Humans are the longest-running model nature has ever trained

and the training is still ongoing

Humans are the longest-running model nature has ever trained — and the training is still ongoing.

One prominent concept in AI is the concept of "value function", which modern AI systems rely heavily on — mechanisms that estimate whether an action is moving the system toward a good outcome long before the final result is known. Instead of waiting for a distant “reward,” a value function gives dense feedback, guiding the model step by step. This dramatically accelerates learning: the system isn’t stumbling blindly through long sequences; it’s constantly evaluating the direction of its trajectory. Every time the system finds a promising value function, it sort of records it and somewhat hard-codes it into the AI brain. You can hear more about these value functions in the recent podcast with Ilya Sutskever, the Israeli-Canadian computer scientist and OpenAI co-founder who was part of the board decision that temporarily removed Sam Altman from OpenAI.

A question pops up: Training today’s frontier models requires millions of GPUs and vast amounts of data. So why did we humans become the role model for AI — with all the AGI ambitions — without any of those GPUs or datasets?

I came to a realization that what’s fascinating is how closely this mirrors what evolution has done for humans. Nature has spent millions of years refining our internal value functions — survival instincts, social rewards, fear systems, curiosity loops, the satisfaction of problem-solving. These aren’t abstractions; they are genetically embedded reward mechanisms that help our brain decide whether we’re on a promising path. Each generation inherits a slightly improved reward-shaping system, based on which behaviours historically increased the chances of survival and reproduction. Those proven, working value functions are then hard-coded into our DNA, so we are born with preset, hard-coded, working value functions so we continue the "training" from there. How fascinating.

This is why humans don’t need massive GPUs or trillion-token datasets to become intelligent. Our learning process is supported by extremely powerful, evolution-built value functions that compress millions of years of trial-and-error into biological priors. We are born with tuned reward gradients — nudges that guide exploration, cooperation, caution, creativity, and persistence long before we understand why they matter. Evolution did the long, expensive optimization; culture and individual experience merely fine-tune the checkpoint.

In that sense, human intelligence isn’t just “biological hardware.” It’s the product of the longest-running training loop in history — one that shaped not only how we think, but what we care about. Our value functions are our compass. And that compass is the reason a slow-trained, low-compute biological model can outperform machines that consume far more raw data and energy.

Humans are the longest-running model nature has ever trained — and the training is still ongoing.

Written by Steven Nguyen

Author

Ai Base Network (ABN), ABN ASIA was founded by people with deep roots in academia, with work experience in the US, Holland, Hungary, Japan, South Korea, Singapore, and Vietnam. ABN Asia is where academia and technology meet opportunity. With our cutting-edge solutions and competent software development services, we're helping businesses level up and take on the global scene. Our commitment: Faster. Better. More reliable. In most cases: Cheaper as well.

Feel free to reach out to us whenever you require IT services, digital consulting, off-the-shelf software solutions, or if you'd like to send us requests for proposals (RFPs). You can contact us at [email protected]. We're ready to assist you with all your technology needs.