Geoffrey Hinton, the 'godfather of AI', quits Google and sounds warning

The Turing Award winner left the Alphabet-owned technology giant so he can speak freely about his concerns on AI

Artificial intelligence pioneer Geoffrey Hinton is worried that future iterations of AI could become a threat to humans. Reuters

Geoffrey Hinton, one of the leading voices in the field of artificial intelligence and a pioneer of deep learning, has left his role at Google and spoken of the dangers some of his creations may pose.

The Turing Award winner, who created technologies that became the underlying foundation for today's AI craze — including advancing the concept of neural networks — left the Alphabet-owned technology giant after a decade so he can speak freely about his concerns about AI, he told the New York Times in an interview.

From eliminating jobs to the threat of AI becoming sentient as it can learn on its own by self-analysing huge amounts of data, Mr Hinton, 75, voiced regrets about the innovations that he had a hand in creating.

“I console myself with the normal excuse: If I hadn’t done it, somebody else would have,” he said.

It is hard to see how bad people could be prevented from doing bad things with his creations, he said.

Mr Hinton stood by talking freely about AI in a tweet, while clarifying that he was not criticising Google, whom he said “acted very responsibly”, in the Times article.

“I left so that I could talk about the dangers of AI without considering how this impacts Google,” he said.

Google has not officially announced his departure. He told the San Francisco-based company last month that he was leaving, and last Thursday spoke to Alphabet chief executive Sundar Pichai, according to the Times.

He declined to reveal what they discussed.

Mr Hinton's decision to leave Google comes at a critical time as the company is locked in a battle for AI supremacy, particularly in generative AI, which rose to prominence thanks to the emergence of Microsoft-backed OpenAI's ChatGPT.

Generative AI is emerging as the latest battlefield for tech companies seeking advantage from the technology.

It can produce data including audio, code, images, text, simulations, 3D objects and videos. While it takes cues from existing data, it is also capable of generating new and unexpected output, according to GenerativeAI.net.

ChatGPT became a sensation because of its advanced conversational capabilities. It can even answer questions from patients “more accurately and empathically than doctors”, according to a study from the University of California published last week.

Google released Bard in February, using information from the web to offer new, high-quality responses, it said at the time.

However, days later, Bard made an error in a promotional video during a company event in Paris, wiping $100 billion off Google's market value.

In the same month, Microsoft's own Bing chatbot caused alarm after becoming threatening — including speaking of desires to steal nuclear codes, create a deadly virus or to be alive.

Even Twitter chief executive Elon Musk threw his hat into the ring, saying last month that he plans to build a “truth-seeking” AI platform.

“The idea that this stuff could actually get smarter than people — a few people believed that,” Mr Hinton told the Times.

I don’t think they should scale this up more until they have understood whether they can control it,

Geoffrey Hinton

“But most people thought it was way off. And I thought it was way off. I thought it was 30 to 50 years or even longer away. Obviously, I no longer think that.”

Mr Hinton was also concerned about the economic ramifications of advanced AI: while chatbots like ChatGPT complement humans, they could replace repetitive tasks, and could ultimately “upend” the labour sector.

“It takes away the drudge work; it might take away more than that,” he said.

Roughly 18 per cent of work globally could be automated by AI, with a bigger impact on developed than emerging markets, Goldman Sachs said in a March report.

In the US alone, about two thirds of jobs are exposed to some degree of automation by AI, with most having a significant — but partial — share of their workload (25-50 per cent) that can be replaced by AI, the study showed.

Mr Hinton is worried that future iterations of AI could become a threat to humans because of their unexpected behaviour. He also dreads the day that truly autonomous weapons — “killer robots” — may become a reality.

The emergence of weaponised tech could stem from threat actors successfully using operational technology environments to cause human casualties by 2025, research firm Gartner has previously warned.

Acknowledging the dangers of AI, Mr Hinton said the best course of action is to have a united and global effort to control the technology, and maybe even form an overarching organisation that will regulate the industry.

“I don’t think they should scale this up more until they have understood whether they can control it,” he said.

Billions of dollars have also been poured into generative AI, as investors see the potential of technology, a recent study from CB Insights showed.