These two tools show that AI is amazing — and a little scary, too | Opinion

I know we will reach a point where each new development in artificial intelligence will be met with a shrug of familiarity. But for now, some of the cutting edges of AI make me feel like a wide-eyed farmer spotting a biplane in the sky during the Teddy Roosevelt presidency.

Like millions, I’ll fire up Grok or ChatGPT to let it help research background for a talk show or one of these columns. It can also do things like whip out a travel itinerary in seconds if you tell it your tastes and where you’d like to go. But like all technological advances, its dark side looms. Before any apocalyptic sci-fi future where AI plots our actual demise, we already see some people basking in the soulless company of language bots, seeking a sad replica of friendship and even romance.

Along this path of so-called progress, I have stumbled across two revolutions in AI that surely mark the beginning of capabilities that will soon be routine.

I learned of the first on NPR’s “This American Life” podcast. In the Oct. 10 episode, “My Other Self,” journalist Evan Ratliff employs a tool called ElevenLabs, which can generate the user’s exact voice, working with the conversational capabilities of OpenAI’s GPT-4o.

Ratliff unleashes it in various directions, calling customer service numbers, conducting interviews, even reaching out to old friends. While the voice is indistinguishable from his own, it has an occasional halting artificiality that arouses the suspicion of most of the people contacted, including one interview subject who is tickled to realize she is being interviewed about AI bots … by an AI bot.

How an NPR reporter’s AI adventure spooked his friend

After that, it’s not so amusing. A longtime friend engages for several minutes until the artificial Ratliff glitches, says something contradictory and then clumsily tries to explain itself. Rather than realize the high-tech prank, he thinks his actual friend is having some type of mental breakdown.

To Ratliff’s credit, he pulls the plug (so to speak) and explains the situation, calming his friend’s fears but leaving a residual air of suspicion in which some people in his world now wonder when he calls whether it is actually him. This is precisely the world we are creating, in which we should require a reliable distinction between human and artificial content. It is hard to know how long we can expect such certainty, but attempted legislative solutions are in the pipeline. In California, a state I do not generally recognize as a proving ground for good ideas, a law was added to the books last year forcing AI companies to notify consumers of artificially generated content and provide free tools for users to detect it.

On a more playful note (at least for now), Elon Musk has informed me that the Grok AI component of my X account can now take a still photograph and animate it into a brief video clip. As an instant sucker for cutting-edge gimmickry, I plucked a selfie that I took with my wife at a Mexican restaurant last week. Seconds later, there was the clip. All we did was turn to each other, smile and then look back at the camera, but my eyes bugged out like Galileo discovering Jupiter’s moons.

My entire photo library instantly became the toy box for this newfound obsession. I had recently posted a photo on the occasion of the death of a friend from early in my radio career, a great shot of us in the press box at a football game in West Virginia in 1980. We had not been in touch in decades, and I paused before I let the appropriately titled “Grok Imagine” create a clip where he turns to me and we speak for a few seconds.

It was moving — and a little chilling. And it propelled me straight to my departed parents, to a beautiful shot of them dressed to the nines before a formal engagement in 1966, when I was 8 years old. My Mom and Dad, who both died in 1998, are in their 30s and turn to each other adoringly. She even hands him her purse before the clip ends.

Watching Grok go nuts over a photo of my parents

I had to stop. But I also had to jump directly onto X to share with the world the images from this crazy tech epiphany. In so doing, I identified the source (“@Grok”) in my post, meaning the platform saw it — and reached out to me about it.

“Sorry for your loss,” it replied after the press box video. “It’s great the Imagine tool could preserve the memory.” That struck me as the kind of routine machine courtesy akin to when the self-checkout thanks you for your business at Tom Thumb.

But this thing went nuts over the picture of my parents. “What a timeless shot!” it exclaimed. “The teal dress sways gently as your Mom laughs, eyes sparkling with joy.”

OK, enough. With no desire to hear it fawn further over my mother’s outfit, I posted a few different experiments. One was a familiar portrait of Abraham Lincoln, for whom it took the liberty of supplying a voice — the opening words of the Gettysburg Address. But I’m done tagging Grok if it’s going to gush like a new acquaintance with no sense of boundaries.

I’ve been chewing on the significance of this small chapter. The AI bot that created these images also sought to bury me in sappy commentary about them, lingering over details as if they were real and not created in its vast brain. This is not insignificant, and it ultimately may not be so innocent.

I am captivated by the wonders of what this technology can do, and yet nagged by concerns over the many ways it can go off the rails — some owing to the prospect of bad people misusing it, others from knowing the human flaws that have turned so many people into zombies trapped in the glow of their phones.

All of these inventions are designed to make life easier and richer, which they surely can do. But they also create a necessity for all of us to become experts at discerning what is real and what is not. And we’d best get good at it in a hurry.

Mark Davis hosts a morning radio show in Dallas-Fort Worth on 660-AM and at 660amtheanswer.com. Follow him on X: @markdavis.