Home » News and Events » News » Deep learning’s world champion is concerned about AI future

Deep learning’s world champion is concerned about AI future

“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,” Professor Yoshua Bengio said in the interview with MIT Technology Review’s senior editor for AI, Will Knight.

Yoshua Bengio is well-known for being an expert in deep learning since the day deep learning was just an academic curiosity. He is a professor at University of Montreal, also the co-founder of Element AI in 2016 which has built a successful company helping firms explore AI applications.

In the interview with Will Knight, Mr. Bengio shared that AI will be the game changer to improve people’s lives everywhere. Yet, a few notes were taken about how AI should be developed.

Firstly, the democracy in AI research might lead to power concentration as big companies tend to get more and more powerful due to people’s priority. Whereas “it’s dangerous to have too much power concentrated in a few hands,” said Professor Yoshua.

Secondly, there is a need of establishing laws and treaties to prevent lethal uses of AI in addition, and set up a defensive system to prevent it as some countries can secretly develop AI weapons.

Furthermore, to achieve a human-level AI that satisfy human’s need in long term, we also need a long-term strategy as well as investment. And machine learning will be the foundation to make a major leap in AI innovation. “We need to be able to extend it do things like reasoning, learning casualty and exploring the world”, said Professor Yoshua. As machines are not able to project themselves in different circumstances, they need model. Hence, we need machines to be able to discover these casual models, enable them to learn from their mistakes.

To avoid casualty, the research of AI needs to follow certain set of rules to keep AI development under control. This is what the AIWS 7-layer Model is developing.