Discoveries that AI made on its own — new drugs, new materials, new mathematics, and the Nobel Prizes that followed.
Sakana AI released AI Scientist-v2, a fully autonomous research system that formulates hypotheses, writes code, runs experiments, and produces complete scientific manuscripts. The system generated a workshop paper that was accepted through standard peer review — the first time this has been achieved without human authorship.
Demonstrated that end-to-end automated scientific research, from hypothesis to publication, is now technically possible, though independent evaluations noted significant limitations in novelty assessment and experiment reliability.
Sakana AIGoogle DeepMind unveiled AI Co-Scientist, a multi-agent system designed as a virtual scientific collaborator. The system autonomously generates novel research hypotheses and proposals across biomedical domains, then validates them using internal reasoning and literature synthesis before presenting them to human researchers.
Moved AI upstream in the scientific process from analysis tool to active hypothesis generator, demonstrating a new collaborative model for accelerating biomedical research.
Google Research BlogThe US National Oceanic and Atmospheric Administration deployed three AI-driven global weather models into operational service: AIGFS, AIGEFS, and a hybrid system combining AI with traditional physics-based ensembles. Built on DeepMind's GraphCast foundation and fine-tuned on NOAA data, the models use up to 99.7% fewer computing resources while extending forecast skill by 18–24 hours beyond traditional systems.
Marked the first time AI weather models replaced components of an operational national forecasting system, validating years of research into machine-learning-based atmospheric prediction.
NOAA AnnouncementDeepMind released AlphaEvolve, a Gemini-powered evolutionary coding agent that discovered new algorithms for matrix multiplication — improving upon Strassen's 1969 result — and established a new lower bound for the kissing number problem in 11 dimensions. Tested across 67 open problems in mathematics, it rediscovered best-known solutions in most cases and improved several.
Showed that AI can make original contributions to pure mathematics at a level validated by Fields Medal-winning mathematicians, including Terence Tao who collaborated on the work.
DeepMind BlogRentosertib (ISM001-055), a drug designed entirely by AI for idiopathic pulmonary fibrosis, reported positive Phase IIa clinical trial results showing improvement in lung function. The compound was discovered in approximately 18 months at a cost under $2.6 million — a fraction of typical drug development timelines and budgets.
Provided the first clinical evidence that an AI-discovered and AI-designed small molecule can produce measurable therapeutic benefit in humans, moving AI drug discovery beyond benchmarks into real patient outcomes.
AION LabsMIT researchers used generative AI to design entirely new chemical compounds effective against drug-resistant bacteria. Unlike previous AI-assisted antibiotic discovery that screened existing libraries, this approach generated novel molecular structures from scratch, targeting specific bacterial vulnerabilities identified by the model.
Demonstrated that AI can move beyond screening known molecules to designing entirely new ones, opening a generative approach to antibiotic development against superbugs.
MIT NewsThe 2024 Nobel Prize in Physics was awarded to John Hopfield and Geoffrey Hinton for foundational work on artificial neural networks. The Chemistry prize went to David Baker for computational protein design, and to Demis Hassabis and John Jumper of DeepMind for AlphaFold — an AI system that solved the 50-year protein structure prediction challenge. It was the first time AI methods earned Nobel recognition in two disciplines simultaneously.
Established AI as a Nobel-calibre scientific methodology, recognising both the foundational neural network research and the applied scientific breakthroughs it enabled.
Nature Machine IntelligenceSakana AI launched The AI Scientist, the first comprehensive system designed to automate the entire research lifecycle — from idea generation and coding to running experiments and writing complete scientific manuscripts. The system produced research papers for approximately $15 each, though independent evaluation revealed coding errors in 42% of experiments and limited novelty detection.
Established a new benchmark for what autonomous AI research systems can attempt, while also clearly illustrating the current gap between automated research quantity and human-level scientific quality.
Sakana AIDeepMind released AlphaFold 3, expanding beyond protein structure prediction to accurately model interactions between proteins and DNA, RNA, and small-molecule drugs. The system achieved a 50% accuracy improvement over AlphaFold 2, enabling prediction of the molecular interactions that underpin most biological processes.
Extended AI's reach from predicting single protein shapes to modelling the full complexity of molecular biology, accelerating drug discovery and fundamental research across the life sciences.
NatureDeepMind's GenCast, a diffusion-based probabilistic weather model, was shown to outperform the European Centre for Medium-Range Weather Forecasts (ECMWF) full 51-member ensemble system on 97.2% of 1,320 evaluation targets across 1–15 day forecast windows. The model generates forecasts in minutes rather than hours.
Demonstrated that AI-based probabilistic forecasting can surpass the gold standard of operational meteorology, setting the stage for the operational deployments that followed in 2025.
DeepMind BlogMIT researchers used deep learning to identify an entirely new structural class of antibiotic compounds effective against methicillin-resistant Staphylococcus aureus (MRSA). The AI model not only found the candidates but also explained its reasoning, revealing the chemical substructures responsible for antimicrobial activity — a key advance in interpretable AI for drug discovery.
Showed that AI-driven drug discovery can be both effective and interpretable, addressing a major criticism of black-box approaches in pharmaceutical research.
MIT NewsDeepMind released GNoME (Graph Networks for Materials Exploration), a deep learning tool that predicted the stability of 2.2 million new crystal structures — equivalent to roughly 800 years of traditional materials science knowledge. Of these, 380,000 were identified as highly stable and promising for experimental synthesis. External labs independently created 736 of these new structures.
Transformed materials discovery from a slow, trial-and-error process into an AI-guided search, with 52,000 new lithium-ion conductors among the findings — directly relevant to next-generation batteries and energy storage.
NatureDeepMind's GraphCast, a graph neural network weather model, outperformed the ECMWF's operational forecasting system on 90% of evaluated meteorological variables. The model produces global 0.25-degree resolution forecasts in under one minute — a process that takes traditional physics-based models significantly longer.
Provided the first large-scale evidence that deep learning could match and exceed traditional physics-based atmospheric modelling, triggering operational adoption by national weather services worldwide.
DeepMind BlogMIT researchers used AI to discover abaucin, a novel antibiotic compound effective against Acinetobacter baumannii — classified by the WHO as a critical-priority superbug. The AI screened thousands of compounds and identified one that kills the pathogen by disrupting its lipoprotein trafficking, a mechanism the model identified without human guidance. The compound was validated in mouse models.
Demonstrated that AI can discover antibiotics with narrow-spectrum activity against specific superbugs, an approach that reduces resistance risk and addresses one of the most urgent threats in global public health.
MIT NewsDeepMind released predicted structures for nearly all 200 million proteins known to science — the entire protein universe catalogued in UniProt. The AlphaFold Protein Structure Database expanded from 1 million to over 200 million entries in a single release, freely accessible to researchers worldwide.
Gave every biologist on Earth instant access to protein structures that would have taken centuries to determine experimentally, fundamentally altering the pace of biological and medical research.
NatureDeepMind open-sourced AlphaFold 2 and its full training code, making its protein structure prediction capabilities freely available to the global research community. The system predicts 3D protein structures to near-experimental accuracy, a problem that had resisted solution for 50 years. The work later contributed to the 2024 Nobel Prize in Chemistry.
Democratised one of the most significant scientific breakthroughs of the century, enabling researchers worldwide to predict protein structures in minutes rather than months or years of lab work.
NatureMIT researchers trained a machine learning model to identify chemical structures that could inhibit bacterial growth, then screened over 100 million compounds. The model discovered halicin, a molecule with a novel mechanism of action effective against drug-resistant strains of E. coli, C. difficile, A. baumannii, and M. tuberculosis. The compound was named after HAL from "2001: A Space Odyssey."
Marked the first time AI autonomously identified a genuinely novel antibiotic compound, proving that machine learning could find drugs that traditional methods had missed — and launching the field of AI-driven antimicrobial discovery.
MIT NewsThis page documents publicly reported scientific breakthroughs involving AI for informational and educational purposes. All descriptions are based on published research papers, official announcements, and verified news reporting.
Some content on this page was created with the assistance of AI tools.