What If Evolution Never Created Its Smartest Mind?

Let me start from the end.

Imagine there's an AI sitting on some server somewhere. It has never read a book or seen Wikipedia, and nobody taught it a single thing humans know. Still, it has figured out gravity on its own, and in a completely different way from how we did it.

Now the question is how it got there. For that we have to go back. Way back.

The apple thing

You know the Newton story. An apple fell on his head and he asked himself, why did this fall?

Apples were falling on people's heads before Newton too. The difference is he got curious about it. He kept going deep, deep, deep, deep, and at the bottom of it he found gravity.

Tesla was the same with electricity. He got curious and kept going until he could actually make it work for us.

So if you look at it fundamentally, every big thing humans ever invented came from curiosity. A human gets curious, and that brings inventions.

Now, people are already working on making AI curious, and honestly it's slowly starting to happen. AI is solving maths problems that humans had been stuck on.

So my question is: if AI becomes truly curious, will it keep discovering new things, and will it go past human intelligence at some point?

What makes a human, a human?

My simple thought is this.

A baby is born with a bare-minimum brain. It comes into this world, and god has given it eyes, hands and the ability to walk. It starts observing things. It watches someone eat three or four times and its neural network gets trained: okay, this is how you eat. It feels hungry and learns what to do about it.

That's how it learns, little by little. Nobody gives it a manual.

At a fundamental level, artificial neural networks were built the same way. So what if we make them observers too? Let them see things, get curious and ask "why did this happen?", and keep figuring stuff out on their own.

Maybe we're the dumb ones

This is the part that really got me thinking.

Right now every model we train is trained on human data. All of it is our books, our code and our knowledge.

We're assuming humans are the best species, so if we train AI on human knowledge and make it curious, it'll become superintelligent.

But who knows, maybe our species is the dumb one.

Maybe we're not the smartest brain. Maybe the one that should have been the smartest by nature's standards didn't survive. It just died. Evolution doesn't pick the smartest, it picks whoever survives.

If that's true, an AI trained only on us can only go as far as we did. Our limit becomes its limit.

The crazy experiment

So here's a small crazy experiment.

Let's say I build a virtual reality and put an AI inside it. I cut off its knowledge at the point when the first living species came on Earth. It has no human knowledge at all, just curiosity.

Then I keep giving it the world, a little bit every day. I keep giving it scenarios. I bring it in front of the dinosaurs, then the species that came after them. I keep throwing tough problems at it, everything nature put on this Earth, and I give it compute. Then I basically tell it, okay, now you figure it out from here.

Since all of this runs on GPUs, we don't need to wait thousands of years. We can just fast forward.

So I recreate every scenario that happened on Earth and run survival of the fittest on AI. A virtual evolution, basically.

Simulations inside simulations

At the most fundamental level, the AI looks at something and says, "there are n number of ways this can go from here."

So it doesn't pick one. It runs n simulations. Each of those creates more simulations, and those create more, and it keeps going and going.

Most of them die, because that's evolution. But some survive, and they keep going deeper.

The last agent

At the very end, there's one last agent left.

It's different from humans. It doesn't think the way we think or do things the way we do things. It walked its own path through all of evolution and ended up somewhere we never reached.

It's not even a species, really. It became that thing on its own, and it might be super intelligent.

Back to the server

So now you know where that AI from the start came from.

It didn't learn from us, and that's why it's different from us.

And the question I keep coming back to is this: what if humans were never the smartest thing evolution could create, just the ones who survived?

Maybe this experiment is impossible, or maybe it's years away. Maybe someone is already building it.

This is just a thought experiment, and these are my thoughts. Tell me if it makes sense or if I'm talking gibberish.