Using biology as a lens for AI, and noticing how adaptation, feedback, emergence, and evolution may explain intelligence better than ordinary software thinking.
Biology textbooks often present astonishing things as if they are ordinary facts.
A cell repairs itself.
A body regulates temperature.
An immune system learns the shape of an invader.
A brain rewires after experience.
A species changes across generations because the environment keeps applying pressure.
The tone of a biology textbook can make these things feel almost normal. It explains them calmly, labels the parts, gives the mechanism a name, and moves on. But if I pause for a moment, the whole thing feels unbelievable.
Life is full of systems that are not designed in the neat way software is designed. They are grown, selected, repaired, adapted, layered, and reused. They work with noise. They tolerate failure. They change because the environment pushes back.
That may be a better way to think about AI than traditional software engineering.
In software, we are used to clear instructions. A developer writes logic. The machine follows it. If the input is this, do that. The shape of the system comes from intention.
AI feels different.
The system is trained more than programmed. It changes through examples, feedback, loss, reward, repetition, and adjustment. Its behaviour is not always obvious from the code that created it. Some abilities appear only when the system becomes large enough, trained enough, or placed in the right environment.
That sounds closer to biology than to ordinary software.
Biology understands emergence. A single neuron is not a thought. A single ant is not a colony. A single gene is not an organism. Intelligence, behaviour, and adaptation often appear at the level of the system, not the isolated part.
AI also seems to have this quality. Individual parameters are not ideas. Training examples are not understanding by themselves. But when the pieces interact at scale, something more interesting appears.
This does not mean AI is alive.
That is not the point.
The point is that the conceptual lens matters. If we treat AI only as software, we may expect too much predictability and too much direct control. If we treat it more like a biological system, we become more attentive to feedback loops, adaptation, incentives, environments, and unexpected behaviour.
A model is shaped by its training environment the way an organism is shaped by its ecological environment. What it sees, what it is rewarded for, what errors are corrected, what pressures repeat, and what capacities are allowed to develop all matter.
Biology also teaches humility.
Living systems are difficult to reduce to one clean explanation. They have layers. They have tradeoffs. Something useful in one environment can become harmful in another. A trait that looks wasteful may have hidden value. A small feedback loop can create a large change over time.
That kind of thinking feels useful for AI.
Maybe the most important question is not only, “What did we code?”
It is also, “What kind of system are we growing?”
That question changes the mood. It makes AI less like a machine waiting for instructions and more like an adaptive system being shaped by data, goals, constraints, and human feedback.
There is something fascinating and slightly uncomfortable about that.
We are building systems that learn from the world, and the world is messy. We are asking them to adapt, but adaptation always comes with surprises. We want intelligence, but intelligence may not arrive as a clean software feature. It may arrive as a set of behaviours emerging from pressure, scale, and feedback.
Biology has been dealing with that mystery for a long time.
Today, it made me think that AI may become easier to understand if I stop imagining it as only code and start seeing it as a system under selection.
Not alive.
But definitely closer to growth than to simple execution.