Meta's Top AI Scientist Douses Hopes for Human-Level AI

Meta's LeCun on the fear of AI

Meta's leading artificial intelligence authority, Yann LeCun, reiterated skepticism this week that systems will reach advanced general intelligence soon, rebuking more bullish takes.

The prominent researcher suggested today's hyped language models might attain animal-like cognition over the coming years but lack common sense, preventing human parity.

LeCun aimed thinly veiled criticism at Nvidia's CEO, claiming human-like AI would be created within five years, a timeline he called unrealistic.

Instead, massive data gaps mean existing systems can pass professional exams but not unload dishwashers like 10-year-olds.

Caution regarding general intelligence

The text comprises a "very poor source of information" for teaching multifaceted real-world competence, LeCun added at a Meta anniversary event.

He has previously dismissed predictions of machines surpassing or threatening humanity as "preposterously ridiculous."

The tempered perspective from an icon like LeCun pours cold water on unchecked AI optimism that has recently been pervading tech and media circles.

It may also help contextualize today's models' genuine near-term potential versus longer-term limitations.

Tempered Perspectives in AI

After all, matching a pet's perceptive and responsive intelligence could still transform industries through automating routine tasks.

But companies touting creative destruction must be realistic about current constraints before claiming the finish line is within sight.

So while AI hype may sell more GPUs, as LeCun snarked, responsible voices like his are vital for grounding inflated expectations of imminent and sweeping disruption.