This week, Nvidia CEO Jensen Huang spoke out and said what has long been on everyone’s mind: AGI is already with us. Not imminent. Not “in five years.” Here. Today. If that stops you in your tracks, you’re in good company. You can read the full interview here, and I encourage you to do so if you’re interested in this sort of thing. According to the piece in the India Times: “Asked if AGI has been achieved, Huang said, ‘Depending on how you choose to define it, yes, we have achieved it’.”
Now, that right there is the money quote. Because while this sounds like a big, attention-grabbing headline, let’s unpack what that might actually mean, practically speaking. What is AGI, anyway? The problem is, there’s really no one “right” definition of AGI. I mean, you can ask five different experts what AGI is, and you’ll get six different answers.
Maybe more. So, when Jensen Huang says “depending on how you choose to define it,” that’s a big, fat caveat right there. If, for example, you define AGI as “a machine capable of performing any intellectual task that a human can,” then…yeah, okay. We’re getting pretty damn close. But if you define AGI as “autonomous reasoning over a long period of time, and the ability to set its own goals,” then…not so much.
Interestingly, if you go to the website of OpenAI, one of the leading voices on AGI, and the company behind DALL-E and ChatGPT, you’ll see that they describe AGI as “a future software technology capable of performing most economically valuable work.” Note the word future. Not present.
Can it run a $1B company?
At one point, Jensen was asked whether AI could run a $1B company. His response: “In the near term, yes. In the long term, probably not.” Now, to me, that actually sounds like one of the most reasonable, nuanced positions in this entire debate. We’ve seen AI used to generate legal briefs, write code, even do scientific research.
But running a company long-term, making strategic decisions under uncertainty, and dealing with pesky humans and all their drama? That’s still firmly in the realm of humans, for now. So, yeah, AI is getting good. But let’s not get carried away here.
Why did Jensen say this?
Look, it’s impossible to know someone’s motivations for certain, but let’s not be naive here. Jensen Huang is the CEO of Nvidia. His company’s graphics cards are what power a lot of the AI research going on today. So when he speaks, that’s not just a philosophical musing, that’s a positioning statement. And let’s not forget that there’s a war going on here to win the future of AI.
Companies like Google, Facebook, Amazon, and others are investing tens of billions of dollars into AI research, data centers, hiring talent away from universities…even when those companies experience setbacks, like Meta did when it delayed the release of its new AI model, that just illustrates how difficult this is: So in that context, going out and saying “we’ve already achieved AGI” is a very deliberate, very public declaration of “we’re winning.”
So…is it true?
So here’s the thing. A part of me thinks Jensen might actually be right, in the very narrow sense that if we define AGI as “general-purpose intelligence that can achieve human-level performance in many areas,” then yeah, we’re pretty close. But there’s another part of me, maybe the more cynical part, that thinks we still haven’t achieved something essential.
Maybe it’s judgment. Maybe it’s common sense. Hell, maybe it’s even consciousness. I dunno. But we’re not all the way there yet. And maybe that’s okay. Because maybe the real headline here isn’t “AGI is already here” so much as “we’re now living in a world where you could even ask that question.”
Because once you start wondering to yourself, “Wait a minute…is this already General Intelligence?”…then you’re already living in a different world than you were a year ago.
Ultimately, I don’t think Jensen’s comment is meant to be taken as a factual statement so much as a goad, to get us all to rethink our assumptions of just how far we’ve come. Are we underestimating AI? Almost certainly. Are we overestimating it? Also almost certainly.
But the mere fact that the question of whether AGI is already here no longer seems laughably hypothetical, that’s what we ought to be paying attention to.