Sign In

Communications of the ACM

ACM Careers

We Analyzed 16,625 Papers to Figure Out Where AI Is Headed Next

View as: Print Mobile App Share:
deep learning, illustration

An analysis of all 16,625 papers available in the "artificial intelligence" section of the arXiv open-source database of scientific papers found three major trends: a shift toward machine learning during the late 1990s and early 2000s, a rise in the popularity of neural networks beginning in the early 2010s, and growth in reinforcement learning in the past few years.

Much current news about artificial intelligence is thanks to deep learning. But deep learning represents just a small blip in the history of humanity's quest to replicate intelligence. When you zoom out on the whole history of the field, it's easy to realize that it could soon be on its way out.

Every decade has essentially seen the reign of a different technique: neural networks in the late '50s and '60s, various symbolic approaches in the '70s, knowledge-based systems in the '80s, Bayesian networks in the '90s, support vector machines in the '00s, and neural networks again in the '10s. The 2020s should be no different, says Pedro Domingos, a professor of computer science at the University of Washington, meaning the era of deep learning may soon come to an end.

From Technology Review
View Full Article


FengTyan Lin

To complement this article, I have done a parallel study, which analyzed 18,624 artificial intelligence related papers published between 2009 and 2018 from Web of Science core collection database. It shows that, in the last 10 years, although machine learning is the dominant keywords and still increasing dramatically, knowledge-based systems are not dying, instead, it keeps moderately increasing. As for the number of papers whose keywords includes deep learning, it only shares 11.87% of those of machine learning. Since the number of papers concerning deep learning is still moderately increasing, it is probably too early to say that the era of deep learning may soon come to an end in 2020s. By identifying keywords with lower frequencies but very high growth rates, I suggest that blockchain, internet of things, big data, and image recognition will be shining stars in the near future.

Displaying 1 comment