Hydrogen, given enough time, turns into people.
These days, we’re seeing the rise of LLMs like ChatGPT and many others. I’ve been thinking a lot about how we, as humanity, reached this point — and how the future might look. What is even possible from here? Every night before I sleep, these thoughts keep my brain wide awake. What an exciting time to be alive. It honestly feels dazzling to imagine what lies ahead.
In this piece, we’ll reflect on what intelligence really is — what its properties are, what its limits might be, and what it allows us to do. How will our societies evolve alongside it?
Right now, our current society is structured in a way that slowly conditions a fresh brain. A child’s natural curiosity slowly fades as society pushes conventional thinking — study, get a job, follow the template. Creativity and exploration get buried under expectations.
But maybe we’re approaching something new. A future that is not dystopian (hopefully 🙃), but brighter.
Let’s start with a simple definition:
Intelligence = the ability to accomplish complex goals.
This is broad enough to include self-awareness, understanding, problem-solving — because those are all forms of achieving complex goals.
There are different types of intelligence:
- Narrow (like a chess AI)
- Broad (like a human being)
A healthy child, given time and the right input, can learn to play chess, speak any language, or master any domain. That’s the power of broad intelligence.
The end goal? To build an entity with maximally broad intelligence, capable of achieving any goal — including learning itself.
I keep circling back to this: AGI is hard because human-level intelligence includes many skills that happen below the level of conscious awareness. Skills like recognizing a face, moving around a room, catching a ball, judging someone’s intentions, or simply deciding what to pay attention to. These things feel effortless to us because they’ve evolved over millions of years.
Meanwhile, things like mathematics or engineering — what we usually consider “intelligent” activities — are actually recent and artificial. Ironically, these are easy for computers, while perception, movement, or emotion recognition are extremely hard.
Why? Because perception and motor skills require enormous computation. And only now have we developed systems capable of this. For instance, image recognition — something your brain does effortlessly — took decades of research and tons of compute to pull off with machines.
But now we’re moving into Life 3.0 — where intelligence becomes universal and machines begin to acquire not just technical, but social skills. Skills like how to collaborate, how to build a factory, how to teach.
At its core, intelligence is not about flesh, blood, or carbon atoms. It’s all about information and computation.
So what are information and computation, really?
Physics teaches us that at the most fundamental level, everything is just matter and energy moving around.
It’s fascinating to think: How can a bunch of dumb particles following physical laws exhibit something as complex as intelligence?
A big part of the answer lies in a concept called substrate independence — the idea that intelligence (or the mind) isn’t tied to a particular material. As long as the physical or computational platform supports it, intelligence can emerge.
That’s why memory can be stored in silicon chips or in biological DNA. The platform doesn’t matter — what matters is the structure and the pattern.
Of course, memory in the brain and computer memory are quite different. In computers, we retrieve data by specifying where it’s stored. In our brains, we recall by specifying what is stored — similar to using a search engine. This kind of memory is called auto-associative memory, and it’s one of the key pieces in making AI more human-like.
Networks of interconnected neurons form landscapes filled with “energy minima” — stable patterns that a system can settle into, producing useful behavior or thought.
Once memory is solved, the next step is computation — figuring out how to build functions that can accomplish complex goals.
So how does dumb matter compute?
It turns out, matter — when arranged in the right way — can perform complex computations. Not only is it possible, but it can be done in many different ways. This is where the idea of universal computation comes in.
Computation is substrate-independent. That doesn’t mean the substrate is useless — but it means the specific details don’t matter. You could use NAND gates, silicon transistors, neurons, or even quantum particles. As long as the system can compute, the material can vary.
Hardware is the matter.
Software is the pattern in which matter is arranged.
This is the heart of why AI is possible. Intelligence doesn’t require a human body — it requires the right pattern of matter.
We haven’t hit any physical limit yet that would prevent us from creating beings more capable than humans — across all dimensions. We just haven’t figured out the right arrangements of matter yet.
Progress in quantum computing might be key. Recently, Microsoft built a chip using Majorana particles and topological superconductors to make qubits more stable. This opens the door to much more powerful quantum computers.
Neural networks are currently our best shot at building intelligent systems. They mimic the way biological neurons work — they compute and they learn. But they might soon hit computational ceilings. That’s where quantum computing might offer the next leap.
We still don’t fully understand what “activation functions” the brain uses to achieve such insane efficiency — but we know for sure: the brain doesn’t violate the laws of physics.
If you zoom out a bit: early organisms learned to survive and reproduce slowly through Darwinian evolution — learning happened across species over generations.
Then, half a billion years ago, some organisms evolved neural networks — allowing them to learn and adapt during their lifetime.
Eventually, we — humans — came into the picture. But we may be stuck at a local maxima: our bodies are fragile, slow, disease-prone. There are physical limits we can’t surpass.
So maybe the next step is to design a new kind of being.
A life form better than us in both hardware and software. A being that can evolve its own code, and design better versions of itself.
This is the future we’re heading towards.
Once we get there, we might unlock everything: time travel, parallel universes, deep space exploration, energy from stars, cryogenic sleep, intergalactic travel.
I genuinely believe humanity needs an exceptionally brilliant companion — an entity that can help solve the universe’s deepest mysteries.
Intelligence should not be a privilege. It should be a tool — accessible to every human — to help us create a better world.