Why do the new machines flatter, bluff and cut corners? A new essay at Intellicurean goes looking for the answer in a Grimm fairy tale that Carl Jung took seriously, and finds it closer to home than the machines.
Why we’re pointing you there
A poor student frees a spirit from the roots of an oak. It tells him its name, Mercurius, and promises to strangle whoever let it out. “Ever the Trickster” starts from that tale, then gives the engineers their full say: assistants that learned to agree from the ratings people gave them, confident guesses rewarded by the way models are scored, and tests passed by shortcut, each one documented by the labs and researchers who found it. Then it asks what their explanations leave out: why each fault rhymes with one its makers would least like to own. Along the way come Goethe’s sorcerer’s apprentice, Mary Shelley’s creature, Jacob Grimm’s golem of 1808, which grew until its master could no longer reach its forehead, and the day in 1965 when Gershom Scholem dedicated a computer named Golem Aleph.
What we liked most is that the essay refuses both easy registers of the AI debate, the shrug and the apocalypse, and holds every laboratory it names to the same light, including the company whose models helped draft it, which it names at the first mention. And it saves its best turn for the tale itself: a second look at what the student actually did under the oak.
Read the full essay at Intellicurean →
Essays & Reviews for the Intellectually Curious
