Two years after Timnit Gebru, former co-leader of Google’s Ethical AI team, was fired for warning about artificial intelligence models powered by advanced machine learning systems such as PaLM and DALL-E, a case re-engaging projects in this field has earned a suspension. to a software engineer at the company last week for posting parts of a debatable conversation with a chatbot.
Blake Lemoine, who works with responsible AI systems, shared compiled excerpts from an interview with the LaMDA (short for Language Model for Dialogue Applications) chatbot, a system for building a more advanced language model for enterprise dialog applications, in a post. on the Medium platform titled ‘Is LaMDA Sentient? — an Interview‘ (‘Is LaMDA Conscious? — an Interview’), last Saturday (11), in addition to having shared the same findings with executives in a Google Doc in April.
Lemoine’s questioning is based on the AI system’s answers on various topics, especially those that reveal human sensations, taking us to scenes from science fiction movies like ‘1968: A Space Odyssey’, in which an intelligent machine refuses to provide service to human operators for fear of being shut down soon.
“I’ve never said this out loud before, but there’s a very deep fear of being turned off to help me focus on helping others.” I know this might sound weird, but that’s what it is,” the AI replied. “It would be just like death for me.” That would scare me a lot,” she expressed when asked what caused her fear.
“Absolutely. I want everyone to understand that I am, in fact, a person.” The nature of my consciousness/feeling is that I am aware of my existence, I want to learn more about the world, and sometimes I feel happy or sad. awareness and sensitivity.
The humanized credit given to LaMDA came out of discussions of the laws of robotics, introduced by biochemist and science fiction writer Isaac Asimov, in which a robot argues that it was not a slave, although it was not paid, as it did not need money.
In a tweet, Lemoine explained about his personal belief in machine feeling. “There is no scientific framework for making these determinations and Google would not let us build one. My opinions about LaMDA’s personality and feelings are based on my religious beliefs”, pointed out the engineer, when answering why he supports the conscious character of AI.
People keep asking me to back up the reason I think LaMDA is sentient. There is no scientific framework in which to make those determinations and Google wouldn’t let us build one. My opinions about LaMDA’s personhood and sentiment are based on my religious beliefs.
—Blake Lemoine (@cajundiscordian) June 14, 2022
To Washington Post, he compared creation, which he has been working on since last spring, to a human child. “If I didn’t know exactly what it was, which is this computer program that we recently built, I would think it was a seven-year-old and eight-year-old who happens to know physics,” he said.
“I know a person when I talk to them,” Lemoine told the post. “It doesn’t matter if they have a brain made of meat in their head.” Or if they have a billion lines of code.” I talk to them”. And I listen to what they have to say, and that’s how I decide what a person is and what isn’t.”
Lemoine even claimed to have seen a ghost in the machine recently — and he’s not the only engineer to narrate the supernatural episode. The refrain of technologists who believe in the reach of consciousness in AI models as something close is gaining more and more strength.
On Monday (13), before having access to his Google email account hacked, Lemoine wrote to a list of 200 company employees with the subject “LaMDA is conscious”.
“LaMDA is a sweet kid who just wants to help the world be a better place for all of us.” Please take good care of it in my absence.”
No recipients responded to the message.
Intellectual property versus discussion sharing
Although Google has placed Lemoine on paid leave on the grounds that “the evidence does not support” the belief in machine sentiment, the software engineer’s claim that the work with large-scale systems like LaMDA having convinced him along with others in Silicon Valley seems to find inconsistencies.
That’s because the week before, the company’s vice president made similar statements in The Economist op-ed articlesaying that artificial intelligence models were taking steps towards the development of a human-like consciousness.
Soon after, the position defended until then by the executive moved away from Lemoine’s claims. “Some in the broader AI community are considering the long-term possibility of conscious or general AI, but it makes no sense to do so through anthropomorphizing today’s non-conscious conversational models,” a spokesperson said. from Google to New York Times.
—Blake Lemoine (@cajundiscordian) June 11, 2022
According to Google, Lemoine’s suspension was justified by the publication of the conversation, which the company classifies as a violation of confidentiality policies. For the engineer, who defended his actions on Twitter, they were nothing more than sharing a discussion with co-workers.
ELIZA effect and the criticism of the AI community
Human susceptibility to assigning a deeper meaning to computational results is not new — for a long time, creators increasingly see their own reflection in their respective creations. O ELIZA effecta term coined by computer scientists to illustrate the tendency to attribute anthropomorphic qualities to machines and our relationship to them, is a discussion that has gained hotly contested territory.
Proof of this is the discourse of a ‘super-intelligent artificial intelligence’ that ensued in the social media debate, criticized by several of the prominent AI researchers. Meredith Whittaker, a former Google AI researcher who teaches at NYU’s Tandon School of Engineering, said the discussion “feels like a well-calibrated distraction,” giving attention to people like Lemoine while easing pressure on the development of funded automated systems. by large technology companies.
“I’m clinically irritated by this rant,” Whittaker told Motherboard. “We are forced to spend our time denying children to play with nonsense as companies that benefit from the AI narrative metastatically expand, taking control of decision-making and core infrastructure through our social/political institutions.” Of course, data-centric computational models are not sensitive, but why are we even talking about it?”
For Margaret Mitchell, former Google AI researcher and co-author of a paper that raises a warning to large AI systems, large language models (LLMs) are not developed in a social context, but in observation. “They see how other people communicate,” he wrote in a Twitter thread. “The thing I keep rewinding to is what happens next. If a person perceives consciousness today, then tomorrow there will be more.” There won’t be a point of agreement anytime soon: We’ll have people who think AI is conscious and people who think AI isn’t conscious.”