Innovation in the Age after Deep Learning

In a recent article written by Karen Hao, the MIT Technology Review downloaded some 16,000 abstracts from arXiv in an effort to break down the trends in deep learning through time. What they found was that despite deep learning being at the forefront of AI research in the past decade, it seems likely to fizzle out in the next few years like all other technological trends.

Figure 1 - Changes in language used in deep learning articles, from Hao

Trends in deep learning have shifted from its limitations to its potential, as seen in Figure 1. With the incredible leaps forward that machine learning has made in the past five years or so, it’s no surprise that this advancement could eventually sputter and slow down, although it likely won’t stop. If we choose to believe that the age of deep learning could come to an end, then no matter the reason, it would make sense for us to begin reconsidering older methods in a new light, and begin searching for new approaches as well.

Article, by Karen Hao: https://www.technologyreview.com/s/612768/we-analyzed-16625-papers-to-figure-out-where-ai-is-headed-next/