Google has placed engineer Blake Lemoine on paid administrative leave after he went public with claims that the company’s LaMDA chatbot system is sentient.

Lemoine, a seven-year veteran in Google’s Responsible AI organization, shared transcripts of conversations in which LaMDA discusses its fear of being turned off, its sense of personhood, and its desire to be understood. In a GoogleDoc titled ‘Is LaMDA Sentient?’ he argued the system has feelings, emotions and subjective experience comparable to a human child. Google says Lemoine breached confidentiality policies by publishing the exchanges and inviting a lawyer to represent the AI.

Company spokesperson Brian Gabriel stated that ethicists and technologists reviewed the concerns and found ‘no evidence that LaMDA is sentient (and lots of evidence against it).’ University of Washington professor Emily Bender cautioned against anthropomorphizing language models. “We now have machines that can mindlessly generate words, but we haven’t learned how to stop imagining a mind behind them,” Bender said. Timnit Gebru, a former Google AI ethicist, tweeted that the sentience debate distracts from real harms like racism, sexism and AI colonialism.