xAI chief Elon Musk and Meta’s chief AI scientist Yann LeCun, recently engaged in friendly banter on X, and it looks like there is no end in sight. And LeCun is clearly winning the argument, or more accurately, is successfully derailing Musk’s plans to build a new supercomputer with 100,000 NVIDIA GPUs and achieve AGI next year.
Interestingly, the drama is unfolding on X rather than on Meta’s Threads, which was launched about a year ago. Meta and xAI are competitors, with both social media platforms building open-source LLMs that can provide real-time information.
Not sure if this is a deliberate move by Mark Zuckerburg or LeCun – who understands LLMs like no other – to pollute X with pessimist views, so Grok fails. “My relationship with X/Twitter is love/hate,” quipped LeCunn, further fuelling speculation.
This development comes in the backdrop of xAI announcing a Series B funding round of $6 billion to expand its team, making xAI the second-most valuable AI startup at a $24 billion valuation, outperformed only by OpenAI, which is at an $86 billion valuation.
It is also impressive to see it surpassed Anthropic in less than a year, which now stands as the third-most valuable AI startup at an $18 billion valuation.
Advises People to Not Join xAI
LeCun believes that LLMs that power generative AI products such as ChatGPT will never achieve the ability to reason and plan like humans, or achieve AGI. LeCun is of the opinion that animals are more intelligent than AI.
“General intelligence, artificial or natural, does not exist. Cats, dogs, humans and all animals have specialised intelligence,” said LeCun recently.
The banter between LeCun and Musk started after the latter invited people to join xAI’s mission.
“Join xAI if you can stand a boss who claims that what you are working on will be solved next year (no pressure),” responded LeCun, advising interested candidates against joining Musk’s company.
Further, he said that he likes Musk’s cars, rockets, solar panels, and satellite network but dislikes his vengeful politics, conspiracy theories, and hype.
LeCun believes he is politically correct because he is a “scientist, not a business or product person” unlike Musk.
When Musk questioned his contribution, as to how much research he conducted “in the last five years,” LeCun candidly replied saying, “Over 80 technical papers published since January 2022.”
A few days ago, LeCun clarified that FAIR has roughly 500 scientists and engineers, and he doesn’t run FAIR; Joelle Pineau does. “In fact, I don’t run anything, I’m nobody’s boss. I’m the Chief AI Scientist: I provide ideas and advice to teams,” he said.
xAI vs Meta
Musk recently told his investors that his goal is to have the supercomputer operational by fall 2025. Once completed, this supercomputer — comprising NVIDIA’s flagship H100 GPUs — will be at least four times larger than the biggest existing GPU clusters, including those built by Meta Platforms for training AI models, he told investors.
Currently, xAI is reportedly set to spend $10 billion on Oracle Cloud servers.
In an interview with Norway Wealth Fund CEO Nicolai Tangen on X Spaces, Musk revealed that training the Grok 2 model requires approximately 20,000 Nvidia H100 GPUs. He added that training the Grok 3 model and future versions will necessitate 100,000 Nvidia H100 GPUs.
In April, xAI introduced Grok-1.5V, a first-generation multimodal model. In addition to its strong text capabilities, Grok can process a wide variety of visual information, including documents, diagrams, charts, screenshots, and photographs.
Meanwhile, Musk’s Tesla, now headquartered in Austin, is also developing a Dojo supercomputer. NVIDIA provided 35,000 H100 GPUs to Tesla, paving the way for the breakthrough performance of FSD Version 12, their latest autonomous driving software based on Vision.
As far as Meta is concerned, LeCun recently confirmed that the company has obtained $30 billion worth of NVIDIA GPUs to train their AI models. Enough to run a small nation or even put a man on the moon in 1969.
Earlier this year, Zuckerberg announced that Meta is building massive compute infrastructure to train Llama 3 and will acquire 350,000 H100s by the end of this year, aiming for a total of almost 600,000 H100s worth of compute.
During the latest NVIDIA earnings call, the company said that the big highlight of the quarter was Meta’s announcement of Llama 3, their latest LLM, which was trained on a cluster of 24,000 H100 GPUs.
Llama 3 powers Meta AI, a new AI assistant available on Facebook, Instagram, WhatsApp, and Messenger. Using Meta AI, users can access real-time information from across the web without having to bounce between apps.
Interestingly, Meta AI Assistant is quite similar to xAI’s Grok, as both are generative AI applications on social media platforms and generate real-time information. Recently, Grok has faced criticism for generating hallucinated content on several occasions.
Meanwhile, Meta is yet to release the Llama 3 400B models, which are still in training. The company said that over the coming months, they will release multiple models with new capabilities, including multimodality, the ability to converse in multiple languages, a much longer context window, and stronger overall capabilities.
OpenAI is Likely to Reach AGI Before xAI and Meta
According to reports, Microsoft and OpenAI are also working on plans for a data centre project that could cost as much as $100 billion and include an AI supercomputer called Stargate, set to launch in 2028.
OpenAI chief Sam Altman recently proposed the concept of Universal Basic Compute, in which everyone would have access to a portion of GPT-7’s computing resources.
“I wonder if the future looks something more like Universal Basic Compute than Universal Basic Income, and everybody gets like a slice of GPT-7 compute,” said Altman, in the recent episode of the All-In Podcast.
Most recently, OpenAI announced that it had started training its next frontier model, GPT-5, and anticipates that it will bring the next level of capabilities on their path to achieve AGI.
Sadly, there is no convincing LeCun.
A few days ago, he took a dig at OpenAI, sarcastically saying: “Come work at ClosedAI. With AGI just around the corner, your shares will be worth 42 sextillionnollars. We can claw back your vested shares if you quit, unless you sign a non-disparagement agreement. Oh wait, sorry, we didn’t realise our contract was this harsh until someone published an article on it. You can keep them.”
