Slate essay: Turing's imitation game began as a drag show

The Turing test you remember from textbooks is a Q-and-A between a human and a machine to see if the machine can pass. The version Alan Turing actually proposed in 1950 had a third character, a woman, and started as a parlor game in which a man tried to convince a judge that he was the woman. Patricia Fancher's essay in Slate walks through that detail and lands somewhere I had not seen before. The original imitation game was drag.

The move from man-pretending-to-be-woman to machine-pretending-to-be-woman is what Fancher means when she writes that 'A.I. is parallel to a man performing as a woman. It was a computer in drag.' Turing, she argues, 'imagined both gender and intelligence to be fluid,' which lands differently when you remember that in 1952 he was convicted of gross indecency and given chemical castration, and that Queen Elizabeth II's posthumous pardon did not come until 2013.

The present-day hook is what current AI products have done with that legacy. Fancher's line is that 'A.I. is quickly becoming the most boring version of a woman it can be,' and she points at OpenAI's May 2024 Sky demo, the voice that, in her phrasing, 'sounded uncomfortably close to Johansson's anyway' even though Scarlett Johansson had twice declined to lend it. The compliant feminine assistant is the default now; the performance Turing actually wrote about, where the whole point was that the machine could surprise you, is not.

The honest caveat is that this is a humanities essay, not a product critique with numbers behind it. Fancher is making a reading of Turing's 1950 paper and an argument about voice assistants, not measuring anything, and the article does not get into what users say they want or how the bigger labs describe their persona choices internally. What it does well is force a question that usually gets skipped. When a company picks a voice and a personality for an assistant, the choice to default to deferential femininity is a choice, not a technical inevitability, and it inherits a much narrower idea of 'humanlike' than Turing himself was working with.

What is worth watching is whether any of the bigger labs treats persona as a place to actually experiment, or whether the Sky-style controversy pushes everyone toward the safest, most generic voice on offer.