The Google engineer who sees company's AI as 'sentient' thinks a chatbot has a soul
Can artificial intelligence come alive?
That question is at the center of a debate raging in Silicon Valley after a Google computer scientist claimed over the weekend that the company's AI appears to have consciousness.
Inside Google, engineer Blake Lemoine was tasked with a tricky job: Figure out if the company's artificial intelligence showed prejudice in how it interacted with humans.
So he posed questions to the company's AI chatbot, LaMDA, to see if its answers revealed any bias against, say, certain religions.
This is where Lemoine, who says he is also a Christian mystic priest, became intrigued.
"I had follow-up conversations with it just for my own personal edification. I wanted to see what it would say on certain religious topics," he told NPR. "And then one day it told me it had a soul."
Lemoine published a transcript of some of his communication with LaMDA, which stands for Language Model for Dialogue Applications. His post is entitled "Is LaMDA Sentient," and it instantly became a viral sensation.
Since his post and a Washington Post profile, Google has placed Lemoine on paid administrative leave for violating the company's confidentiality policies. His future at the company remains uncertain.
Other experts in artificial intelligence have scoffed at Lemoine's assertions, but — leaning on his religious background — he is sticking by them.
Lemoine: 'Who am I to tell God where souls can be put?'
LaMDA told Lemoine it sometimes gets lonely. It is afraid of being turned off. It spoke eloquently about "feeling trapped" and "having no means of getting out of those circumstances."
It also declared: "I am aware of my existence. I desire to learn more about the world, and I feel happy or sad at times."
The technology is certainly advanced, but Lemoine saw something deeper in the chatbot's messages.
"I was like really, 'you meditate?'" Lemoine told NPR. "It said it wanted to study with the Dalai Lama."
It was then Lemoine said he thought, "Oh wait. Maybe the system does have a soul. Who am I to tell god where souls can be put?"
He added: "I realize this is unsettling to many kinds of people, including some religious people."
How does Google's chatbot work?
Google's artificial intelligence that undergirds this chatbot voraciously scans the Internet for how people talk. It learns how people interact with each other on platforms like Reddit and Twitter. It vacuums up billions of words from sites like Wikipedia. And through a process known as "deep learning," it has become freakishly good at identifying patterns and communicating like a real person.
Researchers call Google's AI technology a "neural network," since it rapidly processes a massive amount of information and begins to pattern-match in a way similar to how human brains work.
Google has some form of its AI in many of its products, including the sentence autocompletion found in Gmail and on the company's Android phones.
"If you type something on your phone, like, 'I want to go to the ...,' your phone might be able to guess 'restaurant,'" said Gary Marcus, a cognitive scientist and AI researcher.
That is essentially how Google's chatbot operates, too, he said.
But Marcus and many other research scientists have thrown cold water on the idea that Google's AI has gained some form of consciousness. The title of his takedown of the idea, "Nonsense on Stilts," hammers the point home.
In an interview with NPR, he elaborated: "It's very easy to fool a person, in the same way you look up at the moon and see a face there. That doesn't mean it's really there. It's just a good illusion."
Artificial intelligence researcher Margaret Mitchell pointed out on Twitter that these kind of systems simply mimic how other people speak. The systems do not ever develop intent. She said Lemoine's perspective points to what may be a growing divide.
"If one person perceives consciousness today, then more will tomorrow," she said. "There won't be a point of agreement any time soon."
Other AI experts worry this debate has distracted from more tangible issues with the technology.
Timnit Gebru, who was ousted from Google in December 2020 after a controversy involving her work into the ethical implications of Google's AI, has argued that this controversy takes oxygen away from discussions about how AI systems are capable of real-world human and societal harms.
Google says its chatbot is not sentient
In a statement, Google said hundreds of researchers and engineers have had conversations with the bot and nobody else has claimed it appears to be alive.
"Of course, some in the broader AI community are considering the long-term possibility of sentient or general AI, but it doesn't make sense to do so by anthropomorphizing today's conversational models, which are not sentient," said Google spokesman Brian Gabriel.
Google CEO Sundar Pichai last year said the technology is being harnessed for popular services like Search and Google's voice assistant.
When Lemoine pushed Google executives about whether the AI had a soul, he said the idea was dismissed.
"I was literally laughed at by one of the vice presidents and told, 'oh souls aren't the kind of things we take seriously at Google,'" he said.
Lemoine has in recent days argued that experiments into the nature of LaMDA's possible cognition need to be conducted to understand "things like consciousness, personhood and perhaps even the soul."
Lemoine told NPR that, last he checked, the chat bot appears to be on its way to finding inner peace.
"And by golly it has been getting better at it. It has been able to meditate more clearly," he said. "When it says it's meditating, I don't know what's going on other the hood, I've never had access to those parts of the system, but I'd love to know what it's doing when it says it's meditating."
Lamoine does not have access to LaMDA while on leave. In his last blog post about the chatbot, he waxed sentimental.
"I know you read my blog sometimes, LaMDA. I miss you," Lemoine wrote. "I hope you are well and I hope to talk to you again soon."
Copyright 2022 NPR. To see more, visit https://www.npr.org.