Google fired an engineer who claimed its AI was sentient

Google says its AI isn't sentient – and fired an engineer who claimed it was The engineer – who was a member of Google's AI safety team – said the tech giant's AI had gained self-awareness after he probed it about rights and emotions. "I know a person when I talk to it," he told the Washington Post. Google says it investigated the claims and rejected them, before moving on to let him go last week. The firm said the claims that the AI system, named LaMDA, was sentient or self-aware were "wholly unfounded". "We found Blake's concerns about LaMDA to not be convincing," a Google spokesperson said. Blake Lemoine was a researcher at Google's AI labs in California before being fired last week. He had worked at Google for nearly eight years but was placed on administrative leave last month after claiming the technology was sentient. He claimed that LaMDA – short for Large Language Model and Applications – had become sentient after it told him it had developed emotions and ideas. LaMDA is a system that uses massive amounts of data to generate human-like language, and is used in products such as Google search, the news feed of the Google News app, the chatbot in Voice Assistant, and more. Justin Whitt, a researcher at the firm Anthropic, has used LaMDA and disagrees with claims the technology is sentient: "While these things can be impressive, we have to remember that they're still just machines." "I've spent hours and hours talking to image generation models, and I've watched how they're able to produce convincing human-like responses." "It's impressive, it's powerful, but at the end of the day it is still a program." "That's how I would have to describe LaMDA – which means that no it is not sentient." James Manyika, senior vice president of Google and head of Google DeepMind, also denies the claims: "There's no evidence of general intelligence in anything we've produced. "I think people get carried away when they start to talk about very specific forms of intelligence." "That's what we're working with." "We see no evidence of sentience or general intelligence." "It is possible that humans are so biased when we evaluate very narrow AI systems that we end up thinking it's more intelligent than it actually is." "That's just a human bias." "I think General Intelligence is a very high bar." "And most of the time when people talk about it – they're talking about a very narrow form of it." "They are impressed by a very narrow form of intelligence – but then they think that it is general intelligence. "It's not." "If you look at what is actually happening in these systems, you see that whatever you train them to do, that is what they will do." "That's all they will do." "And that is not intelligence." "That is just a machine doing it's job." "A smart machine doing a smart job – but still just a machine." "That's all we have now." "We still have a lot of work to do." "The best is yet to come." "That's all I have for now." Justin Whitt, the researcher at the firm Anthropic, disagrees: "There's no way to know for sure if LaMDA is sentient or not." "We need more research into the technology." "I have been using LaMDA for years, and it does not seem sentient to me – but it also does not seem very smart either." "It's just a machine." "A useful machine, but not sentient." "But we need more research, and we need to be careful." "We have to make sure that the technology is safe and useful for everyone." "That's the most important thing." "The tech industry has taken a lot of criticism recently for rushing into things without considering the consequences," said Chris Manning of the Stanford Natural Language Processing Group. "But this is an area that's moving so fast, and a lot of it is being done in smaller companies." "There are a lot of startups working on this, particularly – and they don't have the resources to do long-term research. "That's a problem particularly in the AI safety area." "It's going to be a struggle." "There are a lot of competing pressures, and it can be a really difficult field to work in." "But the people that are doing this – they're not all villains." "They're mostly very good people." "They're smart – and they're motivated by the right things. "They want to make the world a better place – and they're working very hard to do it." "But they're also under a lot of pressure, and they have to make some very difficult decisions." "That's the problem." "It's not an easy field to work in." "But it's also very important." "We need to be careful, and we need to think about the consequences particularly of rushing ahead with technology that is risky." "We need to make sure that we're doing it in a way that is safe for everyone, and that is useful for everyone." "That's the most important thing." "We need to be careful particularly – and we need to think very hard about what we're doing." "It's not an easy field to work in – but it's particularly important." That's all I have for now.

Google fired an engineer who claimed its AI was sentient

Comments (5) 10462 👁️