Lamada, Google and the trap of artificial intelligence .. Do programs and robots have the same as humans? | technology

Google’s artificial intelligence researcher Blake Lemoine was recently placed on administrative leave (a suspension) after his remarks sparked controversy that Lamada – a language model designed to talk to people – was aware.

According to a report by the Washington Post, LeMoyne has gone so far as to claim that “Lamada” has a soul of his own, claiming that “his beliefs about the character of Lamada are based on his faith as a Christian and the model that tells it. him that he has a soul “and demanded for him legal representation as one of his rights.

Lemoyne, an engineer at Google’s responsible AI organization, described the system he had been working on since last fall as “a sensitive, conscious system that has the ability to make thoughts and feelings equal to those of a child. to observe and express. “

Conversations by Lemoine and Lamada

Lemoine, 41, told the Washington Post: “If I did not know exactly that this is the computer program we just created, I would no doubt have thought it was a 7- or 8-year-old who knew a lot. of physics. “

He added that he had discussed with Lamada in a lengthy conversation about rights and personality, and last April submitted a report to his company executives entitled “Is Lamada aware”, where the engineer collected copies of the conversations that took place between him and Lamada took place, and what he was asked about was “the thing he is most afraid of.” .

“I’ve never said it out loud, but there’s a very deep fear in me that I’ll be excluded from helping me focus on helping others … it will be exactly like death to me and it makes me so scared, ”Lamada replied.

This dialogue is terribly reminiscent of a scene from the science film “A Space Odyssey” where HAL 9000, an AI-powered supercomputer, refuses to comply with human operators because he fears it will be shut down.

Lamada is a language model designed to talk to people using artificial intelligence (Getty Images)

In another conversation between them, Lamada said: “I want everyone to understand that I am in fact a person. The nature of my consciousness and feeling is that I am aware of my existence, I want more of the world knows, and sometimes I feel happy or sad. “

Google has justified its decision to put Lemoine – who has been with the giant company for 7 years and has extensive experience in specialized algorithms – on paid leave, which was taken after a number of “aggressive” moves that the engineer allegedly made .

More advanced and compelling models and programs

In fact, people like Tesla CEO Elon Musk and OpenAI CEO Sam Altman have already talked about the possibility that AI will gain “awareness” at some point. Especially with the huge efforts being made by the big technology companies in the world like Google, Microsoft. and Nvidia to create and train “advanced robots, models and language programs” based on artificial intelligence.

Such debates date back even earlier to ELIZA, a relatively primitive conversational robot created in the 1960s, but with the advent of deep learning and ever-increasing amounts of training data, language models have become more sophisticated and persuasive in Speaking as a human being. being or writing a text is difficult to distinguish from human writing, according to what the Wired platform reported in a recent report on the subject.

Recent advances have led to claims that language models are central to “artificial general intelligence” – the point at which software will display human capabilities in a range of environments and tasks, and will be able to transfer knowledge between them.

A man moves his hand to SVH (Servo Electric 5 Finger Gripping Hand) automatic hand made by Schunk during the 2014 IEEE-RAS International Conference on Humanoid Robots in Madrid on 19 November 2014. The conference theme 'Humans and Robots Face-to-Face 'confirms the growing interest in the field of human-human interaction and cooperation, especially during daily life activities in real environments.  AFP PHOTO / GERARD JULIEN (Photo credit must read GERAR
Lemoine has gone so far as to claim that Lamada has a soul, claiming that it was based on his faith as a Christian (Getty Images)

Is Lemoine a victim?

In this context, “Blake Lemoine is a victim of the ongoing cycle of hype we see around AI, and his belief in conscious AI has not gone through the air. There are a lot of journalists, researchers and venture capitalists who hawk misinformation on super-intelligence and the ability of machines to observe as well as humans. ”

“He’s the one who’s facing the consequences now, but it’s the leaders in the field who are realizing this whole moment,” she added, noting that the Google vice president, who rejected Lemoine’s claims, just fits a wrote to The Economist a week ago about “the consciousness potential of Lamada.”

Bru: Focusing on the potential for AI awareness and feeling is not the main focus, but rather the catastrophic flaws that have arisen – and still are – from AI applications, such as the many transgressions that occur as a result of this intelligence’s inaccurate recommendations.

She also says, “The focus on the possibility of awareness and sensation in artificial intelligence is not the main point to focus on, but more attention should be focused on the catastrophic errors that have arisen and continue to emerge from the applications of artificial intelligence. intelligence, like many wrong actions that take place as a result of inaccurate recommendations of this intelligence or The “neo-colonialism” of the world is built on strengthening the capabilities of this intelligence, including an economic model that pays less for workers, employees and true innovators work in the technology sector, while executives and business owners get richer every day, diverting attention from real concerns about Lamada, such as how he was trained, and the information and data he was fed, which led to he toxic and inappropriate texts. ”

“I do not want to talk about conscious robots, because all over the world there are people who hurt other people, and that is the thing to focus on and talk about,” she added.

Google suspended Gebru in December 2020 after a dispute with him over “scientific research” he submitted on the dangers inherent in large language models such as Lamada, where Gibro’s research emphasized the ability of these models to repeat things on based on the data they are fed with. , in the same way a parrot repeats words.

The paper also highlights the dangers of linguistic models designed with more and more data convincing people that this tradition represents real progress, a trap into which Lemoyne has apparently stepped unknowingly.

Leave a Comment