In a recent interview, emeritus Professor of New York University, Gary Marcus, talked about the trajectory of artificial intelligence (AI) development. A loud critic of the hype, Marcus cut a frank figure with the assertion that true AI really is something in the far-off despite recent progress.
“We’re still basically working with the same technology,” said Marcus bluntly. “It’s still designed in the same way. It takes a lot of statistics from words and tries to predict what happens next.” He highlighted the persistent issue of hallucinations, or AI systems generating nonsensical output, adding: “Nobody has solved hallucinations, and I don’t think we can with large language models.”
Marcus expressed concerns about the massive investments pouring into AI, questioning their prudence.
“People are putting in too much money too fast. I think it’s premature,” he asserted, citing examples like Sam Altman’s proposal to raise $7 trillion for AI development. “That’s like Series H money or something like that when you establish your business model, and that hasn’t happened here.”
Addressing the risks, Marcus warned: “On the financial side, the stuff is not really going to sustain itself because it doesn’t work very well.” He also cautioned about the potential misuse of AI for misinformation and scams, exploiting the systems’ vulnerability to inaccuracy.
While acknowledging the excitement around AI’s promise, Marcus remained skeptical. “I do think eventually we will solve AI but I think we’re a decade away,” he concluded, urging patience and pragmatism in the face of unrealistic expectations.
Featured image: 4 November 2022; Gary Marcus, AI Scientist, NYU on Centre stage during day three of Web Summit 2022 at the Altice Arena in Lisbon, Portugal. Credit: Piaras Ó Mídheach/Web Summit via Sportsfile, Wikipedia
