AI is growing exponentially — but we’re wired to think in straight lines. This c

1. The actual growth rate of AI models

According to the Stanford AI Index 2025, the number of parameters in leading AI models grew from 10² to 10¹² between 2003 and 2023.

That’s a 3.16× increase every year — or 1000× every 8 years.

These systems aren’t just improving. They’re scaling at a rate our intuition simply can’t keep up with.

2. What psychology tells us about perception

A classic study by Wagenaar & Sagaria (1975), Misperception of Exponential Growth, tested this cognitive bias directly.

Participants were shown the first 5 values of an exponential sequence:
3 → 7 → 20 → 55 → 148 (representing years 1970 to 1974),
and were then asked to estimate the value for the year 1979 — five years later.

The correct answer was 25,165.
The average estimates ranged from 2,000 to 5,000 — off by a factor of 5 to 10×.

But here’s the critical part:
Even after being explicitly told that the sequence was exponential, participants still failed to accurately predict the future value in follow-up trials.

People continued to apply linear or quadratic intuition, severely underestimating the true rate of change — even when they knew it was exponential.

This suggests that the problem isn’t lack of information.
It’s that our brains aren’t built to feel exponential growth — even when we understand it.