PROVIDENCE, R.I. [Brown University] — Yann LeCun is known as one of the world’s preeminent artificial intelligence researchers and pioneers. He’s also known for being blunt and not shying away from controversy. During a talk at Brown University on Wednesday, April 1, he did not disappoint on either count.

“AI sucks,” LeCun told a standing-room-only crowd in Pembroke Hall. “We have systems that can manipulate language, and they fool us into thinking they are smart because they manipulate language. But in fact, they are completely helpless when it comes to the physical world.”

Their helplessness stems from the fact that modern AI systems actually have no concept of what the world is or how any of their actions might impact it.

“The result is that everybody these days in AI is talking about agentic systems — systems that can produce actions in the world — and almost none of those systems at the moment are capable of predicting the outcome…  of their actions,” he continued. “It's a very bad way to produce an action… if you're not able to predict the consequences of it. In fact, it might be dangerous.”

In an event organized by Brown University’s Office of the Provost and the Data Science Institute, LeCun’s comments came during the latest installment of the Lemley Family Leadership Lecture Series. LeCun is a professor of computer science and data science at New York University and executive chairman of the AI startup AMI Labs. In 2018, he was awarded the Turing Prize, regarded by many as the Nobel Prize of computer science, for his development of the convolutional neural networks that power much of modern computer vision.

LeCun says he believes that the next frontier in AI is developing systems that not just understand language, but can create their own abstract models of the world around them. Those models are crucial to developing systems that can take safe and meaningful actions.

“A world model is a predictive system that, given the current state of the world… and given an action that you imagine taking, [you can] predict the next state of the world that will result from this action or intervention,” LeCun explained. “If you have such a world model that predicts what the world is going to be after you take an action, you can use that for planning.”

Creating systems with such models will require a new approach, LeCun says. Rather than training systems on mountains of text, world models will require the ability to process different kinds of noisy data from different kinds of inputs — images, video, audio, scientific data and other sources that the developers of large language models (LLMs) tend to ignore. LeCun sees the LLM approach as essentially a dead end.

“Here's another controversial statement: There's literally hundreds of billions invested in an industry that basically is counting on the fact that LLMs [are] going to reach human-level intelligence,” he said. “It's complete BS.”

Investors seem to be coming around to LeCun’s point of view. LeCun’s new company, AMI Labs, recently announced that it had raised over $1 billion to develop its world model approach.

During a question-and-answer session following his prepared remarks, LeCun addressed concerns about the impact that AI is having on education, particularly at colleges and universities. He says he’s encountered many students who think perhaps going to college isn’t worth it.

“That’s not true,” LeCun said. “The trend that we've been seeing over the last decade is that there is more demand for more advanced degrees. Certainly that's true in computer science. There's more demand in industry for people with Ph.Ds.

“The growth of the economy relies on technological innovation. Technological innovation relies on scientific breakthroughs. And those come about by people who do research. So research is becoming more and more crucial to economic expansion.”

Despite his dim view of the current state of AI, LeCun remains optimistic about its future. He sees huge potential for AI to assist in making tremendous scientific progress in areas like materials science, catalysis and other areas of basic science. Still, he says, it’s a long road to creating machine intelligence that rivals humans.

“At the very best, we might be convinced that we're on a good path towards human intelligence — but not yet at human intelligence — within five years,” he said. “But it's going to take a while, and it's almost certainly much harder than we think, because in the past 70 years… of AI, it's always been much harder than we thought.”