Will AI destroy us... or save us?

April 9, 2026·51 min·3 clips

Each time machines beat humans at something—chess, then Go—humans redefine what 'uniquely human' means, in a pattern called the 'receding horizon.'

1. Throughline (NPR) examines the history of artificial intelligence and the persistent human question of whether machines can—or should—replicate human intelligence. 2. Three main expert voices shape the episode: George Tsarkadakis, an AI researcher and author of In Our Own Image; Francis Collins, geneticist and leader of the Human Genome Project; and Meredith Broussard, data journalism professor at NYU and author of More Than a Glitch. 3. The episode's driving question is why humans keep wanting to create intelligent machines, and what this desire reveals about human nature itself rather than about technology. 4. Tsarkadakis traces the impulse back to the 'big bang of the human mind' around 40,000 to 60,000 years ago, when Homo sapiens began making art and telling stories—the first evidence of humans projecting hopes and fears onto the unknown. 5. The episode shows how science fiction shaped real AI development: Tsarkadakis recalls seeing the robot Robbie in the 1956 film Forbidden Planet as a child in Athens, and he later surveyed scientists who consistently cited a book, comic, or film as the spark for their careers. 6. Historian Stephanie Dick explains that early computers were so large they filled basement rooms, produced intense heat and noise, and required workers to carry boxes of punch cards; these machines were operated largely by women doing what had previously been done by human 'computers.' 7. Dick identifies 'the most disturbing part of AI history': the men who built early AI looked at these industrial machines and concluded that their own minds were fundamentally the same—symbol-processing machines—an identification she ties to elite, homogeneous backgrounds that encoded specific biases into the field from the start. 8. The 1956 Dartmouth conference, convened by mathematician John McCarthy, opened with the declaration that every feature of human learning 'can in principle be so precisely described that a machine can be made to simulate it'—a statement Broussard calls 'enormously grandiose.' 9. The conference was, in practice, full of conflict and produced no coherent field, yet became a commemorated origin myth; a plaque at Dartmouth reads 'on this site artificial intelligence was born.' 10. Broussard argues this origin myth erases AI's roots in industrialization, capitalism, and colonial legacies—specifically the assumption that reason belonged only to certain kinds of people, a premise Babbage illustrated when he wished calculations 'had been produced by steam' rather than by the working-class people he found annoying. 11. The Turing test, proposed by British mathematician Alan Turing, was based on a parlor game about gender performance, and Broussard argues it reduces intelligence to the ability to 'tell a convincing lie, to put on the performance of being something that you're not.' 12. Chess was chosen as the intelligence benchmark by early AI because elite white men considered skill at chess a universal marker of intelligence—but Broussard argues this reflected their own biases, not a universal standard. 13. From the 1970s through the 1980s the AI field cycled through hype and disappointment, with funding drying up during periods known as 'AI winters'; Broussard notes that outside the US, especially in China and Russia, AI research flourished during this same period. 14. In 1997, IBM's supercomputer Deep Blue defeated reigning world chess champion Garry Kasparov in a six-game match in Manhattan, with Kasparov resigning after Deep Blue's 19th move in the final game—the first time a machine had defeated a reigning world champion under tournament conditions. 15. Kasparov described the defeat as 'painful' and said 'the history of us competing with machines will be over soon,' while a New York Times article responded by dismissing chess as 'a small problem' and challenging machines to beat humans at Go. 16. The episode introduces the 'receding horizon': each time a machine beats humans at a task—chess, then Go—humans redefine their uniqueness around some new capability, cycling through poetry, creativity, and emotion as the last bastions of the human. 17. Collins, who completed the Human Genome Project in 2000 by mapping three billion DNA base pairs with 2,400 scientists, warns that reducing humans to their genetic code misses dimensions that DNA cannot explain—such as why Beethoven's Third Symphony can bring him to tears. 18. A demonstration with ChatGPT writing Collins' biography introduces the episode's AI literacy angle: ChatGPT attributed his M.D. to Yale rather than North Carolina and omitted his 12-year NIH directorship under three presidents, then issued a boilerplate apology. 19. The episode is structured as a three-part narrative documentary with movie clips, archival news audio, and ChatGPT interaction interspersed with expert interviews, giving it a collage-style format. 20. Listeners drawn to the philosophical and historical roots of the AI debate will find the episode substantive; those seeking technical depth on current large language models or policy specifics will find the treatment more impressionistic.

As heard by us

A grounded look at AI that moves from Hollywood myth into industrial history.

The episode begins with the Hollywood version of AI, then strips that away and treats the term as a system that learns from data, changes its configuration, and is not a magic wand or a Terminator.

Read the full review in PlayNext →

Why you'd press play

If AI feels like a movie monster, this episode peels back the Hollywood gloss and looks at the machinery underneath.

Read the full recommendation in PlayNext →

Listen to the show on

Apple Podcasts Spotify

Play clips from this episode More from Throughline