Climate change a bigger threat to humanity than 'science fiction' predictions on AI, UCC scientist says

'Technology has certainly developed at a considerable pace, but there's still a long way to go before we see the kind of changes that have been predicted.'

Hype around artificial intelligence’s threat to humanity is “overstated”, and people should be far more worried about climate change, according to one of the world’s leading AI experts.

Barry O’Sullivan, a professor at University College Cork’s School of Computer Science, also believes it is “foolhardy” to think AI is anywhere near being capable of always successfully completing even basic, mundane tasks.

Instead, he says much of what is being said about AI is “still in the realm of science fiction”.

Ahead of a two-day conference on AI at UCC from Tuesday, he said narratives by the likes of Geoffrey Hinton — the so-called “godfather of AI” — are too “gloomy and doomy”.

Dr Hinton, the Nobel Prize-winning computer scientist who quit Google in 2023 after developing the search engine’s AI tech, believes the very AI systems he helped create pose a serious and “scary” threat to humanity.

The renowned neural network pioneer has suggested AI “agents” could soon have artificial superintelligence (ASI) that will surpass the abilities of even the most intelligent human minds.

Dr Hinton helped develop machine-learning systems based on the human brain that led to the creation of generative AI chatbots like ChatGPT — that can generate content that does not rely on pre-programmed responses. He believes the world could have ASI in as little as five years.

“There are so many ways it could get rid of people,” he said in one interview.

He also said it was just a short matter of time before AI would “replace everybody” involved in “mundane intellectual labour” — so-called white collar jobs.

Professor O’Sullivan, who represents the European Union at the Global Partnership on Artificial Intelligence, said: “Some remarks Geoffrey makes are far too strong.

“He's moved to a very gloomy and doomy outlook in terms of AI. But to a large extent, the kind of scenarios he's looking at are still in the realm of science fiction.

“The technologies just aren't showing progress towards this sort of super intelligence, certainly not in 20 years, maybe not even 100."

Prof O’Sullivan, who also sits on the Government's AI Advisory Council, is due to give the keynote address on Tuesday at the AI For Business Leaders: Ethics, Regulation and Business Value conference.

“I'll be focusing on the ethical aspects of AI, the regulatory aspects of AI, and how important it is to be developing and deploying trustworthy AI systems,” he said.

But I'm also going to focus a little bit on the hype, because there's a hyped narrative around AI about the extent to which it's changing the world, which I think is overstated.

“I think it's important we get a realistic sense of where AI is, and how companies can actually start to engage in it.”

He will also talk about how AI will impact the world of work, although he believes much of the narrative around that has also been overstated.

“Technology has certainly developed at a considerable pace, but there's still a long way to go before we see the kind of changes that have been predicted.

“These doomsday scenarios are interesting thought experiments, but there are much more immediate concerns about the technology.

“I think, the very fact that we're talking about these very doomsday scenarios, in a sense, it's a good thing, but [it] also focuses the mind on ensuring that those doomsday scenarios don't occur.

“We shouldn't be overly focused on doomsday scenarios because of all the doomsday scenarios you need to be worried about, I would suggest climate change is a much more immediate one, and one that we should be so focused on much more and coming up with actual approaches to dealing with it.”

He said he was not "poo-pooing" what Geoffrey Hinton is saying, but he feels the other man is far more negative than is needed at this point.

Dr Hinton was asked for a comment.