Catalyzing breakthroughs in science
By proving it could navigate the huge search space of a Go board, AlphaGo demonstrated the potential for AI to assist us higher understand the vast complexities of the physical world. We began by attempting to unravel the protein folding problem, a 50-year grand challenge of predicting the 3D structure of proteins – information that’s crucial for understanding diseases and developing recent drugs.
In 2020, we finally cracked this longstanding scientific problem with our AlphaFold 2 system. From there, we folded the structures for all 200 million proteins known to science and made them freely available to scientists in an open-source database. Today, over 3 million researchers all over the world use the AlphaFold database to speed up their necessary work on every little thing from malaria vaccines to plastic-eating enzymes. And in 2024, it was the respect of a lifetime for John Jumper and I to be awarded the Nobel Prize in Chemistry for leading this project, on behalf of your entire AlphaFold team.
Since AlphaGo’s win, we’ve applied its groundbreaking approach to many other areas of science and arithmetic, including:
Mathematical reasoning: Essentially the most direct descendant of AlphaGo’s architecture, AlphaProof learned to prove formal mathematical statements using a mixture of language models and AlphaZero’s reinforcement learning and search algorithms. Alongside AlphaGeometry 2, it became the primary system to attain a medal-standard (silver) on the International Mathematical Olympiad (IMO), proving AlphaGo’s methods could unlock advanced mathematical reasoning and laying the inspiration for our most capable general models.
Gemini, our largest and most capable model, recently went even further. A sophisticated version of its Deep Think mode achieved gold-medal level performance on the 2025 IMO using an approach inspired by AlphaGo. Since then, Deep Think has been applied to much more complex, open-ended challenges across science and engineering.
Algorithm discovery: Just as AlphaGo looked for one of the best move in a game, our coding agent AlphaEvolve explores the space of computer code to find more efficient algorithms. It had its own Move 37 moment when it found a novel strategy to multiply matrices, a fundamental mathematical operation powering nearly all modern neural networks. AlphaEvolve is now being tested on problems starting from data center optimization to quantum computing.
Scientific collaboration: We’re integrating the search and reasoning principles pioneered with AlphaGo into an AI co-scientist. By having agents ‘debate’ scientific ideas and hypotheses, this technique acts as a collaborator able to performing the rigorous considering mandatory to discover patterns in data and solve sophisticated problems. In validation studies at Imperial College London, it analyzed a long time of literature and independently arrived at the identical hypothesis about antimicrobial resistance that researchers had spent years developing and validating experimentally.
We’ve also used AI to higher understand the genome, advance fusion energy research, improve weather prediction and more.
As impressive as our scientific models are, they’re highly specialized. To attain fundamental breakthroughs like creating limitless clean energy or solving diseases that we don’t understand today, we want general AI systems that may find underlying structure and connections between different subject areas, and help us to provide you with recent hypotheses like one of the best scientists do.
Way forward for intelligence
For an AI to be truly general, it needs to grasp the physical world. We built Gemini to be multimodal from the start so it could understand not only language, but in addition audio, video, images and code to construct a model of the world.
To think and reason across these modalities, the newest Gemini models use a number of the techniques we pioneered with AlphaGo and AlphaZero.
The subsequent generation of AI systems may also must have the option to call upon specialized tools. For instance, if a model needed to know the structure of a protein it could use AlphaFold for that.
We expect the mix of Gemini’s world models, AlphaGo’s search and planning techniques, and specialized AI tool use will prove to be critical for AGI.
True creativity is a key capability that such an AGI system would want to exhibit. Move 37 was a glimpse of AI’s potential to think outside the box, but true original invention would require something more. It could must not only provide you with a novel Go strategy, as AlphaGo impressively did, but actually invent a game as deep and chic, and as worthy of study as Go.
Ten years after AlphaGo’s legendary victory, our ultimate goal is on the horizon. The creative spark first seen in Move 37 catalyzed breakthroughs which might be now converging to pave the trail towards AGI – and usher in a brand new golden age of scientific discovery.
