Yoshua Bengio

Alan Turing, founder of the field in 1950, thought that once we have machines that understand language, we might be doomed because they would be as intelligent as us.

He wasn't quite right.

So we have machines now that understand language, but they lag in other ways like planning.

So they're not, for now, a real threat, but they could in a few years or a decade or two.

So it is that realization that we were building something that could become potentially a competitor to humans, or that could be giving huge power to whoever controls it.

And destabilizing our world, threatening our democracy.

All of these scenarios suddenly came to me in the early weeks of 2023, and I realized that I had to do something, everything I could about it.

It is emotionally difficult.

And I think for many years, I was reading about the potential risks.

I had a student who was very concerned, but I didn't pay much attention.

And I think it's because I was looking the other way.

And it's natural.

It's natural when...

You want to feel good about your work.

We all want to feel good about our work.

So I wanted to feel good about all the research I had done.

I was enthusiastic about the positive benefits of AI for society.

So when somebody comes to you and says, oh, the sort of work you've done could be extremely destructive, there's sort of unconscious reaction to push it away.

But what happened after ChatGPT came out is really another emotion that

that countered this emotion.