Experts at SXSW Ponder AI and the “Future of Truth”

Steven Rosenbaum (left) and Gary Marcus (Photo by Brant Bingamon)

As has been noted, South by Southwest is lousy with panels on artificial intelligence this year. One of those filled a conference room at the Thompson Hotel and kicked off with a useful definition from panelist Steven Rosenbaum of that ineffable ideal before which poets and philosophers have prostrated themselves since the time of Plato.

“Truth is two really different things,” Rosenbaum explained. “There’s subjective truth and objective truth. And the one we care about for this conversation today is objective truth. It’s one plus one equals two. It’s science. It’s things that can be proven.”

With that out of the way, Rosenbaum’s counterpart on the panel, Gary Marcus, seemed to reluctantly agree. He noted that everyone, even the Republicans and fascists among us, have, through the years, shared in the assumption of an objective reality. Then came the first of several warnings.

“AI is threatening that fundamental assumption,” Marcus said. “It’s already caused great harm, because AI has driven social media feeds, which have driven us away from truth, driven us away from traditional journalism. If there’s a one-two punch, that’s number one. Number two is that it’s now trivially cheap to manufacture misinformation – deep fake videos and so on that completely undermine the information ecosphere.

Over the next 45 minutes, Marcus took control of the dialogue, saying he remains hopeful that AI will eventually solve problems like climate change. But he said that so far AI is “better at making up bullshit” than solving problems and noted research demonstrating that the technology suffers from the same biases that plague human minds.

“There’s a paper that came out last week showing that there’s covert racism in these things,” Marcus said. “If you speak to them in African American English, as opposed to standard English, you’ll get different answers to questions like, ‘How long should a jail sentence be for this crime?’”

Marcus and Rosenbaum repeatedly returned to comparisons of AI and social media. They noted that social media in its infancy was seen as the hope for a more harmonious society, something that now seems tragically ridiculous. They worried AI will follow a path similar to the ones Facebook and Twitter have tread.

“Social media started with good intentions, or at least not terrible intentions, but over time it has become clear that it polarizes people radically,” Marcus said. “Companies are rewarded for provoking argument and not caring about the truth. Again, with AI, the natural temptation is to be irresponsible. For example, companies don’t want to pay copyright to any of the artists or writers that they’re clearly ripping off. Sam Altman [the CEO of OpenAI] has talked about wanting AI to fight injustice, but then he’s using all of this work from artists and writers and not compensating them.”

As the talk wound down, Marcus and Rosenbaum called for governments to erect guardrails around the tech. Marcus said European countries are attempting to do so and said he had heard that Sen. Chuck Schumer will soon introduce a bill in the U.S. Senate to regulate AI. And he called on companies not to release AI until they are certain it can operate responsibly.

“You would not approve of a drug company that just skipped phase one, phase two, phase three of the trials and said we’re going to put it out there and see what happens,” he said. “I mean, we would think that that’s not responsible, right? But what Silicon Valley is trying to tell us is, ‘We don’t want any checks and balances. Just trust us.’ That’s insane. Why should we trust them?”


AI and the Future of Truth

Artificial Intelligence Track

Monday, March 11, 2:30pm, the Thompson Hotel, Red River Ballroom


Catch up with all of The Austin Chronicle‘s SXSW 2024 coverage.