intelligence is cheap, finding it is not

When we look at modern AI, it is tempting to assume that intelligence requires giant GPU data centers, megawatts of power, and planetary-scale infrastructure.

That conclusion is wrong.

Those massive data centers are not a requirement of intelligence itself. They are merely a symptom of our current, immature algorithms. The human brain proves something profound: general intelligence can run on about 20 watts.

There are no exotic physics or violations of natural law involved. The brain runs on chemistry, electricity, noise, fragility, and slow biological parts. It is not magic. It obeys physics and relies on messy electrochemical signaling. It is noisy, failure-prone, and easily damaged.

And yet, it produces intelligence.

This fact alone tells us that the algorithms for intelligence are physically possible. If evolution could discover them accidentally, then they are discoverable by us.

the catch: evolution already paid the bill

The reason biological intelligence feels impossible to replicate is not because the brain is mysterious. It is because we are ignoring the massive compute cost that produced it.

Evolution ran the largest optimization process in history:

• Around 500 million years of complex nervous systems
• Trillions of organisms alive in parallel
• Blind trial and error
• Optimizing for survival, not intelligence
• Wasting more than 99.99% of all attempts

The algorithm in the human brain is not elegant because evolution was smart. It exists because evolution could afford to be dumb endlessly.

comparing the costs

Consider modern AI. Training a frontier model today costs on the order of 10²⁵ to 10²⁶ FLOPs. That sounds enormous until you compare it to biology.

Even if a single evolutionary evaluation equals many FLOPs, evolution likely used millions to billions of times more total optimization effort to stumble onto human-level cognition. It was spread over geological time, but it was still real compute.

The human brain did not emerge because evolution understood intelligence. It emerged because evolution had the luxury of failing forever. Humans do not have that luxury.

the real question

This completely reframes the debate around Artificial General Intelligence. It is not really about whether humans can build intelligence at all. The harder question is whether humans can compress billions of years of blind, massively parallel search into a few decades of directed, constrained search.

That is the real difficulty.

It also explains why today's AI feels so brute-force, inefficient, and power-hungry. We are trying to compensate for what evolution did slowly by doing it loudly.

the quiet implication

If we succeed, intelligence will become cheap again. It might eventually become even cheaper than biology.

However, the path to that future does not run through bigger data centers forever. It runs through better algorithms, better inductive biases, and systems that learn how to learn.

The brain shows that intelligence can be cheap. Evolution shows that finding it is not. That tension is where the future of AI really lives.

author's note: I'm just a student who is curious about intelligence and AI, not an expert, not a researcher, and definitely not someone with definitive answers. This is an attempt to think clearly about hard questions, not to claim authority over them.