Meta's AI chief believes that AI super intelligence is unlikely to occur in the near future and is cautious about the potential of quantum computing.

• This week, Meta, Facebook's parent company, hosted a media event in San Francisco to commemorate the 10th anniversary of its Fundamental AI Research team.
• Yann LeCun, Meta's chief scientist, predicts that society will see "cat-level" or "dog-level" AI before achieving human-level AI.
• Unlike other tech giants such as Google and Microsoft, Meta is not heavily investing in quantum computing.

Yann LeCun, Meta's chief scientist and a deep learning pioneer, believes that current AI systems are far from achieving sentience, which would enable them to possess common sense and surpass their current capabilities of merely summarizing large amounts of text in imaginative ways.

In contrast to CEO Jensen Huang's prediction that AI will be "fairly competitive" with humans in less than five years, surpassing people at numerous mentally demanding tasks, his perspective differs.

At a recent event celebrating the 10th anniversary of Facebook's Fundamental AI Research team, LeCun stated that he knows Jensen and believes the CEO stands to benefit greatly from the current AI boom. He added, "There is an AI war, and he's providing the weapons."

As long as researchers at firms such as OpenAI continue their pursuit of AGI, they will need more of Nvidia's computer chips, according to LeCun.

Researchers believe that society will see "cat-level" or "dog-level" AI before human-level AI, according to LeCun. However, the technology industry's current emphasis on language models and text data will not be sufficient to create the advanced human-like AI systems that have been envisioned for decades.

"Training language models requires a vast amount of reading material, yet even with 20,000 years worth, they still struggle to comprehend basic equivalencies like A=B=A," LeCun explained.

LeCun stated that the fundamental aspects of the world are not comprehended by individuals due to the kind of training they receive.

LeCun and other Meta AI executives have been researching how to adapt transformer models to work with various data types, such as audio, images, and videos. By discovering the hidden correlations between these different types of data, these AI systems could potentially perform even more impressive feats.

Meta's research includes software that can help improve tennis skills while wearing the company's Project Aria augmented reality glasses. Executives demonstrated a demo where a person wearing AR glasses while playing tennis received visual cues to properly hold their tennis rackets and swing their arms. The AI models required to power this digital tennis assistant need a combination of 3D visual data, text, and audio to provide speech.

The development of multimodal AI systems represents the next frontier, but it won't come cheap. As more companies such as Meta and Google parent research more advanced AI models, Nvidia could gain an edge, particularly if no other competition emerges.

The AI hardware of the future

Nvidia's graphics processing units have been widely used for training massive language models, making it the biggest benefactor of generative AI. Meta utilized 16,000 Nvidia A100 GPUs to train its Llama AI software.

Will the tech industry require more hardware providers due to Meta and other researchers' ongoing development of advanced AI models?

"Although it's not necessary, it would be beneficial," LeCun stated, emphasizing that GPU technology remains the top choice for AI.

The future computer chips may not be referred to as GPUs, according to him.

LeCun stated that new chips, which are not graphical processing units, but rather neural deep learning accelerators, are expected to emerge.

Quantum computing has garnered the attention of tech giants such as Meta, IBM, and Google, but LeCun remains skeptical about its potential. Many researchers outside Meta believe that quantum computing machines could revolutionize data-intensive fields such as drug discovery by performing multiple calculations with quantum bits instead of conventional binary bits.

But LeCun has his doubts.

"Quantum computing can solve certain problems more efficiently than classical computers, but it is not always the case," LeCun stated.

"The practical relevance and feasibility of constructing useful quantum computers are not as clear," LeCun stated.

Mike Schroepfer, a former tech chief and current Meta senior fellow, stated that he evaluates quantum technology periodically and believes that practical quantum machines may emerge eventually, but the time horizon is so far in the future that it is irrelevant to current activities.

Schroepfer stated that the establishment of an AI lab a decade ago was due to the evident commercial potential of the technology within the upcoming years.

WATCH: Meta on the defensive amid reports of Instagram’s harm