TWiT 1086: The Great Beagle Migration - Pope Leo XIV's 1st Encyclical & Ferrari's 1st EV

Where this was said

Can AI Write Like a Human? The Uncanny Valley of Prose

At 1:24:25 · chapter starts 1:24:10

The panel digs into the question of whether AI will ever match human literary creativity. Gary Rivlin makes his most memorable argument of the episode: AI is not artificial intelligence, it's alien intelligence — it has ingested the entirety of human creative output but possesses no common sense, no lived experience, and no understanding. ^{[1] — Gary Rivlin "AI is obviously artificial intelligence, but I think it's alien intelligence. It knows everything and understands nothing. It has no common…" 1:24:25 Play} He uses the New York Times's Curtis Sittenfeld experiment — same prompt given to a celebrated novelist and to an AI — to argue that great writing is immediately recognizable because it is, by definition, not the lowest common denominator. But Leo Laporte complicates the picture: a more recent NYT Kevin Roose quiz found that more than half of readers preferred AI-written passages and couldn't distinguish them from human writing. Gary Rivlin's son (age 12 in 2023) could correctly identify AI-generated text 80-90% of the time, suggesting younger, AI-native readers may have better detection abilities. The consensus: basic journalism and PR writing is already a commodity for AI; true literary feature writing is not, but the gap is narrowing.