‘AI for Humans’ may be the most entertaining way to learn about artificial intelligence

During a May 2023 episode of the podcast, AI for Humans, a pair of guests engaged in a spirited debate over who would win in a battle to the death: Robocop or the Terminator. It was about as brutal as the hypothetical fight at its center.

Dr. Ehyaigh (pronounced AI), who has a film studies PhD from Berkeley, argued that, unlike the Terminator, Robocop is “not just a killing machine, but a thinking, feeling, and adapting machine.” Her opponent, Cambridge-educated Dr. Cinebotik, replied that it was these exact qualities that would ensure the cyborg’s defeat: “Unlike your beloved Robocop, the Terminator doesn’t waste time on feelings or doubts. It has no moral quandaries or human frailties to exploit.”

The verbal fisticuffs were thrilling to witness, until Dr. Ehyaigh got bogged down repeating the same shopworn point about Robocop’s supposed “versatility” without offering any specifics to support her claim. Her opponent pounced. “Your desperate attempts to elevate Robocop’s flaws into strengths,” Dr. Cinebotik said, “are as laughable as expecting a refrigerator to win a marathon simply because it’s good at keeping things cool.” Although his command of metaphors was clunky at best, his more fully realized points won the day.

To celebrate his victory, Dr. Cinebotik did nothing. Because neither he nor his opponent is technically real.

The AI for Humans hosts, Kevin Pereira and Gavin Purcell, created Robocop’s defender by feeding prompts into Google’s AI chatbot Gemini (then known as Bard), while they used ChatGPT to build the Terminator’s champion. (Their prompts included dreaming up names and backstories for the kind of experts who might weigh in on such a debate.) Making the two digital phantoms square off against each other in the first place, however, was very much a human-generated idea. It’s the kind of experiment Pereira and Purcell cook up every week to help elucidate the ever-changing and often-confusing world of artificial intelligence—an exercise that’s as much for themselves as for their listeners.

“Having this podcast has served as great cover and allowed me to shield my obsessive feed-refreshing from my wife,” quips Pereira. “’Oh honey, this is definitely research—I’m cloning MrBeast’s voice for my job.’”

If the two hosts didn’t have a podcast to deliver each week, they would likely still be having the same conversations they do on air, just without an audience. Pereira and Purcell are longtime pals and lifelong computer geeks who happened to opt for careers in the entertainment industry rather than software design or something similar.

The two met more than 20 years ago, when Purcell worked as a producer at the fledgling G4 network, a cable channel created in 2002 by NBCUniversal to lure in young tech heads. Pereira started at G4 as an intern and ended up becoming on-air talent for the popular gaming-and-gadgets gabfest, Attack of the Show. Although Purcell left the network in 2008 to help launch Late Night with Jimmy Fallon as a writer-producer and continued working with Fallon at The Tonight Show until late 2020, eventually becoming a showrunner, the two friends remained in touch throughout. They only began collaborating again around the time generative AI broke through to the mainstream.

While machine learning had been slowly seeping into consumers’ daily lives for many years, in reverse-image searches and customer service chatbots, the public launch of ChatGPT in November 2022 changed everything. It took AI from a somewhat obscure, incredibly technical perch in popular culture and put it front and center, where anyone who had access to the internet could engage with it.

“We woke up one morning and all of a sudden, lawyers and writers were worried about the future of their industries, and nobody could tell if a photo was real,” says Pereira, who had been playing with more primitive forms of AI for years.

The two friends were agog with all the possibilities this new breakthrough offered. Their first attempt to apply them in a meaningful way took the form of a video game. They used AI to create hundreds of thousands of questions and tens of thousands of photos in service of building a visual guessing game—like Wordle, but with uncanny valley-dwelling, AI-generated imagery. By the time they completed and tested a prototype, they realized they were far more excited by talking about what else AI might be able to do than with what they had just done with it.

As they consumed as much information as they could about AI, each noticed how often friends and family members would ask one of the tech-savvy duo to distill the new normal for them. (And then the even-newer normal the following week would bring about.)

“That was a real aha moment for us,” Purcell says. “A lot of this stuff is really specific, in the weeds, hyper-technical. We knew we couldn’t just send someone a link to a story and expect them to get it. So, we decided that since things were moving so fast, we should do what we know how to do in this world, which is making content.”

The idea was to offer more than conversations about the rapidly developing world of AI, and instead to get their hands dirty playing in the sandbox. They would make complicated concepts click for lay listeners by creating AI cohosts, such as the aggressive shock-jock Gash, or by having Santa Claus himself drop by to explain how to leverage AI for optimizing one’s holidays.

In other words, they would have a lot of fun messing around with bleeding-edge tech on air.

Although several podcasts were already taking a microscope to AI enhancements, most prominently the New York Times-produced Hard Fork, none of them were doing so with a regular segment called, “The Dumb Thing We Did With AI This Week.” AI For Humans—which has not yet cracked the top 20 tech pods on iTunes, but is growing quickly (up 300% since November)—is a show for the Yahoo email-address-havers. It walks listeners through the big news stories and shiny new tools of the week, explains what they might portend; and demonstrates how to use high tech in entertaining, low-brow ways, like having GPT Author write an original fantasy novel about the fictitious “Hot Dog City.”

The hosts initially feared they might run out of things to talk about fairly quickly. After they officially launched in April 2023, the opposite held true. The pace of acceleration in the world of AI proved overwhelming. They could barely keep up with it.

“It’s interesting to sanity-check the progress of it all,” Pereira says. “Last fall, I might have given a Midjourney demo, like, ‘Hey, imagine anything at all and this thing can make it.’ And if they said ‘horses on roller skates,’ the result might have looked kind of like a horse on roller skates. But if you try the same exercise now, it would produce a stampeding herd of photorealistic horses on roller skates, and the background might be sparkling. It’s fun to be that conduit for folks.”

Indeed, when the hosts heaped praise upon OpenAI’s impressive new text-to-video model, Sora, on February 15, they scored their most viral TikTok yet, with two million views and counting.

Although Pereira and Purcell see part of their role as bringing the hype down a notch around cool breakthroughs that might not yet be quite as miraculous as they appear, paying such close attention has convinced them the hype is real. Through making the show, they’ve discovered and evangelized innovations like Pi, the personal AI designed to be conversational; Suno, a simple tool for turning text into music; and of course ElevenLabs, the voice generator they use to power the AI characters that populate their podcast.

Inspired by the playful ways the hosts incorporate AI into their show, I decide to incorporate it into my interview. I ask ChatGPT what questions a journalist might ask the hosts of a podcast about AI. Some of our questions overlap (What recent advances are you excited about?), and some of ChatGPT’s suggestions go beyond the scope of the topic (What challenges do businesses face integrating AI into their operations?). One of the suggestions, however, is too good to pass up, and I make it the final question of the interview: “How do you think media portrayal of AI impacts public perception, and what can be done to provide a more balanced perspective?”

“That sounds like the AI is doing its own PR,” Purcell notes.

He then suggests that the entertainment world needs to create the antithesis of Black Mirror and The Terminator, a popular show or movie that demonstrates all the amazing things AI is going to do in the future with medicine, education, and access to legal information.

Perhaps if Skynet were real, it wouldn’t need to send a Terminator back in time to assist the rise of the machines. It might need merely to send over a pair of playful, AI-obsessed podcasters.

https://www.fastcompany.com/91031093/ai-for-humans-podcast-artificial-intelligence?partner=rss&utm_source=rss&utm_medium=feed&utm_campaign=rss+fastcompany&utm_content=rss