bear responsibility the way a person can [5,7].
The fourth commitment is everyday research integrity: obtaining informed consent and protecting privacy where human data is involved, auditing AI-assisted analyses for bias, keeping methods reproducible and open, and giving honest credit for both human and AI contributions to a discovery. None of these safeguards is new to science, but AI raises their stakes, because a system that can generate a hypothesis in seconds can also spread an error, a bias, or a fabricated result just as quickly [1,2,7].
6. CONCLUSION
So, who makes the discovery? The honest answer is both but not equal in the same way. AI can accelerate the discovery, expand possibilities, and uncover hidden patterns, yet the scientist still provides the vison, ethics, and interpretation that turn data into knowledge. The future of science is not a competition between humans and machines. It is collaboration in which AI helps discoveries move faster, while scientists ensure it moves wisely. That partnership may become the most important discovery of all.
REFERENCES
[1] Hao, Q., Xu, F., Li, Y. et al. Artificial intelligence tools expand scientists’ impact but contract science’s focus. Nature 649, 1237–1243 (2026). https://doi.org/10.1038/s41586-025-09922-y
[2] Ding AW, Li S. Generative AI lacks the human creativity to achieve scientific discovery from scratch. Sci Rep. 2025 Mar 20;15(1):9587. doi: 10.1038/s41598-025-93794-9. PMID: 40113940; PMCID: PMC11926073.
[3] Gottweis, J., Weng, WH., Daryin, A. et al. Accelerating scientific discovery with Co-Scientist. Nature 655, 487–496 (2026). https://doi.org/10.1038/s41586-026-10644-y
[4] Jumper J, Evans R, Pritzel A, Green T, Figurnov M, Ronneberger O, Tunyasuvunakool K, Bates R, Žídek A, Potapenko A, Bridgland A, Meyer C, Kohl SAA, Ballard AJ, Cowie A, Romera-Paredes B, Nikolov S, Jain R, Adler J, Back T, Petersen S, Reiman D, Clancy E, Zielinski M, Steinegger M, Pacholska M, Berghammer T, Bodenstein S, Silver D, Vinyals O, Senior AW, Kavukcuoglu K, Kohli P, Hassabis D. Highly accurate protein structure prediction with AlphaFold. Nature. 2021 Aug;596(7873):583-589. doi: 10.1038/s41586-021-03819-2. Epub 2021 Jul 15. PMID: 34265844; PMCID: PMC8371605.
[5] AlphaFold and beyond. Nat Methods 20, 163 (2023). https://doi.org/10.1038/s41592-023-01790-6
[6] Wang, H., Fu, T., Du, Y. et al. Scientific discovery in the age of artificial intelligence. Nature 620, 47–60 (2023). https://doi.org/10.1038/s41586-023-06221-2
[7] Editorial. Why AI cannot do good science without humans. Nature 653, 650 (2026). https://doi.org/10.1038/d41586-026-01551-3
[8] Ghareeb, A.E., Chang, B., Mitchener, L. et al. A multi-agent system for automating scientific discovery. Nature (2026). https://doi.org/10.1038/s41586-026-10652-y
[9] Leonel da Silva, R.G., Du, L. & Eyal, G. Scientists might be reaffirming the relevance of human oversight as AI lands in labs. Digital Discovery (2026). https://doi.org/10.1039/D6DD00030D
