AI Winter

AI winter refers to a period of time during which funding for activities geared toward developing human-like intelligence in machines is lacking. AI (Artificial Intelligence) winter is characterized by decreased funding in artificial intelligence research, but it often coincides with a drop in public interest as well.

When funding sources dry up and companies quit investing in AI-related research and development, the rate of innovation in the field slows as it is left to only the most dedicated academics. AI winter is thought to occur when the limits of the current technology result in less dramatic progress, leaving academic institutions to work on incremental improvements until another discovery in AI is made.  

AI winter has been used to describe a few years or even decades during which interest and development of AI has essentially halted. Artificial intelligence suffers some unique public relations problems that most areas of technology do not face to the same extent. The goal of artificial intelligence was arbitrarily set in the 1950s by Alan Turing. He proposed the imitation game test, where a computer would have to be indistinguishable from a human. Since then, the doomsday vision of AI sees it replacing humanity as soon as it can imitate it. Unfortunately for AI researchers, this terrifying vision of a runaway singularity can discourage new funding even as the reality of just how far AI is from passing the Turing test may disappoint current funders. It is more disappointment than fear that saw funding for AI drop in the late '70s and then again in the late '80s to early '90s. Technology in computers improved immensely during that time, but AI lagged.

Artificial intelligence, as Turing envisioned for his test, is still a ways off. Computers have used their superior memory and processing power to beat top players in chess, Go and even Jeopardy, but these tend to be limited applications. The concept of AI and its goals have undergone a positive change. Rather than striving to match the generalist mind that humanity has been blessed with, AI now seeks to specialize in particular tasks through techniques such as deep learning.

Artificial intelligence machines now have the capabilities to teach themselves how to get better at things as diverse as recognizing the content of images, understanding natural language and anticipating a person's next action on a mobile device. These successes have pushed off the arrival of another AI winter because they are commercially viable. An AI that can guide a user through an online transaction is worth money, as is one that can answer a question in an online chat, rather than have a user come into a physical location or make a phone call. These tangible benefits have companies and governments investing in AI research in house, as well as in academic institutions.

As long as AI continues to advance in a direction where companies can see a potential cost savings or profit, the sector will be too hot for any kind of AI winter to set in.