AI is too dumb to regulate now, according to Meta’s head of artificial intelligence research in an interview with the Financial Times. Yann LeCun, Meta’s chief AI scientist, tried to dismiss the most overblown fears of AI while offering a libertarian take against any real attempt to regulate the rapidly expanding field of artificial intelligence.
LeCun is a Turing Award winner and leading figure in computer vision and neural networks, and his work in the field is cited among other major AI researchers for why we have our current boom in artificial intelligence. The renowned researcher compared regulating AI now to what would have happened if governments hindered the early internet. He says putting a stopper on AI now would be like regulating jet airlines before they were even invented.
His comments centre around how any regulation today would be centred on “existential risk” for systems “that can even rival cat in terms of learning capabilities, which we don’t have at the moment. He called it “counterproductive” while saying AI sceptics “want regulatory capture under the guise of AI safety.”
LeCun’s opinion attacks people like Geoffrey Hinton, a fellow “godfather of AI,” who came out recently in favour of AI regulation. Hinton implied that artificial intelligence based on our current large language models would become so advanced it could harm humanity, whether through bad human actors or through some kind of self-awareness.
LeCun was entirely dismissive of Hinton’s fears of the supposed inevitable singularity, saying most folks were too influenced by movies like Terminator imagining a time when thinking machines become smarter than regular meatbag humans. Instead, he said that AI models “just do not understand how the world works. They’re not capable of planning. They’re not capable of real reasoning.”
What he’s referring to is the concept of artificial general intelligence, or AGI. Companies like OpenAI claimed that the shift to real machine intelligence is near, but LeCun called this “over-optimistic” since it requires many “conceptual breakthroughs” to reach that stage in AI. Simply put, he’s saying scientists have no real idea how to get from ChatGPT to Skynet, at least not yet.
It’s a nice change of pace to hear a prominent AI researcher try to dampen the hype surrounding AI, but then LeCun is making a strawman of most AI ethicists’ complaints. He’s not mentioning the terrible bias shown by AI art generators, language models, or even autonomous vehicles. Nor does he mention how AI is helping give rise to a new wave of online slop and disinformation.
LeCun was in charge of Meta’s Galactica AI research bot. The assistant was supposed to help researchers speed up their work, but Meta pulled the bot offline last November after scientists found it was incorrectly citing text and even making up journals. LeCun whined about the bots’ removal, saying “It’s no longer possible to have some fun by casually misusing it. Happy?”
Eventually, AI will get smarter, even more intelligent than humans, LeCun admitted. But he claimed that instead of doing harm, he imagined AI would be assisting us or tackling bigger societal challenges like curing cancer or fixing climate change. You know, the usual tech evangelist stuff.
“Intelligence has nothing to do with a desire to dominate. It’s not even true for humans,” he told FT. “If it were true that the smartest humans wanted to dominate others, then Albert Einstein and other scientists would have been both rich and powerful, and they were neither.”
That’s ignoring the worldwide celebrity status that Einstein had during his lifetime, but sure, intelligence isn’t a necessary marker for being a bastard. It’s similar to what he’s said in the past where he claimed that society has the capability “to prevent evil from having infinite power.” But that optimism won’t help those facing job loss because their company thinks it can save money by replacing staff with AI.