Evolutionary computation: The next deep learning

Description

Deep learning (DL) has transformed much of AI and demonstrated how machine learning can make a difference in the real world. With DL, the massive expansion of available training data and compute gave neural networks a new instantiation that significantly increased their power. Evolutionary computation (EC) is on the verge of a similar breakthrough. Risto Miikkulainen explains why.

While DL is focused on modeling what we already know, EC addresses a different but equally far-reaching problem: creating solutions that do not yet exist. EC accomplishes this not by following a gradient (like most DL and reinforcement learning approaches) but by doing massive exploration: using a population of candidates to search the space of solutions in parallel, emphasizing novel and surprising solutions. Thus, EC makes a host of new AI applications possible, from designing more effective and economical buildings, vehicles, and websites to discovering more effective and efficient behaviors for robots and virtual agents to more effective and cheaper health interventions and growth recipes for agriculture. In that sense, EC is the next frontier of AI beyond deep learning.