AI Revolutionizes Science: AlphaFold's Impact and Evolution
The world of science was forever altered by AlphaFold, an AI system that cracked the protein folding code. But its journey is far from over, and the future holds intriguing possibilities.
On December 24, 2025, WIRED delved into the past and future of this groundbreaking project with DeepMind's Pushmeet Kohli, whose team won the Nobel Prize in Chemistry for their work.
Unlocking the Secrets of Protein Folding
AlphaFold, an AI creation by Google DeepMind, celebrated its fifth anniversary, marking a significant milestone in the field of artificial intelligence. Over the years, its achievements have been remarkable, culminating in the Nobel Prize win.
DeepMind initially gained fame by teaching AI to master the ancient game of Go, but they soon set their sights on a more profound challenge: protein folding. This complex process, a mystery in modern science, involves amino acids folding to form proteins. AlphaFold2 emerged as a solution, accurately predicting protein structures in 3D with atomic precision.
The project's success led to a comprehensive database of over 200 million predicted structures, covering nearly the entire known protein universe. This resource is now a cornerstone for approximately 3.5 million researchers worldwide, and the Nature article describing AlphaFold's algorithm has been cited an impressive 40,000 times.
Expanding Horizons: From Proteins to DNA and Beyond
The story doesn't end with proteins. AlphaFold 3 represents a significant leap, extending AI's capabilities to DNA, RNA, and drugs. However, this transition isn't without challenges, such as dealing with 'structural hallucinations' in proteins' disordered regions.
But here's where it gets controversial: Is AI truly ready to take on such complex tasks?
The AI Co-Scientist: A New Era of Collaboration
Pushmeet Kohli, vice president of research at DeepMind, envisions a future where AI becomes an integral research partner. He believes in building AI models that can reason and collaborate with scientists, making these tools accessible to researchers worldwide, and tackling ambitious goals like simulating a complete human cell.
The 'AI co-scientist' concept is a significant development, featuring a system built on Gemini 2.0 that generates and debates hypotheses, akin to the scientific method in action. But does this mean AI will replace human scientists? Kohli suggests a shift in roles, with AI handling the 'how' and humans focusing on the 'what,' ensuring that problem identification remains a human endeavor.
AI's Impact on Drug Discovery and Beyond
Kohli highlights the system's ability to facilitate hypothesis generation, rapidly analyzing vast amounts of literature. For instance, in drug repurposing and bacterial evolution studies, AI agents' disagreements led to more refined research proposals, demonstrating the power of AI collaboration.
Looking ahead, Kohli is excited about understanding cells as complete systems and deciphering the genome. He believes that simulating an entire cell is a key to unlocking new possibilities in personalized medicine and beyond.
And this is the part most people miss: AI's role in simulating cells could revolutionize medicine, but how close are we to this reality?
The Future of AI in Science
As AlphaFold continues to evolve, the next five years promise exciting developments. Kohli's vision includes more powerful AI models, widespread adoption by scientists, and the realization of bold ambitions like cell simulation.
The journey of AlphaFold challenges us to consider the potential and limitations of AI in science. Will AI truly transform scientific research, or are there aspects that remain beyond its reach? Share your thoughts and join the discussion on the future of AI in science.
