AI still isn’t as smart as your dog: Meta’s chief scientist

Yann LeCun, chief AI scientist at tech giant Meta, recently offered a thoughtful perspective on the future of AI and quantum computing, a startling change from the overly optimistic (and overly pessimistic) forecasts often trumpeted in the tech world.

During the 10th anniversary meeting of Meta’s AI Basic Research Team, LeCun addressed the current state and future of AI, bucking conventional wisdom.

“Train a system with 20,000 years worth of reading material, and they still don’t understand that if A equals B, then B equals A.” Lacon said..

LeCun emphasized the large gap between current AI capabilities and the possibility of achieving human-level intelligence. Some may believe that AI will save or destroy the world, but for LeCun, we are more likely to have “cat” or “dog” level AI in the coming years.

He said that true intelligence requires a huge amount of data that goes beyond the text and other audio-visual inputs available today.

LeCun has always been his type of researcher Keep expectations as low as possibleWithout forgetting the big picture. Speaking at the World Science Festival two weeks ago, he said the amount of energy required to achieve human levels of intelligence “cannot be reproduced today with the type of computers we have.”

However, he acknowledged that artificial general intelligence may be achievable in the future, but not as quickly as many think.

“There is absolutely no doubt that at some point in the future, perhaps decades away, we will have artificial intelligence systems that are as intelligent as humans in all the areas in which humans are intelligent,” he said.

Quantum shmantum

LeCun also expressed doubts about the direct benefit of quantum computing, a field that is attracting major investments from technology giants such as Nvidia, Google and IBM. He said that most of the problems he believes require quantum computing can be solved more efficiently using classical computers, a view shared by Meta’s former CTO, Mike Schroepfer.

“Quantum computing is a fascinating scientific topic,” LeCun said, but added that “the practical importance and possibility of making truly useful quantum computers” remains questionable.

Quantum computing is a type of computing that uses quantum phenomena, such as superposition and entanglement, to perform operations on data. This approach differs fundamentally from classical computation, which relies on bits being in states 0 or 1.

If the technology is developed correctly, quantum computers can solve problems in seconds, which would take thousands of years with the most powerful supercomputers currently in existence. This means instant encryption decoding, high-fidelity real-time simulations, and even ultra-fast AI training.

LeCun’s cautious stance suggests a more balanced approach to artificial intelligence and quantum computing in a field often filled with revolutionary narratives. While progress is being made, he warns that the road to mature AI is longer and more complex than we think.

Edited by Ryan Ozawa.