🧬 How Google DeepMind's AlphaFold Is Transforming Biology and Pharma
At TED in San Francisco, Max Jaderberg—lead AI specialist at Isomorphic (a Google DeepMind lab)—shared how AlphaFold 3 is revolutionizing drug discovery and deepening our understanding of biology.
Key insights:
1️⃣ AlphaFold and the biological revolution
AlphaFold has driven a breakthrough in predicting the 3D structures of proteins—one of biology’s core challenges for the past 50 years. It was for this achievement that Demis Hassabis and John Jumper from Google DeepMind received the 2024 Nobel Prize in Chemistry.
In just six years since its debut, AlphaFold has helped researchers save what amounts to millions of years of "wet-lab" effort, condensing months of experiments into mere seconds of neural network compute.
Artificial intelligence is taking computer modeling to a whole new level. Now we can move beyond running complex calculations on a cosmic scale and instead dive into detailed modeling of cellular and molecular processes that cannot be fully captured by traditional equations.
2️⃣ Rising costs in drug development
Since the 1950s, bringing a new drug to market has grown exponentially more expensive (Eroom's Law). Meanwhile, computing power—once ruled by Moore's Law and now by "Huang's Law" for GPU performance—has soared. This surge in computational resources offers a path to surmount the challenges in pharma. As Jaderberg notes, scientists and drug designers can now "play" with an AI analog of biomolecular systems, instantly modifying molecular structures to see how well they bind to target proteins—work that used to take months in a real lab.
Now drug developers can "play" with a computer model: they can modify a molecule's chemical structure and see how it binds to a protein. Today, this takes few seconds—whereas it once required months of experiments.
3️⃣ AI agents join the game
Not only can humans interact with these AI analogs, but so can AI agents themselves. Thousands of agents running in parallel on large GPU clusters can rapidly search for potential drug candidates against various mutations, cancer types and other diseases. It's another step toward truly personalized medicine, where a therapy could be tailored to an individual patient's specific protein mutations.
📱 Watch the full TED here.
More on this topic:
🛑 The Largest AI Model for Protein Design
🛑 AI to Search for New Antibiotics
#AITED #Google #DeepMind @hiaimediaen
At TED in San Francisco, Max Jaderberg—lead AI specialist at Isomorphic (a Google DeepMind lab)—shared how AlphaFold 3 is revolutionizing drug discovery and deepening our understanding of biology.
Key insights:
1️⃣ AlphaFold and the biological revolution
AlphaFold has driven a breakthrough in predicting the 3D structures of proteins—one of biology’s core challenges for the past 50 years. It was for this achievement that Demis Hassabis and John Jumper from Google DeepMind received the 2024 Nobel Prize in Chemistry.
In just six years since its debut, AlphaFold has helped researchers save what amounts to millions of years of "wet-lab" effort, condensing months of experiments into mere seconds of neural network compute.
Artificial intelligence is taking computer modeling to a whole new level. Now we can move beyond running complex calculations on a cosmic scale and instead dive into detailed modeling of cellular and molecular processes that cannot be fully captured by traditional equations.
2️⃣ Rising costs in drug development
Since the 1950s, bringing a new drug to market has grown exponentially more expensive (Eroom's Law). Meanwhile, computing power—once ruled by Moore's Law and now by "Huang's Law" for GPU performance—has soared. This surge in computational resources offers a path to surmount the challenges in pharma. As Jaderberg notes, scientists and drug designers can now "play" with an AI analog of biomolecular systems, instantly modifying molecular structures to see how well they bind to target proteins—work that used to take months in a real lab.
Now drug developers can "play" with a computer model: they can modify a molecule's chemical structure and see how it binds to a protein. Today, this takes few seconds—whereas it once required months of experiments.
3️⃣ AI agents join the game
Not only can humans interact with these AI analogs, but so can AI agents themselves. Thousands of agents running in parallel on large GPU clusters can rapidly search for potential drug candidates against various mutations, cancer types and other diseases. It's another step toward truly personalized medicine, where a therapy could be tailored to an individual patient's specific protein mutations.
📱 Watch the full TED here.
🛑 The Largest AI Model for Protein Design
🛑 AI to Search for New Antibiotics
#AITED #Google #DeepMind @hiaimediaen