The Future of Fundamental Science Led by Generative Closed-Loop Artificial Intelligence

Hector Zenil^1*Jesper Nils Tegner^2*Felipe S Abrahão³Alexander Lavin⁴Vipin Kumar⁵Jeremy Graham Frey⁶Adrian Weller⁷Larisa Soldatova⁸Alan R Bundy⁹Nicholas R Jennings¹⁰Takashi Ikegami¹¹Lawrence Hunter¹²Sašo Džeroski¹³Andrew Briggs¹⁴Frederick D Gregory¹⁵Carla P Gomes¹⁶Jon Rowe¹⁷James Allen Evans^18,19Hiroaki Kitano²⁰Ross King²¹

• ¹Computational Medicine, Karolinska Institutet (KI), Solna, Sweden
• ²King Abdullah University of Science and Technology, Thuwal, Saudi Arabia
• ³Universidade Estadual de Campinas, Campinas, Brazil
• ⁴Simulation Institute, New York, United States
• ⁵University of Minnesota Twin Cities, Minneapolis, United States
• ⁶University of Southampton, Southampton, United Kingdom
• ⁷University of Cambridge Department of Engineering, Cambridge, United Kingdom
• ⁸Goldsmiths University of London, London, United Kingdom
• ⁹The University of Edinburgh, Edinburgh, United Kingdom
• ¹⁰Loughborough University, Loughborough, United Kingdom
• ¹¹Kabushiki Kaisha Riken, Kumagaya, Japan
• ¹²University of Colorado Anschutz Medical Campus School of Medicine, Aurora, United States
• ¹³Institut Jozef Stefan, Ljubljana, Slovenia
• ¹⁴University of Oxford, Oxford, United Kingdom
• ¹⁵US Army Combat Capabilities Development Command Army Research Laboratory, Adelphi, United States
• ¹⁶Cornell University, Ithaca, United States
• ¹⁷University of Birmingham, Birmingham, United Kingdom
• ¹⁸The University of Chicago Chicago Center for Contemporary Theory, Chicago, United States
• ¹⁹The University of Chicago Department of Sociology, Chicago, United States
• ²⁰The Systems Biology Institute, Tokyo, Japan
• ²¹University of Cambridge Department of Chemical Engineering and Biotechnology, Cambridge, United Kingdom

Artificial intelligence is approaching the point at which it can complete the scientific cycle, from hypothesis generation to experimental design and validation, within a closed loop that requires little human intervention. Yet, the loop is not fully autonomous: humans still curate data, set hyperparameters, adjudicate interpretability, and decide what counts as a satisfactory explanation. As models scale, they begin to explore regions of hypothesis and solution space that are inaccessible to human reasoning because they are too intricate or alien to our intuitions. Scientists may soon rely on AI strategies they do not fully understand, trusting goals and empirical pay-offs rather than derivations. This prospect forces a choice about how much control to relinquish to accelerate discovery while keeping outputs human relevant. The answer cannot be a blanket policy to deploy LLMs or any single paradigm everywhere. It demands principled matching of methods to domains, hybrid causal and neurosym-bolic scaffolds around generative models, and governance that preserves plurality and counters recursive bias. Otherwise, recursive training and uncritical reuse risk model collapse in AI and an epistemic collapse in science, as statistical inertia amplifies flaws and narrows the investigation. We argue for graded autonomy in AI-conducted science: systems that can close the loop at machine speed, while remaining anchored to human priorities, verifiable mechanisms, and domain-appropriate forms of understanding. Keywords: AI4Science; AI-conducted science; closed-loop discovery; epistemic singularity; human–machine collaboration; time complexity of AI + human; hypothesis generation; interpretability; model collapse; cognitive collapse; domain–method alignment; graded autonomy; neurosymbolic scaffolds.