Machine-learning researchers make many decisions when designing new models. They decide how many layers to include in neural networks and what weights to give inputs at each node. The result of all this human decision-making is that complex models end up being “designed by intuition” rather than systematically, says Frank Hutter, head of the machine-learning…