In an interview with Lex Fridman in March 2026, NVIDIA’s CEO, Jensen Huang, said, “I think we have achieved AGI.”
Huang is the boss of one of the world’s most valuable companies, and it’s also one of the most important, as it is building the infrastructure and providing for the artificial intelligence boom. Yet, those words barely made a ripple.
AGI, or artificial general intelligence, has been the stated goal of the AI sector for a few years now. It is described as the tipping point, where AI’s ability exceeds human capacity.
Of course, some AI does already do that, but this is a broader metric, hence the phrasing general intelligence.
Not everyone believes AGI is close
So, why did Huang’s words barely make a ripple? Well, there are a few reasons. For a start, many AI experts, including those who lead rival AI companies, claimed he was wrong.
Secondly, there is a consensus that the leaders of AI companies are trying to talk up their achievements and their future influence.
That’s not a criticism, as such; it’s simply a reality of an industry that has the stock market on its back. When an AI CEO is talking on a podcast, he or she is not really talking to the host; they are talking to Wall Street.
Nevertheless, there are some out there who believe that AGI is out of reach, at least with the current iteration of AI based on LLMs.
And parts of the argument are compelling, especially when you consider that there seem to be laws of diminishing returns playing out with each new model released by companies like OpenAI and Google.
The focus now, though, seems to be more on agentic AI and less about chatbots like ChatGPT, and that’s more in the area where Huang was getting at.
The agentic web is an interesting barometer for achieving AGI, because it is more in line with the idea of constructive AI behavior.
For example, the agentic web could see agents – as exemplified by OpenClaw – out in the wild, building applications, trading on financial markets, and so on.
There is a school of thought that one day we will all have AI agents acting on our behalf, doing everything from posting on socials to playing casino games without our direct input, but that, in itself, does not equate to AGI.
AGI will be recognized ‘in hindsight’
Yet, here’s the rub. AGI is arguably something that will be recognized in hindsight. What do we mean by that? Well, it’s not a specific benchmark that is supposed to be reached. After all, how would we measure it?
Indeed, that’s one of the reasons that Huang’s statement was greeted with such skepticism; it’s not like landing on the Moon. No, AGI is something – if it is achieved – that we will look back on and think, “that was the tipping point.”
As for what happens next, it remains to be seen. There are some who talk of concepts like superintelligence, which is a step beyond AGI.
But there are also growing skeptics that AI has limitations that will never surpass human creativity and ingenuity. The lines are being drawn between the two camps, and it is getting increasingly fractious.
