Deep-learning pioneer calls for democratization of AI

The technology could pose more of a danger when in the hands of just a few countries or organizations.
23 November 2018

Yoshua Bengio lectures at Data Science Summer School. Source: Ecole polytechnique/Flickr

One of the pioneers of deep-learning (DL), Yoshua Bengio, has called for the democratization of artificial intelligence (AI) knowledge and techniques so that the technology’s benefits can make a positive change everywhere, not just in the hands of a few.

“We could collectively participate in a race [for AI], but as a scientist and somebody who wants to think about the common good, I think we’re better off thinking about how to both build smarter machines and make sure AI is used for the well-being of as many people as possible,” he said in an interview with the MIT Technology Review.

Deep-learning technology tries to imitate the work of neurons in a brain to replicate thought. With it, software can literally learn to recognize patterns in images, sounds – indeed, any data that is digitizable.

The neocortex in a brain (that makes up around 80 percent of most brains’ mass) comprises of an immensely complex array of interconnected neurons. But via powerful computers and improved mathematical formulae, data scientists can model more layers of neurons more effectively.

CGI rendition showing interconnected neurons. Source: Pixabay

This esoteric area of abstracted mathematics existed only in academe for many years, but recently it has become mainstream, being used in commercial products as well as research environments.

Scientists have long worked in transnational groups that actively eschew political influence and ideology, but the open nature of the scientific publication and peer review means that the techniques behind AI and deep-learning have become to be used in military settings, and also by potentially suspect groups.

Bengio states that there’s a certain inevitability to the movement of talented individuals in the DL field towards monetary reward, rather than experts seeking academic accolades for the sake of pure scholarship.

“[…] AI research by itself will tend to lead to concentrations of power, money, and researchers. The best students want to go to the best companies. They have much more money, they have much more data. And this is not healthy. Even in a democracy, it’s dangerous to have too much power concentrated in a few hands.”

As a result of this natural drift towards existing centers of power, money, and influence, Bengio disdains the situation of the developing world with regards AI’s potential benefits, stating, “The potential for AI to be useful in the developing world is even greater. They need to improve technology even more than we do, and they have different needs.”

He calls for easier migration to improve the speed of AI development: “We could make it easier for people from developing countries to come here. It is a big problem right now. In Europe or the US or Canada, it is very difficult for an African researcher to get a visa. It’s a lottery, and very often they will use any excuse to refuse access. This is totally unfair.”

Also in the interview, Bengio points to the latest research in causality, reasoning, and learning as areas in deep-learning that have significant importance. Noting that humans themselves don’t have a perfect causal model of reality – that’s why we make mistakes.

However, for machines to achieve intelligence, “we need to be able to extend [AI] to do things like reasoning, learning causality, and exploring the world in order to learn and acquire information. If we really want to approach human-level AI […], we need long-term investments, and I think academia is the best place to carry that torch.”

Moreover, if a military power were to develop an AI-powered “killer robot”, our best defense might be building AI-powered defensive technology:

“There’s a big difference between defensive weapons that will kill off drones and offensive weapons that are targeting humans. Both can use AI.”