Developments·Paper
Chalmers-built AI proposes, runs and learns from biology experiments in closed loop
Researchers at Chalmers University of Technology in Sweden built a closed-loop AI system that generates biological hypotheses, plans tests, instructs laboratory robots, analyzes results and refines subsequent questions. Tested on Saccharomyces cerevisiae, it combines LLMs with formal logic, biological databases, machine learning, automated cell cultivation and mass spectrometry. The study was published in the Journal of the Royal Society Interface.
3 engagement2 sourcesFri, Oct 2, 2026, 10:52 AM
AI Overview
- Chalmers University researchers built a closed-loop AI that generates hypotheses, designs tests, instructs lab robots, analyzes results, and refines follow-up questions
- The system, published in the Journal of the Royal Society Interface, combines LLMs with formal logic, biological databases, machine learning, automated cell cultivation, and mass spectrometry
- Researchers tested the system on Saccharomyces cerevisiae, using yeast's vast unexplored genetic and metabolic data despite it being one of biology's most-studied organisms
Entities
Related
Developments·Paper
Google reports real-world validation of Gemini-based Co-Scientist across materials science, biology, and CS
An arXiv paper extends and validates Co-Scientist, a Gemini-based multi-agent system for end-to-end scientific research. The system now moves beyond in silico hypothesis generation into execution-grounded closed-loop workflows, including interfacing with a semi-automated chemical vapor deposition reactor. It spans hypothesis generation, experimentation, and manuscript generation across materials science, biology, and computer science.
An arXiv paper extends and validates Co-Scientist, a Gemini-based multi-agent system for end-to-end scientific research. The system now moves beyond in silico hypothesis generation into execution-grounded closed-loop workflows, including interfacing with a semi-automated chemical vapor deposition reactor. It spans hypothesis generation, experimentation, and manuscript generation across materials science, biology, and computer science.
0 engagement1 sourcearxivThu, Aug 27, 2026, 06:58 AM
Developments·Benchmark
Autonomous Science Bench: arXiv benchmark tests AI agents on budgeted lab experiments across biology, chemistry, materials science
Researchers introduce Autonomous Science Bench, a benchmark of six budgeted design-build-test-learn campaigns in biology, chemistry, and materials science. Agents must choose experiments, receive real lab measurements, and deliver predictive models, evaluated via calibrated discovery and learning gates. In a pilot of eight agent-model configurations over 144 trials, the best succeeded in 67% of trials, with the gates exposing distinct failure modes.
Researchers introduce Autonomous Science Bench, a benchmark of six budgeted design-build-test-learn campaigns in biology, chemistry, and materials science. Agents must choose experiments, receive real lab measurements, and deliver predictive models, evaluated via calibrated discovery and learning gates. In a pilot of eight agent-model configurations over 144 trials, the best succeeded in 67% of trials, with the gates exposing distinct failure modes.
0 engagement1 sourcearxivTue, Oct 13, 2026, 02:18 PM
Developments·Paper
Survey charts past and future of AI Scientists automating science
A new arXiv survey reviews the past and future of AI Scientists—machines that automate science by originating hypotheses, designing and executing experiments, and revising beliefs. It traces the lineage from Adam, the first machine to make novel discoveries through physical experimentation, to integrated scientific agents. The survey frames AI Scientists as connected to literature, formal knowledge, simulations, and physical labs.
A new arXiv survey reviews the past and future of AI Scientists—machines that automate science by originating hypotheses, designing and executing experiments, and revising beliefs. It traces the lineage from Adam, the first machine to make novel discoveries through physical experimentation, to integrated scientific agents. The survey frames AI Scientists as connected to literature, formal knowledge, simulations, and physical labs.
0 engagement2 sourcesarxivFri, Aug 14, 2026, 03:50 PM
Developments·Paper
When Scientific Cognition Is No Longer Scarce
An arxiv paper, 'When Scientific Cognition Is No Longer Scarce,' argues that AI will make computer-based scientific work abundant while experiments, well-supported conclusions, and accountable institutional authority remain scarce. It names this shift the 'scarcity inversion' and examines its consequences for national laboratories and scientific governance.
An arxiv paper, 'When Scientific Cognition Is No Longer Scarce,' argues that AI will make computer-based scientific work abundant while experiments, well-supported conclusions, and accountable institutional authority remain scarce. It names this shift the 'scarcity inversion' and examines its consequences for national laboratories and scientific governance.
0 engagement1 sourcearxivWed, Oct 7, 2026, 03:25 PM
Developments·Paper
Lab positions A.I. agents as 'co-scientists' to speed drug discovery, but says research is only a proof of concept
Lab presents AI agents as 'co-scientists' to speed drug discovery, framed as proof of concept
Lab presents AI agents as 'co-scientists' to speed drug discovery, framed as proof of concept
0 engagement1 sourcenytimes.comSun, Sep 20, 2026, 05:43 AM