The Universe Before Time: Can AI Discover Physics Outside of Space?
Space-time is not fundamental, and we now found a way to explore reality beyond space-time | Mind Beyond Flesh-2
Disclaimer: I'm not a physicist — just someone fascinated by the deep ideas at the edge of science. This post reflects my understanding of complex topics that professionals spend years studying, so I may have misunderstood or oversimplified some details.
This is the second post in Mind Beyond Flesh, a thought-provoking series that explores the nature of reasoning, intelligence, consciousness, and artificial intelligence, not through the practical MedEd tips you're used to seeing on this blog, but through deeper, more challenging questions. If you subscribed to this newsletter just for MedEd content, feel free to skip this post.
A Mystery at the Heart of Physics
Modern physics is full of wonders, but also some big complexities. It gives us predictions that are shockingly accurate, yet behind the scenes, the math can be surprisingly messy. For example, when scientists try to calculate what happens when tiny particles crash into each other at high energies, like in the giant machines such as large hadron collider at CERN, the math sometimes breaks down. It can spit out infinities or require awkward fixes that feel more like band-aids than true explanations.
To handle these collisions, physicists often use something called Feynman diagram. It is a kind of visual shorthand for figuring out what particles are doing. These diagrams works well, but they’re complicated and hard to scale up for more complex scenarios. In fact, even with powerful computers, it can sometimes take weeks to carry out all the necessary calculations. Some scientists remained skeptical, but others held out hope for a simpler, cleaner way to understand how particles interact, one that might reveal a deeper, maybe even a beautiful structure beneath the surface.
In 2013, that beauty arrived.
The Amplituhedron: A New Kind of Beauty
It’s amplituhedron. It’s not a particle. It’s not an equation. It’s a shape; a dazzling, high-dimensional geometric object that encodes how particles interact. Introduced by Nima Arkani-Hamed and Jaroslav Trnka, the amplituhedron was a radical shift. It suggested that maybe particles don’t interact in space and time. Maybe space and time emerge from deeper, timeless structures.
In this view, calculating a particle collision becomes an exercise in geometry: you measure the volume of the amplituhedron, and voilà: you get the amplitude, the number that tells you what’s likely to happen. No need for weeks of hard calculations.
But there was a catch.
The amplituhedron only worked for an idealized version of the universe: a perfectly symmetrical, supersymmetric realm that doesn’t quite match the messy, broken-symmetry world we actually live in. Real-world physics, like the strong nuclear force that binds atomic nuclei, refused to fit.
We had a beautiful map, but only for a fantasy landscape.
The New Paper: A Map for the Real World
In April 2025, a new paper quietly solved this conundrum. A team including Nima Arkani-Hamed and his colleagues introduced something new: surface kinematics.
Instead of focusing on volumes of abstract shapes, this method draws curves on surfaces, like tracing paths on a punctured disk. Each curve encodes information about how particles interact. When you look at enough of them together, a complete picture of the interaction emerges.
Most importantly, this method works for non-supersymmetric theories, like the ones that describe the real, observable universe. It avoids the infamous mathematical problems (like divide-by-zero errors), preserves the crucial property of gauge invariance (which keeps physics consistent), and can be built up step-by-step using recursion.
In other words, it’s a clean, scalable, and consistent framework for computing particle behavior, in any number of dimensions.
It’s a new map. And this time, it’s for the world we actually live in.
Why This Matters for AI
Now comes the unexpected twist: this new framework might be perfect for AI.
AI struggles with traditional physics math. Equations filled with edge cases, infinities, and symbolic headaches are tough for even the best machine learning systems to grasp. But geometry? That’s something machines are really good at.
Curves on surfaces can be encoded as graphs, networks, and shapes.
Recursive structures are ideal for algorithmic reasoning.
The rules are clean, local, and systematic, no manual exceptions needed.
If you were designing physics from scratch to be AI-readable, you might invent something like surface kinematics. It’s structured, visual, and modular. It has a logic that could be learned.
With it, AI could:
Simulate particle collisions more efficiently
Automate parts of theoretical physics
Discover patterns humans miss
Help design new experiments and even new theories
This is teaching machines the language of nature.
What Could This Lead To?
It’s tempting to stop here — physics gets easier, AI helps out, everyone wins. But the truth is more profound.
If AI learns physics through geometric tools like surface kinematics, it’s not just learning about our universe. It’s learning from a framework that doesn’t begin with space and time at all.
Both the amplituhedron and surface kinematics share a radical implication: space-time is not fundamental. These geometric structures don’t live in space. They exist in an abstract, timeless realm, and what we call space and time may be what emerges when these structures play out.
So when we train AI on surface kinematics, we are training it not on particles or fields, but on something more elemental. We are, in a way, handing AI the blueprint of a universe that doesn’t yet have space or time. And asking it: what else is possible?
Could AI discover new physics — not just beyond the Standard Model, but beyond the structure of reality as we know it?
A New Map, A New Guide
The story of modern physics has always been about finding better maps: clearer ways to describe the universe. From Newton’s equations to Feynman diagrams, each leap has brought us closer to truth.
With surface kinematics, we may finally be taking a step toward a map that captures a new terrain, a geometry hinting at an unknown layer of reality.
And perhaps, with AI as our companion, we’re preparing for the next journey: not just across space, but beyond it.
Yavuz Selim Kıyak, MD, PhD (aka MedEdFlamingo)
Follow the flamingo on X (Twitter) at @MedEdFlamingo for daily content.
Subscribe to the flamingo’s YouTube channel.
LinkedIn is another option to follow.
Who is the flamingo?
