The tech industry is always abuzz with the latest news about artificial intelligence (AI) and how it is continuing to create a significant impact across many industries. However, venture capitalists are now questioning if the technology is just a hype. Back in 2015, Elon Musk announced that his company would release a self-driving car by 2018. When Tesla failed to achieve it by then, Musk moved said release to 2020.
Three months into 2020 and we have yet to see self-driving cars plying roads. As it turns out, it is a difficult feat to achieve. Why? Is it because AI is entering another wave of darkness? Let’s stroll down memory lane to so-called “AI winters.”
The History of AI Winters
An AI winter is a terminology coined by researchers after noticing a pattern in AI projects in the mid-1980s. According to observations since the late 1960s, AI collaborations followed a positive trajectory before declining due to issues that generally arose due to lack of funding. In other words, an AI winter is a time period when interest in and financial support for AI projects and research reduces and dries up.
AI winters have occurred in the past and another one may happen again. To date, we’ve seen two so far, one that occurred in 1974–1980 and the other in 1987–1993. In this post, we will look at the events that led up to the first AI winter.
The Machine Translation Project
One of the earliest studies on machine translation occurred in 1954. At this time, a system analyzed a 25-word dictionary syntactically. That successfully translated English into Russian words. The technology was widely received in light of the height of the Cold War. The hype surrounding the technology led to generous funding to support the project.
The Birth of Perceptrons
By 1958, Frank Rosenblatt introduced the concept of perceptrons or neural networks. His algorithmic experiments led to the discovery of the series-coupled perceptron, which we now know as the “feedforward layout of a neural network.” This AI technology allows machines to classify images (i.e., letters and shapes).
The First AI Winter
After a decade of improvements, AI progress slowed down in the 1960s mainly because many countered the systems developers were building just wouldn’t work.
In the case of the Machine Translation Project, researchers argued that computers would need too much information to work correctly. In 1966, the Automatic Language Processing Advisory Committee even published a report saying that machine translation does not even serve a purpose. That pronouncement eventually led to a decline in funding.
In 1969, Rosenblatt’s work on perceptrons was also highly criticized. A book by Minsky and Papert said perceptrons lacked “connectionism.” They can only solve linear problems. While Minsky and Papert established that only multiple layers could address the issue of connectionism, they said the algorithm necessary to teach a network is absent. And so, like the Machine Translation Project, interest in perceptrons dwindled, followed by a decline in funding.
Now that you know a bit more about AI winters, you may be wondering if another AI winter is indeed coming.
Are We on the Verge of Another AI Winter?
As notable AI projects have recently been pushed back, it is pretty understandable why many, particularly venture capitalists, question if an AI winter is upon us. However, various novel AI projects remain in use to this day. At the beginning of 2020, we saw several AI research projects receive significant funding, in fact.
An example would be an AI expert from the University of Hawaii who received a grant for his human-in-the-loop AI from the National Science Foundation. Travis Mandel’s project uses AI and machine learning (ML) systems to address real-world issues.
Another example is the U.S. Department of Energy (DOE)’s US$17 million research funding for using AI and ML to predict fundamental plasma phenomena, manage facilities, and accelerate discovery.
Even though an AI winter is always a possibility based on historical circumstances, its occurrence always depends on how willing funding agencies are to support projects to give way to breakthroughs. And should an AI winter happen again, so long as believers exist, AI will continue to change the world. In light of recent developments, however, it may be safe to say that an AI winter would have to wait.