Blake Lemoine, the Google engineer who publicly claimed that the company’s LaMDA conversational artificial intelligence is sensitive, has been fired, according to the Big Technology newsletter, which spoke to Lemoine. In June, Google placed Lemoine on paid administrative leave for violating its nondisclosure agreement after she contacted government officials about her concerns and hired a lawyer to represent LaMDA.
A statement emailed to The Verge on Friday by Google spokesman Brian Gabriel appeared to confirm the firing, saying, “We wish Blake well.” The company also says: “The MDA has gone through 11 different reviews and earlier this year we published a research paper detailing the work involved in its responsible development.” Google maintains that it “extensively” reviewed Lemoine’s claims and found them to be “totally unfounded.”
This aligns with numerous AI experts and ethicists, who have said his claims were more or less impossible given current technology. Lemoine claims that his conversations with LaMDA’s chatbot lead him to believe that it has become more than just a program and has its own thoughts and feelings, rather than producing conversation that is realistic enough to make it seem so, as it is designed to do.
He argues that Google researchers should seek consent from LaMDA before running experiments there (Lemoine himself was assigned to test whether the AI produced hate speech) and posted excerpts of those conversations on his Medium account as evidence.
The Computerphile YouTube channel has a decently accessible nine-minute explanation of how LaMDA works and how it could produce the answers that convinced Lemoine without actually being sensitive.
Here’s Google’s full statement, which also addresses Lemoine’s accusation that the company didn’t properly investigate his claims:
As we share our AI principles, we take AI development very seriously and remain committed to responsible innovation. The MDA has gone through 11 different reviews and earlier this year we published a research paper detailing the work that goes into its responsible development. If an employee shares concerns about our work, as Blake did, we review them extensively. We have found Blake’s claims that LaMDA feels to be completely unfounded and have worked to clarify this with him over many months. These discussions were part of the open culture that helps us innovate responsibly. It is therefore unfortunate that, despite long involvement in this matter, Blake has still chosen to persistently violate clear employment and data security policies that include the need to safeguard product information. We will continue our careful development of language models and wish Blake the best.