Jensen Huang and the end of the myth: AGI is no longer a goal

The demystification of an elusive concept

For decades, Artificial General Intelligence (AGI) has operated as the 'Holy Grail' of computing, a mythical frontier comparable to nuclear fusion or the colonization of Mars. Traditionally defined as a system capable of performing any human intellectual task with equal or superior competence, AGI was the theoretical horizon that drove the creation of labs like OpenAI or DeepMind. However, the discourse has mutated drastically. While figures like Sam Altman, CEO of OpenAI, maintain an enthusiastic stance suggesting that AGI could be reached by the end of the year, Jensen Huang, CEO of Nvidia, has adopted a position of radical pragmatism that shakes the foundations of the industry: AGI, under certain parameters, is already here, but its importance is irrelevant compared to the capacity for execution.

This shift in narrative is not accidental. Historically, AI has gone through 'winters' marked by excessive expectations. Comparing the current AGI fever with the overflowing optimism of the 1960s, when it was promised that a machine would translate languages perfectly in less than a decade, is essential to understand why Huang is trying to deflate the term. For Nvidia, the value does not lie in the 'consciousness' or 'generality' of the model, but in its tangible utility.

Beyond semantics: Nvidia's pragmatism

Huang's stance during Nvidia's recent earnings call on August 26, 2024, marks a turning point. By calling the debate over AGI 'nonsensical,' the CEO of the world's most valuable company in AI infrastructure proposes a new frame of reference: the 'profitable token.' This metric replaces scientific abstraction with accounting reality. In a market where companies have invested billions of dollars in hardware (Blackwell and H100 GPUs), return on investment (ROI) is the only metric that will determine the survival of AI-based business models.

This vision contrasts with the Silicon Valley narrative, which often seeks technological 'singularity.' Nvidia, by supplying the nervous system of this revolution, has a privileged perspective: they don't need to sell a utopia, they need their infrastructure to be the engine of a productive economy. The transition from the scientific discovery phase to the industrial deployment stage is the most significant economic event since the mass adoption of the internet in the 90s.

The reality of 'agents' and their limits

Huang's definition of what constitutes current AGI focuses on the ability of models to reflect on their own performance, acquire new skills, and correct errors without constant human intervention. This evolution of 'LLMs' (Large Language Models) into 'autonomous agents' is where the true disruption lies. An agent capable of iterating its own logic to optimize a supply chain process provides more value than a model that simply passes a bar exam.

However, Huang maintains a healthy skepticism regarding total autonomy. He has been emphatic in pointing out that, although an AI can write code, design a logo, or draft a report, the probability that an autonomous agent will be able to conceive, execute, and scale a complex company like Nvidia is, for the moment, nil. This distinction is crucial for the future of work: AI is not replacing leadership or strategic vision; it is acting as a catalyst for operational efficiency that eliminates repetitive tasks but leaves the need for a human governance structure intact.

Why is this shift critical for the market?

Huang's statement has profound implications for the global technology ecosystem:

• The end of free experimentation: Venture capital, which has funded language models without a clear business model, is beginning to demand profitability. The era of 'burning cash' to train increasingly larger models is coming to an end if there is no direct conversion into economic value.
• Infrastructure maturity: By suggesting that we have already reached the level necessary to transform industries, Nvidia is validating its own roadmap. Investors no longer need to wait for a magical 'singularity'; current infrastructure is sufficient to reconfigure sectors like healthcare, manufacturing, and software development.
• Specialization over generalization: The market is discovering that an AI that 'knows everything' is less valuable than an AI that solves a specific problem infallibly and profitably. The trend is shifting toward specialized vertical agents that integrate proprietary data, where competition is measured not by the number of parameters, but by precision and operating cost.

In conclusion, the debate over whether we have reached AGI is, effectively, a desk-bound distraction. What we are witnessing is the deployment of an unprecedented productivity tool. Nvidia's lesson is clear: the future does not belong to those who best define intelligence, but to those who are capable of turning computation into real, scalable value.