We no longer support this browser. Using a supported browser will provide a better experience.

Please update your browser.

Close browser message

in-person @jpmorgan

Created with Sketch.

Dario Gil

Dario Gil

What are the challenges of developing artificial intelligence?

How do we build AI that is secure? You know, because AI models are susceptible to attack. How do we build AI that does not suffer from bias? How do we build AI that can learn with less data? Very often in a business, we have lots of varieties of data, and a lot of it maybe is not sufficiently labeled. So how do we create AI that works in that manner? And lastly, this topic about what is a computational platform? How do we keep lowering the cost of building an AI model and deploying an AI model?

What is the future of AI?

AI used to be a story of just the tech players, right? You know, two, three years ago. AI now in businesses all over the world, everybody's creating their own AI teams, and I like to say that AI is the new IT now. And when IT emerged, right, in the 60s and 70s what happened? Well, what happened is that business the world over sort of intuitively understood that to be a modern corporation, one had to deploy IT. So that intuition that people have that AI is the new IT now needs to be manifested in reality.

How does moral reasoning work with AI?

If you have a self-driving car, how will you make a choice? One avenue is to then understand that, of how humans actually make these choices, and create AI that model that human behavior. So that would be an example of injecting moral reasoning. It is the effort of understanding what community you're serving, what are the moral preferences of that community, and making sure that an AI that is going to act with autonomy, reflects the moral preferences of the community it serves.

How do you respond to critics of AI?

I was arguing that we're in a narrow form of AI today. That we have this general form of AI that is conceptualized and not realized. What often happens is that while we're using this same term, we mean very different things. So when Elon Musk talks about, you know, the dangers of AI, he's really talking fundamentally in its broader sense about this general form of AI. So when the enthusiast of AI respond to that they're talking about present AI, and narrow forms of AI, and the advances that is happening, and they're speaking past each other. So that's why I think it's important that we have these categories of narrow, broad, general, because then we can be more precise in our conversation.


Dario Gil, Director of IBM Research, explains how artificial intelligence, or the “new information technology,” could provide more value to businesses.