The Real Story: Understanding A.I.'s risk to humanity

SOUTH BEND, Ind. -- A new Politico poll shows a majority of Americans think A.I. could destroy humanity. So how concerned should we be with the pace at which A.I. is developing and what’s the solution?

Between new reports of A.I. cyberattacks, warnings from industry leaders, and the recent resignation of an A.I. safety researcher that's been gaining traction on social media, A.I. has been in the news a lot.

Nitesh Chawla, PhD, is a professor of computer science and engineering and Director of Notre Dame's Data, A.I., and Computing Initiative.

“There's been a lot of concerns and conversations happening about— should we be pausing A.I., should we be pausing the development of the rapid pace, and the concerns about you know, the self-recursive improvement of these AI algorithms and systems,” said Chawla.

He says these concerns are valid.

“I don't recall when last we had a technological disruption like this, where, you know, every couple of months we see new abilities, new features, and new capabilities being introduced,” said Chawla. “At the same time, there's also a lot of advantages and benefits coming out of the use of A.I.”

The Anthropic researcher who resigned posted a long thread on X that points to large A.I. companies acting irresponsibly, gambling with people’s lives very notably–he says:
“The people building A.I. earnestly believe that it could kill us all by the end of the decade.”

Another Anthropic employee said he personally thinks there’s a greater than 10 percent chance that could happen.

So ABC57’s Jordan Tolbert asked Chawla about that.

“I think first of all, right, we should not confuse uncertainty with safety, and we also should not confuse a warning with a scientific forecast, like that 10 percent probability. How was it established? What was the data? What was the process behind it?” said Chawla. “I'll sort of discount it from that perspective— that [was it] a single person's opinion, or was it something that was scientifically, empirically established? And none of that— none of those facts are available to us. But at the same time, there is uncertainty.”

Chawla says whether it’s a .5 percent chance, a 10 or 20 percent chance–these concerns should be taken seriously.

“What I can say with confidence is today that the risk is not zero, and the consequences could be enormous, and we have warnings. We have a runway to start building these safeguards now,” said Chawla.

So how could A.I. actually wipe humans out?

Chawla said there are many unknowns with this tech, and he wouldn't necessarily equate it to wiping out humanity—but says there are risks—like A.I. hacking into other systems.

“So, these programs are living within a grid; they're living within a computer network, they're living within a financial system, etc. Everything is digitized now, so there is a risk of those getting compromised,” said Chawla. “But is there risk about the existential scenario— to me— is uncertain. However, the harm today, whether it's fraud, deepfakes, cybercrime, or discrimination, or privacy, etc., those are real, we are seeing it. And so how do we sort of counter these things?”

He says the challenge is really the speed at which A.I. is developing. What a pause may look like both in the U.S. and globally is also an unknown. So, Chawla says the risks need to be identified and the systems evaluated–urgently.

At Notre Dame, they're trying to be part of the solution. The RISE A.I. program is working to help create a framework for responsible A.I.