The article also discusses the shifts in AI research, from knowledge-based systems to machine learning in the early 2000s, and then to deep learning in the 2010s. It also mentions the recent rise of reinforcement learning. The author concludes by suggesting that the 2020s could see the end of the deep learning era and the emergence of a new paradigm in AI research, although it is unclear what this will be.
Key takeaways:
- The field of artificial intelligence has seen a shift towards machine learning since the late 1990s and early 2000s, with a rise in the popularity of neural networks beginning in the early 2010s, and growth in reinforcement learning in the past few years.
- Knowledge-based systems, which encode all human knowledge using rules, saw a decline in the early 2000s. Instead, researchers turned to machine learning, which allows machines to extract rules automatically from data.
- Deep learning, a subset of machine learning, didn't gain popularity immediately. However, a breakthrough in 2012 led to a surge in new research and its popularity exploded. More recently, reinforcement learning, which mimics the process of training animals through punishments and rewards, has seen a rapid increase in mentions in research papers.
- Despite the current popularity of deep learning, the field of AI research is characterized by the rise and fall of different techniques. Some predict that the era of deep learning may soon come to an end, with the 2020s seeing the reign of a different technique or possibly the creation of an entirely new paradigm.