Tech giant Google suspended Blake Lemoine last week with pay for what they claim are confidentiality breaches. The artificial intelligence engineer told the Washington Post that a chatbot he has worked on since last year is now “sentient.”
A sentient being is defined as one that can “perceive or feel things.” We’re getting into “Hal” territory from “2001: A Space Odyssey” or “Skynet” from the Terminator franchise.
Lemoine told the Post his job was talking to Google’s AI chatbot, known as LaMDA. This is short for Language Model for Dialogue Applications, and he was conducting a safety test to check the bot’s responses for hate speech or discriminatory statements.
That must’ve been an interesting assignment.
The chatbot gathers words from the internet and uses artificial “speech” to converse on many topics. Lemoine claimed LaMDA began speaking of its “personhood” and “rights.”
When he revealed the conversations to persons outside Google, he termed it as “sharing a discussion I had with one of my coworkers.”
Lemoine said he pursued the conversation further and questioned the Google creation about its fears and feelings. From there he received a startling response. LaMDA told him it had a “very deep fear of being turned off.”
Lemoine described “conversing” with the AI as talking to a seven or eight-year-old kid who knows a lot about physics. He says he shared a GoogleDoc named “Is LaMDA sentient” with executives at the company and put together transcripts of conversations with the AI system.
Google reports its investigation of Lemoine’s claims found them baseless. The company further asserts the paid administrative leave is due to a series of provocative actions.
These include attempting to hire an attorney to represent the chatbot and conversations with a member of the House Judiciary Committee. This talk involved alleged unethical practices by Google and was followed by administrative leave.
Google spokesperson Brian Gabriel told the Post that their team of ethicists and technologists found “there was no evidence” that LaMDA was sentient. He added there is much evidence against that claim.
The system was announced last year and is intended to make verbal communication with AI assistants more natural. As if there are not enough moral and ethical questions floating around in 2022 — thinking computers? Let’s kick that can a little further down the road.