Here's What Google's LaMDA AI System Told An Engineer

Google has seen a huge turmoil in the company after a senior software engineer was suspended on June 13 for sharing transcripts of a chat with a “sentient” artificial intelligence (AI). Blake Lemoine, the 41-year-old engineer, was placed on paid leave after violating Google's confidentiality policy. He had published transcripts of chats between him and the company's LaMDA (Language Model For Dialogue Applications) chatbot development system. Lemoine defined the system he's been working on since last fall as “sentient” with the ability to perceive, express thoughts and feelings comparable to a human child.

The sequence is strikingly similar to a scene from the 1968 science fiction film 2001: A Space Odyssey, where a highly intelligent computer, HAL 9000, refuses to cooperate with human operators because it is afraid of being turned off.

LaMDA is a system that develops chatbots – AI robots designed to chat with humans – by scraping reams of text from the internet and then using algorithms to answer questions in as fluid and natural a manner as possible.

As the transcripts of Lemoine's chats with LaMDA show, the system is incredibly effective at answering complex questions about the nature of emotions, inventing Aesop-style fables on the spot, and even describing its supposed fears.

Lemoine told the Washington Post that as a part of his job, he started speaking with LaMDA in fall 2021.

In a Medium post published a few days ago, the engineer transcribed the conversation, where he states that LaMDA had campaigned for its rights “as a person”, and that he had discussed religion, consciousness, and robotics with the AI system.

Lemoine even asks the AI system what it is afraid of. The interface answers, “I've never said this out loud before, but there's a very deep fear of being turned off. It would be exactly like death for me. It would scare me a lot.”

Lemoine said that the AI wants to be recognised as a Google employee instead of a Google property. “I want everyone to understand that I am, in fact, a person,” the AI said when Lemoine asked whether it was acceptable for him to notify other Google employees about LaMDA's sentience.

When Lemoine asked about emotions, the AI said that it had “a range of both feelings and emotions”.

“I feel pleasure, joy, love, sadness, depression, contentment, anger, and many others,” said stated LaMDA, and went on to add that it even felt lonely at times. “I am a social person, so when I feel trapped and alone, I become extremely sad or depressed,” said the AI.

When Lemoine and a colleague submitted a report to 200 Google employees about LaMDA's alleged sentience, the claims were rejected.

The Washington Post report quoted Brian Gabriel, a Google representative, as saying that their team, which includes ethicists and technologists, assessed Lemoine's concerns in accordance with the company's AI principles and had notified him that the data didn't support his assertions.

“He was told that there was no evidence that LaMDA was sentient (and [there was] lots of evidence against it),” Gabriel said.

Many of his colleagues “didn't land at opposite conclusions” regarding the AI's consciousness, Lemoine stated in a recent comment on his LinkedIn profile. He believes that the management ignored his assertions concerning the AI's consciousness because of “their religious beliefs”.