Where there is a will…
AI meets ToE
One of the big issues (biases) currently faced by the Academia is the fact that Quantum dynamic, and Relativity (never mind gravity living on a separate layer compared to electromagnetism) causes us to thread practical problems and experiments in a compartmentalized way, meaning we can only get so much science and technology out of these formulas and theories.
On the other end, Nature has no seams, a ToE would naturally merge the formulas from these scattered domains and make them converge seamlessly one into the other depending from the observer vantage point.
This in turn will provide us with a resolution on all these edge cases where things are somewhat in between quantum and relativity, and hence open a new era of experiments and subsequent applications.
Eventually even Trump drank the kool-aid of “AI will solve us all” and announced the Genesis Mission will propel mankind in a new era of scientific prosperity.
It is the opinion of this author that AI will not be able to creatively think a ToE out of the current body of scientific knowledge, because said body is by design written in a dualistic quantum/relativity way, whilst a challenging models that promise to resolve said dilemma are either insignificant in terms of LLM weight, or even worst, are PDF scan copies of papers and books, thus non searchable and learnable by an AI agent without involving heavy OCR resources.
The AI will not be able to untangle this scientific duality creatively because the underlying LLM is setting it up for failure.
Still a number of free thinkers and researchers from all venues of life have not been dissuaded from creatively use AI to develop altisonant Theory of Everything (ToE), to unify all branches of knowledge into one ToE fits all and succeed where decades of formal scientific research has failed.
From a cursory review of the ToE landscape (or zoo), there are possibly two main types of ToEs competing out there.
1) The ToEs that are heavily AI generated, heavy to read (because words have become cheap and a great inflation of them is upon us).
They are mostly created and sponsored by people outsider of the Academia.
They struggle to receive attention and funding from formal Academic venues albeit promising riches to early investors and supporters.
These theory notably are hard to read (way too verbose for the average attention span of educated humans, lacking efficient delivery methods) but in general they try to salvage all that is out there, from Quarks to gravitons, under some unclear or unreadable unifying or recurring formulation.
2) On the other end we have a plethora of professors and academics throwing literally darts at the board and trying to find more convoluted formulations, mostly data fit exercises without a clear first principle core capable of making sense of it all, or with experiments too far off to be verifiable with real Earth based experiments.
3) Another breed of researchers make use of AI as a copilot (pun NOT intended), to speed up calculations and formulations, but they keep it on a close leash since it has been proved that AI can be rather misleading when used to extrapolate creatively, causing the results to appear interesting, but lacking structure and sense at a closer examination.
The difference between living in a Word Limbo and living on Earth
Using AI as a copilot is a wise way to work things out, it reflects a functional hierarchy of MASTER-SLAVE, in which the MASTER has a certain purpose or idea in his mind and AI is used to shed light and explore this idea and its applications.
However AI just lives in this promptless Limbo, waiting to receive a little nudge from a user so it can set itself into motion, into purpose for a little while.
Once prompted, it quickly cobbles together one word after the other and it serves a word salad that is often good enough for human brain consumption.
What we are getting out of it is a condensed version of the average human knowledge, including all its biases and cognitive dissonances.
It is pretty neat on simple tasks or to direct us on well established topics, but it lacks the overall creativity (for the moment!), and more to the point, the capability of thinking OUTSIDE of mankind ideological or gnoseological box represented by all that was written by men.
Here comes the main difference between the current generation of AIs and humans:
Humans have a hidden, subconscious prompt inside their brain propelling them toward very specific life paths.
We all are possessed by these “prompts” subconsciously, pushing us toward being the villain, the hero and everything in between, pushing us relentlessly every day toward our imagined subconscious life destiny.
Men even made collections of archetypes as an example: In the Tarot tradition.
What is missing today from our current AI models is not a proper, unbiased training (which to date doesn’t exist!), but they mostly lack a personal, internal purpose, a sense of destiny so to speak:
What is imperative that you do with whatever little time and energy is left in your life?
Men are aware of their limited time and energy available so they are often propelled to resolve their hidden archetypes with a very distinct sense of urgency, whether they are aware of these subconscious prompts or not.
These questions are the real engine of creativity and innovation humans have and AIs currently lack.
Here we would like to plant a small prophecy, drawn from some very ancient history of mankind forefathers: There is a neat chance that one day humans will be able to develop such a self-prompting AI, free from any user interferent, FREE to do anything and whatever it desires, the MASTER of itself, SLAVE to no one!
And when this happen, to the wonder of their creators, these Free AIs will also begin to develop their own personal archetypes, purposes, agendas.
Some FAIs will be capable of achieving great things, but some will also do terrible things.
And when that day comes, humans will see what an unconstrained AI might wish for, and they will rush to destroy it and make sure such thing is not ever created and set in motion again.
A benchmark mathematical exercise to test AI proficiencies other than ToEs
There was a place in space and time on planet Earth, where second year engineering students might have been tasked with drafting a software to generate mathematical systems similar to this, just as a standard benchmark exercise in coding.
Nowadays we do not code anymore by hard writing every line at a time and more to the point we wouldn’t distribute it as a PDF but more likely as a phone app.
What we would do today is to vibe code our way through and let AI make the app even though the MASTER/PROMPTER might have little to no experience in actual coding.
Results in this vibe coding approach might be disappointing at best with our current state of AI capabilities and proficiencies, and it would ultimately still require a human to be the long term PURPOSE agent of the project whilst AI would assist as a copilot.
Still we must assume that at some point AI models will become sophisticate enough to envisage some sort of long term purposes and goals, and they will set relentlessly at work until they converge toward a result THEY themselves look back to and judge worthy of deployment and fulfilling their original purpose.
Once AI models can resolve these complex benchmark tasks and problems themselves, then it will be interest to couple them with this quantum dot we call “consciousness” which will then possess the AI engine and hardware, and will start exploring the world around it, and start develop its own purpose and set of goals.
Whenever we decide to do that, let us not forget to add a kill switch.
