DRIVING THOUGHTS

A few days ago, I read a New York Times article about the world's leading deepfake expert who no longer trusts his own eyes when determining whether a video is fake or real. Although he helped develop software designed to identify deepfakes, he said the technology used to create fake videos and images is improving too quickly to keep up.

The man is Harry Farid, 60, one of the world's foremost experts in digital forensics. The article, written by Eli Saslow, noted that Farid has spent more than two decades differentiating visual reality from digital deception, from requests made by “governments, human rights organizations, journalists, law enforcement agencies, and countless others who were increasingly confused and deceived by the online world.”

Farid's research has shown that most people can no longer reliably distinguish a real photograph from a digital creation, a real voice from an AI clone, or a genuine video clip from a complete fabrication.

What is most worrisome is this: “Lately, he has been failing his own tests.”

If the world's leading deepfake expert now doubts his own eyes when verifying whether a video, photograph, or voice recording is genuine, how can the rest of us hope to make correct judgments about the information we encounter online? How can we trust what we read, see, and hear when making decisions that affect our finances, health, relationships, and daily lives?

Farid said he remains confident that he can solve almost any AI mystery. The problem, however, is that every investigation takes time. By the time he completes an analysis, "a fake had hardened into a fact. A fact had blurred into doubt."

According to him, the average social media post has a half-life of less than 90 seconds. Think of the damage that can be done in just three minutes—twice the lifespan of an average post. By then, misinformation may already have spread to thousands, perhaps millions, of people.

Here is an even harder truth: creating a fake is cheap. Detecting one is expensive, time-consuming, and increasingly difficult.

One of the algorithmic tools Farid helped develop analyzes a person's mannerisms, vocal inflections, and even blood flow. When a real person speaks, the eyes dilate and the heart pumps blood through the face. Farid can sometimes detect subtle changes in skin color that reveal a person's heartbeat in real time. An AI-generated avatar, by contrast, appears flatlined.

Most of us dismiss deepfakes as someone else's problem. After all, why would anyone spend time and effort creating a fake version of an ordinary person?

Yet even Farid became a victim.

Someone hacked his phone and used AI to clone his voice, placing calls to colleagues and even to his wife. The experience was unsettling enough that the couple created a "safe word" to verify their identities whenever they discussed sensitive matters over the phone.

So why am I writing about this?

The challenge today is no longer simply finding information. The challenge is deciding which information deserves our trust.

Because spotting fakes amid the flood of information we receive every day has become an essential life skill—one that we, and our families, must consciously develop.

We make decisions based on what we hear, read, and see. How can we make good decisions if we cannot distinguish fact from fiction?

Last Sunday, I had lunch with two friends, both senior citizens. One asked whether Iran had once again closed the Strait of Hormuz. Another repeated an opinion about the drowning of two Ateneo basketball players that clearly originated from fake news. Both avoided discussing the drama-filled Senate sessions because, as they put it, "those sounded and looked so unreal. Senators are supposed to be upright people. They don't behave that way."

Comments like these reveal either an inability—or an exhaustion—in trying to determine what is real before forming an opinion.

At the time, Iran had announced a renewed closure of the Strait of Hormuz, a development so significant that I never imagined anyone would suspect it was fake.

The Ateneo tragedy, meanwhile, has generated such a mountain of information that it would be easy for anyone to arrive at a conclusion based on incomplete or misleading reports.

As for the Senate, it is troubling that some people would rather look away than watch public officials behave in ways that seem inconsistent with the dignity expected of lawmakers.

So, if anyone knows of classes that teach the skills needed to distinguish fact from fiction in today's digital world, please send me an email. (Email: [email protected])