Is it possible for artificial intelligence to truly become sentient? This week, Audacy asked one of the people best equipped to answer that question: Nvidia CEO and founder Jensen Huang.
Per the company’s website, Nvidia sparked the “era of modern AI by powering the breakthrough AlexNet neural network,” in 2012 and has since become a leader in the space. In April, the company announced plans to “produce up to half a trillion dollars of AI infrastructure” over the next four years.
KCBS Radio caught up with Huang – who founded Nvidia in 1993 and previously worked at LSI Logic and Advanced Micro Devices – at Lawrence Berkeley National Laboratory.
“I have no idea, and the reason for that is the definition of sentient is hard for me... he said when asked when artificial intelligence could become sentient. “I don’t know the definition of sentient, but AI is going to be able to pass these intelligence tests fairly soon. You know, if you could put a test in front of an engineer, we will come up with a technology that addresses that.”
According to the Merriam-Webster dictionary, sentience is described as the capability of “sensing or feeling,” and to be “conscious of or responsive to the sensations of seeing, hearing, feeling, tasting, or smelling.” When applied to humans, the term is pretty straightforward, but things get more complicated when it is applied to technology.
“Sentient artificial intelligence is defined theoretically as self-aware machines that can act in accordance with their own thoughts, emotions and motives,” said a blog post published by IBM last year. “As of today, experts agree that AI is nowhere near complex enough to be sentient.”
It further explained that discussions about the potential sentience of machines have been going on for decades. In recent years, the discussion ramped up due to an interview in the Washington Post with former Google engineer Blake Lemoine, who told that outlet that LaMDA, Google’s AI chatbot generator, “is sentient because it started talking about rights and personhood, and was seemingly aware of its own needs and feelings,” said IBM’s post. However, it noted that Google’s ethicists have publicly denied the claims and that experts generally agree that AI is not sentient.
However, AI technology is rapidly evolving. For example, Audacy reported just this week on research that indicates Open AI’s o3 model can avoid being shut down to achieve its goals, even if directly ordered to shut down. Audacy has also reported on research that indicates AIs can replicate themselves.
KBCS Radio reported that Huang said AI is currently achieving 60 to 70% on a wide range of exams, including mathematical tests, bar exams, and medical tests. Huang is confident that 100% scores are coming soon.
“As a result of that, AI will be a very useful tool,” he said. “Artificial general intelligence, whatever that definition is, is hard to know. And whether it has human intelligence, hard to know, because we’re just so flexible and so general and we are so adaptable.”