Claims artificial intelligence will 'escape our control' and threaten humanity are 'ridiculous', according to one of its pioneers.

Yann LeCun is one of three 'AI godfathers' who won the 2018 Turing Award for their work on the networks.

The others – Geoffrey Hinton and Yoshua Bengio – have recently warned about the dangers.

READ MORE: AI will be 'impossible' to control and use tactics 'incomprehensible' to humans

But LeCun, who is Meta's chief AI scientist and a New York University professor, disagrees.

He dismissed the idea that we magically discover the secret to AI, turn on the machine and all of a sudden it becomes way more intelligent than us and then within minutes they will have invented all kinds of stuff that we don't understand, and it will 'escape our control'.

"This scenario is ridiculous,'' he said.

He joked: "Getting the system to run for about 20 minutes without crashing is really hard.

"Imagining that the system somehow is going to replicate itself all across the world is just preposterous,'' he added.

Hinton, Bengio and other leading tech experts have warned of the dangers if AI should fall into the wrong hands or become so advanced it decides to dispense with humans.

Execs – including SpaceX, Tesla and Twitter chief Elon Musk – called in a public letter for AI research to be paused so safeguards can be put in place.

LeCun called the letter "wrong and useless" because delaying research would be "pointless".

He suggested the decision by Sam Altman – chief executive of ChatGPT developer OpenAI – to sign the statement could be motivated by "regulatory capture".

"OpenAI is in a position where they believe they are ahead so the right thing to do is slam the door behind you,'' LeCun said.

The EU has voted to approve the AI Act which will attempt to regulate the technology when it becomes law next year.

LeCun said he was not against regulating AI but was opposed to restricting research.

His lab – which is called Fair (Fundamental AI Research) and has access to one of the fastest supercomputers – is committed to producing `open source' work meaning it will be freely-shared.

LeCun has predicted in a few years ChatGPT and other similar models 'will not be used any more' due to their limitations.

He said they do not reason or have a real understanding of the world and are less intelligent than a child or animal.

"So machine learning really sucks,'' he added.

For more shocking stories from the Daily Star, make sure you sign up to one of our newsletters
here

Source: Read Full Article