We live in a historical moment marked by a fascinating and challenging paradox in equal parts: the exponential growth of artificial intelligence. However, our mind, molded for millennia of evolution, resists intuitively understanding these explosive processes. This inability generates uncertainty that affects our predictions, business decisions and society as a whole. The AI is currently developed at an amazing growth rate: every year the computational power required to train leading models is multiplied by approximately 4.6.
To dimension this phenomenon, let’s visualize a chess board and place a wheat grain in the first box, two in the second, four in the third and so on, bending the amount of grains in each box. Our minds can hardly process that the amount of rice in the last box far exceeds the annual world production.
The other day in a seminar we did a similar exercise. We offered two options: to immediately receive 100,000 euros or accept the first day of the month a cent, the second two, the third four, the fourth eight and so on for a whole month. 80% chose the first and thus losing the hypothetical opportunity to win five million euros. I suppose that if it were a real example it would have been 100%. This is the rhythm to which AI is currently evolving.
Moore’s law predicted the duplication of computing power every 18 months. Today, AI widely exceeds this rate as we have seen. Recent models such as GPT-4.5, Deepseek R1, Grok 3, handle computational resources that just five years ago would have been inconceivable. And everything is due to this curve. This curve corresponds to a benchmark not yet saturated by the most advanced models. There are not many. It is worth noting that the temporary strip is very brief, specifically the last 12 months.
We offered two options: to immediately receive 100,000 euros or accept the first day of the month a cent, the second two, the third four, the fourth eight and so on for a whole month
Why does it cost us so much to understand this type of growth? The problem lies in a well -known cognitive bias in economic and behavioral psychology: the exponential growth bias (exponential Growth Bias). This psychological phenomenon causes us to intuitively interpret the exponential curves in a linear way, leading us to drastically underestimate the speed and magnitude of the changes.
Historically, illustrative examples of this bias abound. The propagation of pandemics such as COVID-19 He made it clear how, in his early stages, authorities and general population underestimated the rapid escalation of the virus.
Another classic example is The difficulty in understanding the effect of composite interest in personal finance. Initially it seems insignificant, but over time it generates extraordinary results. Both cases show that our linear intuition is not equipped to correctly anticipate exponential changes.
In AI, this cognitive limitation is clearly reflected in predictions on artificial general intelligence (AGI). For decades, experts have significantly varied their projections on when we will reach a general the year comparable to human intelligence. Interestingly, these predictions are adjusting quickly: what five years ago was estimated by the middle of the century, it could now occur before 2030. If the error rate continues, it could be 2026.
While we discuss these uncertainties, we already live an advance with advanced agents such as Manus and Operator, available to the limited public. These autonomous systems represent a qualitative change: they not only execute tasks and program, but rather reason, plan and operate with minimal human intervention, further accelerating the exponential curve in their business adoption. Looking forward, uncertainty will increase precisely due to our difficulty in internalizing the exponential.
In the coming months, we can expect even more powerful models, really multimodal and knowledgeable in its surroundingswith more sophisticated practical applications, from advanced virtual agents to robots capable of interacting naturally in human environments. This poses an urgent reflection: are we prepared, as individuals, companies and societies, to understand and manage an immediate future characterized by such radical changes? Recognizing and overcoming our cognitive limitation against exponential curves could be a crucial ability to navigate the most important technological challenge of the 21st century where the business fabric will cease to be focused on people to become an AI.
Miguel Martín Lacoma is a consulting director at Aikit
#paradox #exponential #curves #dont #understand #advance
