PERSON
Conrad Waddington
The British embryologist who drew a hillside of branching valleys to explain how a single fertilized egg becomes a liver cell here and a neuron there—and whose epigenetic landscape was rediscovered, name and all, by the engineers training
neural networks two-thirds of a century later.
Every scientist who draws a picture apt enough to escape its field and wander through others, picking up meanings the author never intended, deserves the kind of rereading that the AI moment now demands of Conrad Hal Waddington. In 1957 he published in
The Strategy of the Genes an image that has become one of the most reproduced diagrams in all of biology: a ball rolling down a hillside grooved with branching valleys, coming to rest at the bottom of one channel or another. He meant it to explain how development is at once flexible and reliable—how the same genome yields hundreds of cell types, each produced consistently despite the noise and variation of the biological world. He called the concept
the epigenetic landscape and the buffering of developmental outcomes against disturbance he called
canalization. Machine-learning researchers now title papers "A Waddington landscape for prototype learning" and describe network training