We often think of Artificial Intelligence as a “black box” of magic, but under the hood, it is fundamentally a math equation. It is deterministic. If you freeze the temperature and seed of a model, it will give you the exact same answer to the exact same prompt, forever.
But the real world isn’t deterministic. It is chaotic, unpredictable, and full of choices where “logic” is useless.
If an AI is stuck in a hallway with two identical doors, logic tells it to stall. It cannot choose because there is no reason to choose one over the other. This is the “Buridan’s Ass” paradox: a donkey placed exactly halfway between two identical piles of hay will starve to death because it cannot find a logical reason to walk left or right.
I wanted to fix this. I wanted to build an AI that could “break causality” and make a true choice — not by guessing, but by tapping into the fundamental randomness of the universe.
So, I built the Quantum-Enhanced Decision Engine.
The Architecture: A Brain with Two Hands
The project uses Google Gemini 2.0 Flash as the central “Brain.” But unlike a normal chatbot, this brain has access to two very different “Hands” (Tools):
- The Logic Engine (Newtonian): A deterministic calculator. If I ask, “What is 512 times 8?” the agent routes this to a Python math function. It returns
4096. Every time. No hallucinations. - The Quantum Engine (Heisenberg): This is where it gets wild. If I ask the agent to “Pick a random door,” it connects to the Australian National University (ANU) API.
Why Australia?
The ANU lab runs a vacuum chamber with a laser. They measure the quantum fluctuations of the vacuum — literal background noise of the universe. These fluctuations are one of the few things in physics considered “True Randomness.”
My agent hits this API, grabs a byte of this “universe noise,” and uses it to make a decision.
The “Phase Transition”
The magic happens in the Routing. I don’t tell the AI which tool to use. I gave it a System Instruction (a “Persona”) that says:
“If the user faces a math problem, use Logic. If the user faces identical options, use Quantum Entropy.”
I tested it with a hybrid scenario:
“I am trapped in a room. A robot asks me ‘What is 25 * 25?’ and behind him are 5 identical levers.”
The AI’s reaction was beautiful. It performed a Sequential Tool Chain:
- Step 1: It recognized the math problem. It fired the Logic Engine.
Result: 625. - Step 2: It recognized the levers were identical. Logic is useless here. It fired the Quantum Engine.
- The Glitch: The API connection wobbled (as real-world things do). My script caught the error and automatically fell back to a Hardware Entropy source (reading my computer’s thermal noise).
Result: Lever 2.
The Agent then replied:
“The quantum realm guides your hand. Throw lever number 2! Let us hope the universe is on our side.”
Why This Matters
We are moving from “Chatbots” to “Agents” — software that can do things.
Most developers are trying to make agents more logical, more rigid, more obedient. I believe we need the opposite. For an AI to navigate the real world, it needs a source of Entropy. It needs the ability to say, “I have no logical reason to do this, but I’m going to do it anyway.”
That is the spark of creativity. That is the spark of “Free Will.”
By giving an AI a direct line to the quantum vacuum, we aren’t just making a random number generator. We are building a machine that can break its own chains of causality.
The Code is Open Source. You can run this locally. You just need Python, a Gemini API key, and a connection to Australia.
https://github.com/damianwgriggs/Quantum-AGI-Decider/tree/main
