The Half-Life of Knowing: When Yesterday Stops Predicting Tomorrow

Artificial intelligence has already moved from novelty to infrastructure. It has entered medicine, law, education, engineering, design, finance, leadership, programming, research, and creative work. Some people look at what is happening and see exciting possibility. Others feel something closer to disorientation. Many experience both at the same time.

That response deserves more respect than it sometimes receives. Artificial intelligence is creating more than uncertainty. It is creating volatility. In computational neuroscience and learning theory, volatility refers to something very specific: how quickly the underlying conditions of an environment are changing. The brain is continually estimating this. In a relatively stable world, yesterday is useful evidence for tomorrow. In a volatile world, the brain has to update more aggressively because yesterday’s model may already be losing predictive value. Humans actually adjust how rapidly they learn according to how stable or unstable they believe an environment to be. And that may be one of the most intriguing neurological dynamics of the AI era.

The question facing the brain is no longer simply: What is going to happen?

It is increasingly: How long can I trust what I currently know?

A skill mastered five years ago may remain enormously valuable. Another may suddenly be augmented by a machine. A familiar workflow may change in a month. A capability that once required a specialist may appear inside an application overnight. Recent labor-market research already shows rapid transformation in the task content of some occupations as generative AI changes what work humans perform and which skills retain their value.1

That creates an unusual cognitive environment. The brain builds models so it can prepare. AI keeps introducing information capable of revising those models.

New capability.

New prediction error.

New model.

Another capability.

Another prediction error.

Another update.

The exhausting part therefore extends beyond the amount of change. It is the shrinking useful life of some of our predictions. 

Predictive systems also face another problem: every error should not carry equal weight. The brain has to estimate which signals are reliable enough to justify changing its model. In predictive-processing research, this is often discussed as precision weighting. A highly reliable prediction error should influence learning more strongly than a noisy or unreliable one.

Consider what that means in the modern information environment:

One expert says AI will transform an industry. 

Another says the impact is exaggerated. 

A new model appears. 

A company restructures.

A job disappears.

Another job category emerges.

A viral headline predicts the end of an entire profession.

The nervous system is receiving a flood of possible signals and attempting to answer an extremely difficult question: Which of these deserves an update to the model of my future?

Prediction error itself is essential to learning. It tells the brain that reality has delivered new information. The cognitive burden rises when meaningful errors arrive quickly, repeatedly, and across multiple parts of life while the environment itself remains in motion.

There is, however, another variable that changes the experience dramatically: control.

Research on stress and controllability has repeatedly shown that the nervous system responds differently when an individual has meaningful influence over what happens next.2  Prefrontal systems, particularly networks involving the ventromedial prefrontal cortex, participate in detecting and learning controllability. Experiences of successful control can also produce lasting changes in how later stressors are processed.

That gives us a useful way to think about adaptation to artificial intelligence. The brain does not need a perfect map of the future. It needs agency within the future.

There is an enormous neurological difference between: I have no idea what AI is going to do to me.

and:

I am learning what this technology can do. I am experimenting with it. I am deciding where it belongs in my work. I am strengthening the capabilities that remain distinctly valuable. I know how to learn when the environment changes.

The future remains uncertain in both cases. The prediction model is entirely different. This may become one of the defining forms of resilience in the decades ahead: moving predictive confidence away from knowing exactly what will happen and toward trusting our capacity to respond when it does. That confidence is trainable.

Neuroplasticity allows experience to alter the networks supporting learning, attention, decision-making, and skilled behavior. Every successful encounter with change gives the brain new evidence. Every unfamiliar tool learned, difficult transition navigated, and prediction successfully revised becomes another data point:

I have encountered uncertainty before.

I learned.

I adapted.

I remained capable.

Over time, the prediction itself can change. The brain becomes less dependent on: I need the world to remain predictable. And increasingly capable of operating from: I know how to update when it changes.

History offers perspective here, although it should offer humility as well. Previous general-purpose technologies such as steam power and electricity transformed labor markets profoundly, often across decades, and the benefits and disruptions were distributed unevenly. Artificial intelligence may ultimately prove comparable in scale, although researchers appropriately caution that its full labor-market effects remain too early to determine.

The reassurance, then, does not come from pretending that every outcome will be good. It comes from something more powerful:  Human beings possess a nervous system built to learn from error, update internal models, develop new skills, and reorganize behavior around changing conditions.

The world ahead may become increasingly difficult to predict. Our capacity to become better predictors, and better adapters, can grow with it.

I can train you on these skills.  To inquire, send an email to: jon.law@sync7.co

---

1.  Freund, L. B., & Mann, L. F. (2026). Job transformation, specialization, and the labor market effects of AI (Working Paper).

 2. Ochmann, K. N., & Walter, H. (2025). Learned helplessness and learned controllability: From neurobiology to cognitive, emotional and behavioral neurosciences. Frontiers in Human Neuroscience, 19, Article 12417506.