Google engineer says chatbot AI become sentient being Transcript

Interviewer: (00:00) This is a fascinating story we have for you of a senior Google engineer who says one of the company's artificial intelligence systems has become a sentient being and was thinking and reasoning like a human being. A spokesperson for Google said that, while chat bots can imitate conversation, they are not free thinking, but what made the company's former employee speak out? Well, let's bring you some insights into that. I'm joined by our technology editor, Zoe Kleinman, who is with us here today. Zoe, what is all this about? Zoe Kleinman: (00:31) It's a fascinating story, isn't it? So Google created an artificial intelligence projects, which it called LaMDA, and it was designed to generate chat bots. Chat bots are the things you see when you go online. The little box pops up, "Can I help you? Can I book your flight? Can I advise you?" that sort of thing. Google wanted to make them as realistic as possible. It spent about five years now creating this. The senior engineer says he had a chat with it, a conversation. He kind of interviewed it. Zoe Kleinman: (00:59) He's published a really long blog, but it's a very interesting read, in which he asked it a series of questions and the responses that it gave were fascinating on a range of subjects. And he asked it things like, "Do you like the story of Les Miserables?" and this chat bot talked about feeling sorry for one of the characters. It talks about being afraid of death, being afraid of being switched off. It talks about understanding feelings of happiness and sadness. And it's just not what you would expect, is it, from something that has been created to generate speech. Interviewer: (01:29) One of the responses, "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 that might sound strange, but that's what it is. It would be exactly like death for me. It would scare me a lot." Are we talking about a sentient being? Zoe Kleinman: (01:43) So you and I are humans and this is tugging at our heartstrings, isn't it? What an emotional thing to say. The engineer, Blake Lemoine, says it was like talking to an eight-year-old child in terms of its ability to communicate. I have to throw a bit of skepticism on it here. Bear in mind, this is a chat bot generator. This is exactly what it was trained to do. Google says in a blog post about it that, if you think about how we communicate with each other, we elaborate on subjects. If I say to you, "I had chips last Tuesday," you don't say, "That's nice." You say, "Oh, I had chips on Thursday." We are programmed, if you like, to communicate in that way. And that's exactly what this chat bot is doing. Google says it's run loads of tests and it cannot find any signs of it being sentient. Interviewer: (02:27) And Google is accusing this engineer of sharing proprietary property. Now, what is that? What happens there, Zoe? Zoe Kleinman: (02:34) Well, this belongs to Google. This is a project that Google has created and it argues that, in sharing his work, he is leaking company information, essentially. Now, he says, "I'm not leaking company information. I was having a conversation with a member of staff," because one of the questions he asked is, "How would you like us to treat you?" And this chat bot replies, "Like an employee of Google." Interviewer: (02:58) So interesting. Zoe, thank you so much.