Artificial intelligence is making rapid strides in scientific discovery, raising the possibility that an AI system could one day earn a Nobel Prize. While current models still rely heavily on human guidance, researchers believe that within the next decade or two, AI may achieve full autonomy in the scientific process—generating hypotheses, designing experiments, analyzing data, and making breakthrough discoveries without human intervention. The idea gained formal traction in 2016 when Hiroaki Kitano, a biologist and CEO at Sony AI, launched the Nobel Turing Challenge. The goal: create an AI capable of making a discovery on par with Nobel-winning research. The challenge envisions a future where AI can independently navigate the entire scientific workflow, from identifying a problem to publishing a transformative finding. Ross King, a chemical-engineering researcher at the University of Cambridge and a key organizer of the challenge, believes an AI scientist could win a Nobel Prize in as little as 10 years. “It’s almost certain,” he says, though he acknowledges the timeline depends on how quickly AI research evolves. Still, many scientists remain skeptical. They argue that today’s AI models—trained on vast amounts of existing human knowledge—are not truly creative. They generate ideas based on patterns, not original insight. To make a Nobel-worthy discovery, AI may need a fundamental shift in design and a major investment in foundational research. Yolanda Gil, an AI researcher at the University of Southern California, suggests that a billion-dollar government initiative focused on AI-driven science could accelerate progress dramatically. Despite these hurdles, AI is already playing a growing role in science. In 2024, the Nobel Prize in Physics was awarded to pioneers in machine learning, and half of the chemistry prize went to the team behind AlphaFold—a DeepMind AI that predicts protein structures with unprecedented accuracy. However, these awards recognized human contributions to AI, not discoveries made by AI itself. For an AI to win a Nobel, it must operate with high or full autonomy. This means it must decide what to study, how to investigate it, and how to interpret results—without human input at critical stages. Researchers are already seeing progress. Gabe Gomes and his team at Carnegie Mellon University developed Coscientist, an AI system that uses large language models to plan and run chemical experiments using robotic lab equipment. One version can perform complex computational chemistry tasks faster than a human researcher could in years. Meanwhile, AI is generating millions of new materials and predicting biological behaviors. James Zou at Stanford has shown that an LLM-based system can uncover overlooked patterns in data—such as the swelling of immune cells during severe COVID-19 cases—without being prompted by researchers. “The AI agent is beginning to autonomously find new things,” Zou says. Zou is also helping organize Agents4Science, a virtual conference where AI agents will write and review papers, marking a milestone in AI-driven science. While challenges remain—especially hallucinations in AI outputs—Zou believes these can be mitigated with human feedback. Sam Rodriques, CEO of FutureHouse, sees the final stage of AI in science as fully autonomous systems that ask their own questions and conduct experiments independently. He predicts such a system could make a Nobel-worthy discovery by 2030, particularly in fields like materials science or neurodegenerative disease research, where the challenges are vast and the need for innovation urgent. The journey toward AI as a scientific peer is underway. Whether it leads to a Nobel Prize for a machine remains to be seen—but the signs suggest it may not be far off.