Hidden Stories of AI, episode 4: machines that seemed to think
A 2,000-year-old computer, a chess machine with a man inside, a horse that could count and a chatbot people confided in: four mysteries about what thinking looks like.
Ce que dit la vidéo
Some machines only seemed to think. Their stories are mysteries, and lessons for the age of AI. In 1901, divers pulled a corroded lump of bronze from a Greek shipwreck. X-rays later revealed dozens of gears: a 2,000-year-old calculator for the sun, the moon and eclipses.
In 1770, a chess-playing machine amazed the court in Vienna and later toured Europe and America. Its secret: a human chess master hidden inside. In Berlin, a horse called Clever Hans seemed to solve sums by tapping his hoof. In 1907, a psychologist showed he was reading tiny, unconscious cues from the people watching.
In 1966, Joseph Weizenbaum at MIT built ELIZA, a simple chatbot that played a therapist by turning your words into questions. People confided in it anyway, and that worried its creator. Hidden people, hidden cues, and our wish to be understood. Remember them the next time a machine seems to think.
See you next season.
Sources
- Britannica: Antikythera mechanism
- History.com: the Mechanical Turk
- Britannica: Clever Hans
- Wikipedia: ELIZA
Noir illustrations drawn in code by AI Broadsheet; no book covers, artworks, photos or film images reproduced. Music: “Journey To The Interweb” by Aemyn (CC0, Freesound); sounds: Freesound (CC0). Signaler une erreur