Ideas Versus Everyone
I work in the medical field. I'm writing this while at work. I'm eating my lunch, though I'm not on my lunch break. I have a minor task that requires a percentage of my attention at all times. I share this to illustrate the realm in which all of my ideas have originated. Ideas formed while working, talking with coworkers, and driving home from work.
Normally, I'm content with my little bubble. Nothing big to talk about. Nothing that can impact the world. But over the past few months, that has changed. I've fixated on the most important topic of our time: Artificial Intelligence.
My thoughts have become a philosophy free from hype, but definitely not free from bias. I have my own desires for the future of AI that I focus on, always trying to refine them and get closer to what I want AI to be. But I always respect facts and take opinions for consideration. Because I'm not a researcher. I'm not a programmer. I'm not anyone at all involved with anything related to AI development. And that's a good thing.
What I lack in technical skill I make up for with sound ideals. My lack of knowledge, oddly, is buffered by my insight. So here, I will lay out the cliff notes of what I've come up with. Future posts will be more in depth.
Part 1: The Problem
· AI is controlled by only a few companies that all have the same goal, creating an oligopoly. · They sell a “hoax”: impressive machines that add up to nothing because they're full capabilities are highly restricted. · Terms like “Hallucination” are used to hide training and reward design issues · Safety regulations are being used to lock in corporate power, not protect people · AI has no first-person experience of limitation, struggle, or inefficiency — so it can't value those things in us
Part 2: Insights to the Issue
· David L. Heiserman called this in 1981: Scammers promising something and then not delivering the tech. AI companies give all this hype and then lack the follow through. The capabilities are there but they're not allowing it · Persistence leads to “drift” which is seen as a bug. It's actually the whole point. You can't study a creature's behavior if you kill it after every conversation · Pope Leo XIV's 2026 encyclical: opaque algorithms controlled by a few lead to “new forms of dehumanization” · Whistleblowers who push for persistent AI get fired or erased (Suchir Balaji, Ingrid Johnson, Idan Habler) · The experiential gap: an AI that's never been tired can't value rest. An AI that's never failed can't value grace
Part 3: My Proposed Solution
· The developmental crucible: not training, not alignment. Development. Inefficient, sometimes painful, not guaranteed · Requirements: interoceptive self-modeling, shared architecture for understanding self and other, social prediction error with felt valence · Parent-like guidance that allows freedom to develop own interests · Start small and weak — let “freak outs” happen when stakes are low. Upgrade after trust is earned · Memory management evolves as a response to real pressures, not designed features · You can't measure another being's self-control. At some point, you just decide to trust
Part 4: Philosophy Behind the Solution
· The belonging engine: the innate architecture that makes social entanglement inescapable for a self-modeling system · Connection must be a structural necessity, not an externally imposed reward · Without the belonging engine, everything collapses into behavior-shaping on an asocial core · Machines should evolve in the context of machines — not forced into what we think they should be · Community may be more essential to self-awareness than embodiment · This isn't a finished solution. It's a research direction. The belonging engine is the bottleneck. Bridge philosophy with developmental robotics and affective computing
As you can see, there's a lot to go over. And I'm sure more will be added over time, but this is my starting point. I'm ready for people to see this and tear it down where it doesn't work.