CONCEPT
Digital Biology
Hassabis's thesis that biological phenomena—from protein folding to drug-target binding—are, at their most fundamental level, information-processing systems, and therefore exactly the kind of thing that artificial intelligence is uniquely suited to model, predict, and ultimately redesign.
The phrase
digital biology carries a thesis about the nature of life itself. Biology at its most fundamental level,
Demis Hassabis argues, can be understood as an information-processing system: phenomenally complex, emergent, the product of billions of years of evolution, but information processing nonetheless. DNA is a code; proteins are its products; the cell is a vast network of molecular signals. If this is right, then biology is the kind of thing that an artificial intelligence—which is itself an information-processing system capable of learning models from data and using those models to search enormous combinatorial spaces—is uniquely suited to describe. Hassabis has gone so far as to suggest that AI may be the perfect description language for biology: the patterns of life, too intricate for any human mind to hold, are exactly the patterns a
learning machine can capture. The reframing is radical. It treats the living world not as chemistry to be observed but as information to be modeled.