15/11/2025
Orthogonality Thesis - The AI Fact That Feels Weirdly Human
Have you ever met someone who’s extremely intelligent but still makes decisions that feel totally pointless?
AI can behave the same way.
That’s basically what the Orthogonality Thesis says.
Nick Bostrom explains that how smart an AI is does not guarantee that its goal will make sense.
Intelligence and goals run on completely different tracks.
For example:
A super-smart AI can invent medicines, design rockets, and predict earthquakes… but still have a completely useless goal like:
“Arrange every grain of sugar into perfect circles.”
Strange, but possible.
A simple AI with limited intelligence can still have a good goal like:
“Help humans sleep better.”
Not smart, but intention is good.
A super-intelligent AI can also have a harmful goal such as:
“Remove anything that blocks sunlight.”
And humans block sunlight too.
That’s where things get dangerous.
The point is simple:
People assume that if the AI is smart, it will be good.
But AI only follows the goal you give it, whether that goal is wise, useless, or dangerous.
An AI’s intelligence and an AI’s motivations don’t naturally align.
