Fake or real? Importance of trust in media generated by AI

By now, you’ve likely seen the mind-blowing videos and images that the hottest new artificial intelligence tools can create.

From ChatGPT to Midjourney to Sora to Microsoft’s VASA-1 (which can turn a still image into a realistic looking video), the world is abuzz with the opportunities (and dystopian dangers) these tools pose.

However, as the line between what is really human and artificially created continues to blur such that the two may soon become indistinguishable, we must address an important social construct: trust.

We don’t talk much about trust in our everyday lives because, well, we usually don’t have to.

Trusting in each other often goes without saying. When you open a bottled water, do you smell it to make sure that no one stuck chemicals in it?

Before you open a package, do you shake it to make sure there’s no broken glass inside that could cut you? Of course not, we have a baseline level of trust with each other that allows us to devote our attention and energies to more productive pursuits, such as drinking the water and opening the package. If we had to check or taste or smell or inspect everything we encounter, we’d be exhausted.

Indeed, as sociologist Niklas Luhmann aptly put it: “A complete absence of trust would prevent [one] even getting up in the morning.”

AI tools generate realistic looking and sounding audio and “deepfake” videos which are indistinguishable from the “real” thing.

We’ve seen this happen with spoofed robocalls in the voice of President Biden. This election year, bad actors who are intent on interfering with the democratic process are surely scheming to release deepfake audio and video shortly before the polls open. This forthcoming explosion of AI-generated content poses two problems.

First, how can a democratic society that depends on an informed citizenry function when voters cannot trust or have faith in the images and sounds they hear when scrolling social media or watching television? At least 40 states have proposed laws to deal with the anticipated influx of AI-generated (or “enhanced”) political ads this upcoming political season. Rhode Island is in that forward-thinking group, and its House of Representatives recently approved a bill titled “Deceptive and Fraudulent Synthetic Media in Election Communications.”

Among other things, this bill would have tightly regulated the use of AI-generated political content, especially within 90 days of an election.

Although the Rhode Island Senate did not take up this bill, it was a good start.

But time (and this election season) would tell how effective it would be in addressing the problem of AIgenerated media in politics.

Second, how more difficult will our daily lives become if we have to enter every interaction asking ourselves (or others), “Is the person on this Zoom call real?” or “Was that really my child’s teacher’s voice on that voicemail?” This constant state of skepticism and high-alertness is going to exhaust us over time. It will cause friction, hurt feelings, and make people want to engage less with others (because it’s too much trouble to determine what’s real and what’s not).

While Edgar Allen Poe wisely advised us to “believe nothing you hear and only half of what you see,” he had no idea that there’d come a day when you won’t be able to tell which half to believe.

The solution to this problem is not readily apparent. The best we can do is to put the question on the table and remember the importance of trust in a civil society.

Brian J. Lamoureux is a partner at Pannone Lopes Devereaux & O’Gara LLC and an assistant professor of practice in the Providence College School of Business.

This election year, bad actors who are intent on interfering with the democratic process are surely scheming to release deepfake audio and video shortly before the polls open.