ESHS/HSS Annual Meeting

Alan Turing's Rhetoric of "Experiment" and the Status of the Turing Test

• In Event: CTRL ALT SHIFT. Expériences, Knowledge, Activism, Methods - 1

English Abstract

The Turing test helped launch AI as a field, serving early researchers as a “definition,” a “strong criterion,” and a long-term goal. Since the 1990s, with the Loebner Prize, it has been reinterpreted as a practical experiment or “benchmark” for AI. In the 2010s, a preprogrammed chatbot was even claimed to have passed the test, despite the common-sense understanding that no such technology yet existed. More recently, with the rise of large language models, a new wave of chatbots has been engineered to play online “Turing” tests, accompanied by fresh claims of success. In this talk, I ask whether these online tests could support the claim that LLMs pass the Turing test, and how we should assess the epistemological status of the Turing test today. Has the test in fact been passed? And should we understand it as a controlled experiment or as an experience? To help answer these questions, I will examine Turing's aims and his rhetoric of "experiment" in the context of "Computing Machinery and Intelligence" (1950) and the postwar English controversy over minds and machines. Although Turing had no track record as an experimentalist, he was debating three Fellows of the Royal Society (“take nobody’s word for it”) and aligned himself with Bertrand Russell’s portrait of Galileo. I argue that his empiricist stance satisfied Russell, and effectively functioned as empiricist propaganda --- misleading some later readers and fueling the publicity ambitions of others.

Author

• BERNARDO GONCALVES, National Laboratory for Scientific Computing (LNCC)