Nvidia CEO Jensen Huang tipped his hat to OpenAI on Sunday.
The chipmaking mogul took to X to congratulate OpenAI’s team on the release of the company’s newest and most powerful model, Astra, and made the notable statement that “AGI has arrived,” a nebulous term that typically describes AI models that match or surpass human intelligence.
He also pointed out that Astra was trained on Nvidia’s chips.
“From ChatGPT to o1 to Astra in 4 years,” Huang said in a post on X. “AGI has arrived. Congratulations @OpenAI team.”
OpenAI unveiled Astra on Thursday. The ChatGPT maker called it the world’s “most intelligent and aligned model,” and said it is capable of performing “the most demanding professional work with unmatched speed, accuracy, and judgment.”
Greg Brockman, the company’s president, said on a call with reporters on Thursday: “Welcome to the AGI era.”
AGI stands for artificial general intelligence, a fuzzy concept that nonetheless has become a core goal for leading frontier AI labs.
OpenAI defines it as “highly autonomous systems that outperform humans at most economically valuable work.” The company pitched Astra as a major research breakthrough and a shift in what’s possible to delegate to AI. It’s rolling out to customers this week.
On the Thursday call, Brockman said that in the future, he thinks people will look back and think AGI was created “about this time, and I think it might be about this model.”
“For me personally, I do think we’re there,” he said.
‘Declaring victory without a definition’
It is not the first time Huang has said AGI has arrived. In March, podcaster Lex Fridman asked him to consider a definition of AGI as an AI system capable of starting, growing, and running a tech company worth more than $1 billion.
“I think it’s now. I think we’ve achieved AGI,” Huang replied.
Other prominent tech figures have also said notable AI milestones have been reached. Elon Musk said in January that humanity had “entered the technological singularity,” which he described as a point of no return in which AI accelerates beyond human intelligence.
In May, Marc Andreessen, the cofounder of investment firm A16z, said the AGI threshold had been crossed roughly three months earlier, arguing that several frontier models were already “as smart as a person.”
Not everyone agrees.
Gary Marcus, a prominent AI researcher and critic, said Huang had jumped the gun with his comments about OpenAI achieving AGI.
“Huang gave no evidence and no definitions, which feels to me like an effort at a takeover of a scientific question by corporate fiat,” Marcus wrote on his Substack Sunday evening. “Declaring victory without a definition simply muddies the waters.”
Marcus attached his own 10-point definition of AGI, noting that Astra only meets one or two of those benchmarks.
“By conventional definitions, Astra still falls short,” he said.
The researchers behind ARC Prize, whose benchmark OpenAI highlighted in announcing Astra, also stopped short of calling the model AGI.
They described Astra’s results as a major advance but said success on the benchmark was not proof of AGI, noting that its tests take place in tightly bounded environments that do not reflect the complexity and open-endedness of the real world.
Sam Altman, for his part, told the “Sources” podcast in a recent appearance that AGI is, at best, “a very poorly defined term. I was going to say it’s like an irrelevant marketing term.”
What AGI would mean for everyday life is similarly unclear. It could lead to breakthroughs in medicine and science, or provide everyone with highly intelligent personal assistants. It could also put more jobs under pressure.
When the chips are up
Frontier AI companies like OpenAI, Meta, Anthropic, and Google have grown increasingly reliant on advanced chips manufactured by Nvidia. In a funding announcement in March, OpenAI called Nvidia “the foundation of our infrastructure.”
“Our training fleet and the majority of our inference stack continue to run on Nvidia GPUs,” the company said.
That demand has translated into enormous sales for Nvidia. The company reported $96.2 billion in quarterly revenue in August, more than double the amount reported in the same period a year earlier. Its data center business, which includes its AI chips, generated $89 billion in revenue.
And in his X post on Sunday, Huang suggested there’s plenty more computing power on the way.
“400K GPUs coming online next,” he said, referring to graphics processing units, the advanced chips used to train and run AI models.
