Analysing 16,625 papers to figure out where AI is headed next (2019)

A study by MIT Technology Review analyzed 16,625 papers in the field of artificial intelligence to track its evolution over 25 years. The research suggests that the dominance of deep learning, which has been instrumental in AI advancements, may be waning. The analysis revealed three major trends: a shift towards machine learning in the late 1990s and early 2000s, a surge in neural networks' popularity since the early 2010s, and a recent growth in reinforcement learning. The transition from knowledge-based systems to machine learning was driven by the impracticality of encoding vast amounts of rules manually. Deep learning gained prominence after a breakthrough in 2012, leading to significant advancements in various applications. The rise of reinforcement learning, exemplified by DeepMind's AlphaGo victory in 2015, signifies another pivotal shift in AI research. The study underscores the cyclic nature of AI techniques, with each decade witnessing the rise and fall of different approaches. The future of AI remains uncertain, with competing ideas on what will succeed deep learning as the dominant paradigm.