Artificial intelligence has become a football league. There are divisions in which titles are pursued such as having the most powerful AIthe fastest, the most capable at a specific task either the most versatilebut the goal is to become champion of the Champions League, and that is the AGI. Although from the United States they do not stop talk about artificial general intelligence As something that is about to fall, it is a theory that suggests that, at some point, an AI will be achieved that will surpass humans in all areas of knowledge. For the NVIDIA CEO, that moment has already arrived.
At the moment it’s… smoke.
The AGI has already arrived. It has been in the podcast by Lex Fridman in which Jensen Huangboss of the company that is supporting the AI industry, has pointed out that general artificial intelligence is already here. Fridman posed a question: Could an AI system establish, grow, and operate a $1 billion-plus technology company within five to twenty years?” Huang rephrased the question, stating that current AI systems are already capable of creating a viral web service that briefly generates $1 billion in revenue.
That remains to be seen, but it is not the only one that Huang threw into the air. When Fridman asked if AGI is five, ten, fifteen or twenty years away, Huang responded with a resounding “I think it’s now. I think we’ve achieved AGI.”
Well no. Friedman said that was going to raise some eyebrows, and it took Huang less time than ChatGPT to agree with you if you tell him something isn’t right to qualify his words. Much of the conversation focused on OpenClaw platform, the open source agentic AI framework that has become the pearl of the industry and where everyone wants to have something to say.
In the article of The Verge point out the example of the Tamagotchi and how Huang points out that people can use their AI agents to do all kinds of things like digital influencers or apps to feed a Tamagotchi for you. That would become an “instant success” that can generate billions of dollars, but the possibility that, right now, 100,000 of those agents will build a company like NVIDIA “is zero percent,” he noted.
What is AGI? That is, current AI and agents can create “fleeting and viral hits”, but… that is not the question. That is, one thing is agentic AI, which is something that all the major companies in the sector are already pushing and, basically, it is like a shortcut program for AI to do things for us. And a very different one is the AGI. The artificial general intelligence that Huang says we have already achieved is something very different.
While the agents drink from the same AI that is nothing more than a language model that is dedicated to putting together words that more or less make sense, but that are calculated by algorithms, percentages and probabilities As long as two words go together, AGI is a “real” artificial intelligence, one that thinks like a human. It is something that, as far as we know, is only in the theoretical framework because its technical complexity is overwhelming.
The big difference between an AI and agents and AGI is that, if the first two have language as their core and operate from it, AGI is the human brain whose core is thought. Or so they aspire to create.
Is Huang an AI? After seeing some of their recent statements and this interview, the question that arises is why there are certain profiles that are choosing to “reason” exactly the way an AI reasons. These tools They are designed to prove us right.so as not to confront us and so that we spend more time using that AI and not another.
That’s why ChatGPT or Gemini are so accommodating in their responses when we try to find a way around them. But, as Mashable points out, giving that more convenient answer is also becoming a trend among “real” intelligences.
The wolf is coming. Huang himself already pointed out in 2024 that AGI would be software capable of imitating standard human intelligence, but his examples do not seem to match the idea of what AGI is. The industry seems hell-bent on reaching AGI through language models, but as some point out, It’s a dead end. Yann LeCun is considered one of the godfathers of modern AI. Until recently, he was also the head of AI at Meta and stated that the path to AGI is not language modelsbut the world models.
They are models that will learn from the environment, will be able to imagine scenarios and operate like humans. That has nothing to do with current models that would not be possible without plundering the Internet and human-created culture. It is not at all clear what it will be the spark that will allow that general artificial intelligence will be reached, but the American AI Big Tech never tires of saying that AGI is already here. Now it is, really good.
Huang just stated it to qualify his words in the following sentence, Sam Altman of OpenAI It has been creating expectation for a long time about the AGI when they are not yet able to do a ChatGPT that doesn’t hallucinate, Zuckerberg has assembled a super team to achieve itAnthropic’s Darío Amodei believes that is at the doors and Elon Musk says that Grok 5 could achieve it. At the moment, a lot of promises, and something that creates an app to power a Tamagotchi doesn’t sound like that great revolution in technology either.
Images | NVIDIA
GIPHY App Key not set. Please check settings