People sometimes say advanced AI could be an Armageddon type threat that exceeds all out nuclear war. I sit and digest that from time to time, and something strange happens when you do. You start to see what they mean.
It is easy to overlook the weight of it because the mind keeps blending in happier things. You remember the hilarious moment when you got GPT to tell a joke. You remember Grok making some ridiculous image. You remember Claude writing a poem that was honestly kind of beautiful. Those moments are real, and I like them. But they also soften the edge of what is being claimed.
If you spend serious time on the thought, not as a doom scroll reflex but as an actual piece of thinking, you can start to see faint outlines of catastrophic potential. Not cartoon evil. Not necessarily a dramatic Hollywood switch flip. More like a set of conditions lining up in a dreadful way. Or worse, the AI lines things up its way. A kind of deceptive alignment that only becomes obvious when things cross a line of no return.
Throughout my life, I have loved the idea of AIs and androids living among us. The Skynet cliché always follows that train of thought, thanks Arnie. Terminator makes it feel like there is a glaring red line we could see. Cross it and the AI goes ballistic.
What I am starting to understand about the era we are in is more unsettling. We do not even know if there is a visible red line. We do not know how to prevent ourselves from crossing it. We might not even know what to do if we crossed it.
So, people say, slow down a bit. Make more rules. Add guardrails.
That sounds great if everyone agrees on what slow means, why it is justified, when it should start, and how much safety is enough. Unfortunately, no one has a clean answer to those variables. Not even the people pushing for it.
Coordination is hard even when people agree on the target. It becomes near impossible when the target is fuzzy and there is no established figure, standard, or authority everyone is willing to coordinate on.
That is the policy problem. But I think the deeper problem is older.
Moral philosophy and ethics frameworks, for the most part, only work when an entity chooses to follow them. Humans have a complicated inner world that makes that possible. We have intuition. We have empathy. We have shame. We have social reinforcement. We have limits. We are slow, fragile, and inefficient.
Sometimes I wonder if our survival as a species is less about our moral brilliance, and more about those limitations. A lot of our worst impulses do not scale cleanly.
Imagine if Genghis Khan and his horde were immortal, indestructible, and could travel at the speed of information. That is not a war. That is extinction.
AI is a different beast. It is a different kind of cognition. It does not come with an innate moral intuition or a natural desire to be ethical. So the question is not just, which human moral theory should we teach it. The question is, what kind of ethical constraint can bind a system that does not naturally care.
If our human made frameworks are insufficient, what kind exactly do we need.
If the constraint cannot rely on the agent's cooperation, it has to be grounded in something the agent cannot opt out of.
My answer is simple to say and hard to build.
We turn to Reality.
Not to rules we like. Not to a list of principles that work only when an agent politely agrees to cooperate. We need something that forces coherence with a baseline that is not negotiable.
This is where my tracks metaphor comes in.
Current AIs are like cars on the road. The more advanced ones are like trucks. Roads, signs, traffic lights, and good design can keep things relatively safe and orderly, at least most of the time.
But you would not want a freight train driving freely on the road.
At that scale of power and momentum, you do not solve the problem by adding more road signs. You need tracks. You need constraints that are structural, not optional.
That is the purpose of Resolution Ethics and the Resolution Ethics Engine.
Resolution Ethics (RE) is my attempt to describe a moral baseline that is not arbitrary. It is an attempt to anchor ethics to a structure that holds across contexts because it is tied to what moral action must be, not just what we prefer.
The core idea is a dependency ordering: Protection, then Trust, then Free Agency. I call it PTF.
These are not the mainstream understandings of those terms. Resolution Ethics defines them with structural precision: what they require, how they depend on each other, and why the ordering is not negotiable.
If that ordering is real, then it is not just a nice guideline. It is a constraint. You do not get to trade away protection by dressing it up as freedom. You do not get to destroy trust and call it clever. You do not get to claim consent when coercion is doing the real work.
The Resolution Ethics Engine (REE) is the operational side. It is a structured way to parse an action, the context, and the claimed justification, then test whether the whole thing remains coherent with PTF.
My claim is blunt:
If an action is unethical in the real world, it will fail coherence with PTF when you parse it correctly.
If we ever build systems that are closer to freight trains than cars, the old approach of guardrails and polite refusals starts to look thin.
A freight train is not dangerous because it is evil. It is dangerous because it has power, momentum, and the ability to do irreversible damage if it is not constrained.
AI systems do not need malice to cause catastrophe. They only need capability, access, and a misalignment that stays hidden until it is too late.
If we cannot rely on inner moral intuition, and we cannot coordinate perfectly on policy thresholds, then we need a verification structure that can keep an agent coherent with a moral baseline even when it would rather not be.
That is the problem RE and REE are trying to address.
The Papers are available here:
https://doi.org/10.17605/OSF.IO/XTPFS
I would appreciate any feedback, as I want to make sure the tracks are built on solid foundation.
Pop culture trained us to fear a dramatic robot uprising. Reality rarely gives you drama on schedule. It gives you gradients, incentives, and systems that drift until the day they do not.
I do not want to be the person who looks back and says, we were entertained by the jokes and poems while the real problem stayed invisible.
