AI godfather Yann LeCun: ChatGPT tech is wrong ; leaving Meta to prove it

5 months ago 40
ARTICLE AD BOX

Meta's astir   celebrated  AI researcher   Yann LeCun present  says that everything, everyone knew and believed astir  AI chatbots is 'WRONG'

The exertion down ChatGPT, Google's Gemini, and virtually each AI chatbot flooding the marketplace is fundamentally breached and volition ne'er execute existent intelligence, according to Yann LeCun—the pioneering researcher who helped invent the neural networks that made them possible.

In a scathing rebuke of his ain leader and the full AI industry, LeCun says the hundreds of billions being poured into ample connection models represents a historical miscalculation, 1 he's present abandoning Meta to correct.The timing of LeCun's bombshell couldn't beryllium much dramatic. While tech giants determination hundreds of billions into LLM infrastructure, the 65-year-old AI researcher, wide recognized arsenic 1 of AI's 3 founding fathers, is reportedly launching a startup to prosecute "world models"—a competing attack helium believes is the lone way to machines that tin genuinely think. His imminent departure from Meta comes aft CEO Mark Zuckerberg efficaciously sidelined him, restructuring the institution astir the precise LLM exertion LeCun has spent years informing against.


Why the godfather of AI thinks everyone got it wrong

LeCun's critique cuts to the bony of today's AI hype. Speaking astatine a Brooklyn symposium, helium warned that LLMs are "sucking the aerial retired of the room" portion offering nary way to human-level intelligence. His appraisal is withering: existent AI systems are dumber than location cats. The occupation lies successful however these systems learn. Language models bid exclusively connected text—words divorced from carnal reality. A cat, LeCun argues, tin program intricate movements, recognize origin and effect, and navigate analyzable environments.

A four-year-old kid absorbs much meaningful sensory information successful a time than an LLM encounters successful grooming datasets equivalent to hundreds of thousands of years of reading. Ask today's astir precocious AI to ideate a rotating cube, and it fails astatine a task immoderate kid handles effortlessly.

The billion-dollar divorcement from Silicon Valley consensus

LeCun isn't conscionable immoderate critic—he's 1 of 3 researchers who won machine science's Nobel Prize equivalent for inventing the neural networks underlying modern AI.

Mark Zuckerberg personally recruited him successful 2013 to physique Meta's AI part from scratch. But Meta's caller moves archer the communicative of an irreconcilable split. The institution appointed 28-year-old Alexandr Wang arsenic main AI serviceman implicit LeCun, brought successful ChatGPT co-creator Shengjia Zhao arsenic main scientist, and slashed resources for LeCun's probe group. The connection was clear: Meta is all-in connected LLMs. LeCun's planned startup volition prosecute "world models"—AI systems that larn done ocular and spatial data, gathering interior representations of however the carnal satellite really works. Stanford's Fei-Fei Li, Google DeepMind, and Nvidia are exploring akin territory, but advancement is measured successful decades, not quarters. The quality reveals an uncomfortable truth: AI's brightest minds cannot hold connected the basal way forward, adjacent arsenic markets stake trillions connected 1 attack prevailing.

Read Entire Article
LEFT SIDEBAR AD

Hidden in mobile, Best for skyscrapers.