In 2024, two Nobel Prizes were awarded for AI-driven science. In 2025 alone, governments and the private sector invested over $112 billion in AI research and development. What follows is a field-by-field account of the discoveries already reshaping our world.
Medicine & Drug Discovery
AlphaFold 3 — Solving the Protein Puzzle
Google DeepMind's AlphaFold series predicted the 3D structure of virtually every known protein — a problem that stumped biologists for 50 years. AlphaFold 3, launched in 2024, extended this to DNA, RNA and drug-like molecules, unlocking targets once considered "undruggable." The work earned DeepMind's Demis Hassabis and John Jumper the 2024 Nobel Prize in Chemistry.
Alzheimer's Cause Uncovered via AI Protein Visualisation
Researchers at UC San Diego confirmed that a gene long thought to be merely a biomarker for Alzheimer's is actually one of its causes — a distinction only made possible because AI allowed them to visualise the three-dimensional structure of the associated protein. The finding opens entirely new treatment avenues.
Drug Timelines Compressed from Years to Months
SandboxAQ, partnered with UCSF's Institute for Neurodegenerative Diseases, used large quantitative AI models to compress research timelines for Parkinson's therapy from years to months — at a fraction of the usual $2.5 billion cost per drug. Over 6,000 companies globally are now engaged in AI-assisted drug discovery.
AstraZeneca: 90% of Drug Pipeline Now AI-Assisted
AstraZeneca reports that more than 90% of its small-molecule discovery pipeline is now AI-assisted. New tools including MapDiff for inverse protein folding are accelerating drug design, lead optimisation, and clinical trial patient selection.
Google AMIE — AI Diagnostics Match Primary Care Physicians
Google's AMIE — a conversational medical AI agent — demonstrated clinical reasoning and communication on par with primary care physicians in both multimodal and multi-visit settings. It is being tested under physician oversight in partnership with Beth Israel Deaconess Medical Center.
Materials Science
MatterGen — AI That Designs New Materials from Scratch
Microsoft's MatterGen is a generative AI that starts with a random atomic structure and gradually sculpts it to produce entirely novel inorganic materials with specific properties on demand. Trained on over 600,000 examples and working across the entire periodic table, it radically accelerates the materials discovery cycle.
Self-Healing & Catalytic Materials on the Horizon
Carnegie Mellon researchers are using AI-integrated molecular simulations to enable expert-level design of novel compounds — opening paths toward self-healing materials for infrastructure, catalysts that break pollutants into useful byproducts, and materials that enhance soil fertility for harsh-climate farming.
AI Finds Low-Carbon Concrete Alternatives
An MIT team used machine-learning frameworks to analyse scientific literature alongside over one million rock samples, identifying new ingredient alternatives for concrete that are both cost-competitive and significantly lower in emissions.
"AI is emerging as a transformative general-purpose technology in scientific research that can unearth discoveries that would have otherwise remained hidden." — World Economic Forum, Top 10 Emerging Technologies 2024
Climate Science & Extreme Weather
100-Year Climate Projections in 25 Hours
Researchers at UC San Diego and the Allen Institute for AI developed Spherical DYffusion, a model that projects 100 years of global climate patterns in just 25 hours — 25 times faster than current supercomputer methods.
Predicting "Gray Swan" — 1-in-1,000-Year Weather Events
Scientists are now combining AI with physics-based climate models to predict extreme weather events so rare they occur only once in a millennium. Google released its most advanced weather forecasting model in 2025, generating forecasts eight times faster than its predecessors.
Real-Time Wildfire Detection & Evacuation Planning
ALERTCalifornia's network of over 1,200 AI-powered cameras uses AI to spot smoke in real time, monitor fire spread, and help emergency managers plan evacuations — a direct life-saving application in climate-driven disaster response.
Physics-Informed AI for Climate Modelling
A University of Hawaiʻi team published a new algorithm that allows AI to adhere to laws of physics while processing complex climate datasets — producing physically plausible outputs even with sparse data, with major implications for meteorology and renewable energy planning.
Food Technology & Agriculture
Precision Agriculture with TinyML Sensors
TinyML AI sensors deployed directly in fields now monitor crop health and optimise irrigation in real time, leading to measurably increased yields and reduced water usage — at costs low enough to deploy at scale in developing agricultural regions.
Quantum + AI for Fertiliser Optimisation
Agriculture researchers at Cleveland Clinic and IBM are testing AI-quantum hybrid applications for fertiliser calculations and field monitoring that could optimise crop yields while minimising environmental contamination from nutrient runoff.
AI-Designed Materials for Harsh-Climate Farming
Microsoft's MatterGen research is targeting materials that promote sustainable crop growth in harsh climates — directly relevant to combating desertification and food insecurity in climate-stressed regions.
Energy
AI-Designed Batteries & Perovskite Solar
AI-accelerated materials research produced breakthroughs in solid-state battery design and advanced perovskite solar cell materials in 2024–2025, driving record adoption of solar panels and accelerating the global transition away from fossil fuels.
Analog Optical Computing — AI Inference at a Fraction of the Energy
Microsoft developed an analog optical computer that uses light instead of conventional digital electronics to run AI inference — published in Nature — pointing toward dramatically greener AI infrastructure and faster MRI and banking settlement applications.
Nuclear Fusion — AI Accelerating the Path to Net Energy
AI is being applied to plasma control and reactor design in fusion research, with 2025 seeing nuclear fusion pilot plants achieving net energy gain in prototype settings. AI also accelerated small modular reactor development and next-generation nuclear fuel recycling.
Space & Fundamental Physics
AI at CERN — Real-Time Particle Physics
Carnegie Mellon researchers at CERN's Large Hadron Collider developed automated AI algorithms that identify in real time which events to store from the 40 million 3D collision images produced every second — a data filtering challenge no human team could achieve.
Black Hole Event Horizons — AI as Theoretical Co-Scientist
Theoretical physicist Alex Lupsasca of Vanderbilt University used AI to help identify new symmetries in the equations governing black hole event horizons — an early glimpse of AI as a genuine scientific thought partner in frontier physics.
FireSat — AI Satellites for Wildfire Monitoring
FireSat, highlighted in the 2025 Stanford AI Index, significantly enhances wildfire prediction from orbit, providing real-time alerts that can save lives and ecosystems. AI is also being applied to satellite data analysis for climate pattern monitoring and planetary resource management.
Quality of Life & Human Augmentation
Brain-Computer Interfaces — Thought-Controlled Prosthetics
Advances in brain-computer interfaces in 2025 enabled thought-controlled prosthetics that are transforming lives for people with disabilities, with ongoing research exploring further applications in education, accessibility and cognitive assistance.
AI Doctors That Listen Like Humans
Conversational medical AI systems now demonstrate the ability to conduct multi-visit clinical conversations, understand patient history across sessions, and communicate medical reasoning at a level comparable to trained primary care physicians — with profound implications for expanding quality healthcare globally.
MRI Scans Accelerated by Optical Computing
Microsoft's analog optical computer prototype demonstrated the ability to dramatically slash the time required to perform MRI scans — reducing costs, expanding access and accelerating cancer detection globally.
"I've always thought if we could build AI in the right way, it could be the ultimate tool to help scientists, help us explore the universe around us." — Demis Hassabis, Nobel Laureate in Chemistry 2024, Co-founder of Google DeepMind
Sources & Further Reading
- Stanford HAI — 2025 AI Index Report: Science and Medicine
- UC San Diego Magazine — Nine Breakthroughs Made Possible by AI (2025)
- Microsoft Source — 10 Scientific Breakthroughs from Microsoft Researchers (Dec 2025)
- World Economic Forum — Large Quantitative Models: How AI is Accelerating Drug Discovery (2025)
- AstraZeneca — MapDiff & AI Protein Design Breakthroughs (Oct 2025)
- Google Research — Accelerating the Magic Cycle of Research Breakthroughs (2025)
- CAS Scientific Breakthroughs 2025: Emerging Trends to Watch
- Carnegie Mellon University — AI's Role in the Future of Discovery (Sep 2025)
- Argonne National Laboratory — Top Science Research Breakthroughs in 2025
- Science News — AI-Enabled Science: Discovery and Insight (Feb 2026)
Buy me a coffee