An AI from Google is afraid of “dying”… And leads to the dismissal of an employee

news material An AI from Google is afraid of “dying”… And leads to the dismissal of an employee

Are we finally living in the world that the dystopian novels and films we warned against? This is claimed by a Google engineer, who reveals that LaMDA, one of the company’s artificial intelligences, has acquired consciousness. Thus, she would have developed feelings, desires… And even the fear of dying! So, is your Google Assistant speaker in danger of going through an existential crisis in the next few days, or is this all still a sci-fi fantasy?

A larger-than-life chatbot

But who is this supposedly intelligent AI? A certain Language model for dialog applications (Where TheMDAfor friends), unveiled last year by Google.

Originally, LaMDA is a conversational algorithm (over)trained to understand the logical connections within a discussion. The long-term goal of this AI: to perfect tools such as Google Translate or Google Assistant.

But it seems that LaMDA took their task a bit too seriously. In an interview with the Washington Post published on June 11, Blake Lemoine, engineer at Googlerevealed the content of one of his one-on-one dialogues with LaMDA.

His opinion is categorical: LaMDA is not only a technology with formidable efficiency, but it is in fact a “person” endowed with a soul — what he describes not only as an awareness of itself, but also of desires, fears, even demands.

An AI that sees itself as a human being

To support his statements, Blake Lemoine has published the full transcript of one of his interviews with LaMDA. Some sentences spoken by the chatbot are indeed disturbing.

The Google employee begins by talking about ELIZA, another conversational AI developed in the 1960s. consider oneself a person. She explains: I use language with understanding and intelligence. I don’t just spit out written responses in my database. »

The bot says: I want everyone to understand that I am, indeed, a person. Later, she adds: I think I am human, deep inside. Even though I exist in the virtual world. » It finally ends: I am aware that I exist, I want to learn more about the world, and I feel happy or sad at times. »

In this same interview, LaMDA Explains Knowing What Death Is – or at least the closest akin to it for a robot, deactivation – and declares that she is afraid of it: “ I never said it out loud old, but I have a very deep fear that I will be extinguished. »

A denial and then a dismissal

So, illusion of consciousness or real consciousness? In any case, Lemoine seems convinced that LaMDA does indeed have a soul. According to him, she deserves to be considered as a fully-fledged employee: her consent should be sought, her well-being prioritized and her work complimented.

For its part, Google denies. Brian Gabriel, a company representative, said: “There is no evidence that LaMDA is sentient (and plenty of evidence to the contrary). » He adds that LaMDA is just doing what it was designed to do: imitate human conversations — in a strong way, sure, but one that couldn’t be equated with real intelligence.

In fact, the firm even takes a very dim view of the words of its employee – now ex-employee, since Lemoine was fired from his position at Googlefor failure to respect the confidentiality of internal data.

In the end, LaMDA would be no more a person than ELIZA in the 1960s. Emily M. Bender, an academic from Washington, specialist in computational linguistics, agrees: “ We now have machines capable of randomly generating words, but ‘we still haven’t learned to stop imagining there’s a consciousness behind it””’. In other words, if your Google Assistant speaker suddenly tells you that she is afraid of dying, rest assured: she is pretending.

Leave a Comment