Dyon fields + True AIs must keep lossless versions of everything, they must memorize their source data and references

Yi Ma @YiMaTweets  9.11 is still larger than 9.8, despite can memorize solutions to PhD level questions. Again, memorizing is not understanding and knowledge is not intelligence.
Replying to @YiMaTweets


Dyon fields + True AIs must keep lossless versions of everything, they must memorize their source data and references

Yi Ma, Humans can let the computer memorize for them, if the computer memorizes and gives it back exactly when tasked.

If what is memorized works (is reliable) then you just treat it like a compiled function in computer terms. Just like clicking a button on a calculator. Same as addition subtraction multiplication division, simple integrals, simple derivatives, using Maxwells equation (a particular version that has been tested and verified), a lookup of data from an MeV GeV TeV electromagnetic experiment on dyons, a lookup of data from a climate simulation, the results of compiling all the stuff on the Internet for site:cern.int

Explaining Dyon Fields a bit: [ —–
Julian Schwinger made up the name. But I am generalizing it. A dyon field is any region (voxel) with known or modeled charge and magnetic structure. Any region with moving positive and negative charges however small or fast. Dyons can be made out of gluon plasma. The vacuum can be modeled as overlapping dyon fields, however temporary or small. The Higgs field can be nicely modeled in terms of potentials, gradients, charge, flows, rigidity, and many other properties — using electromagnetic units, which can be converted into any other units. The biggest reason Schwinger could not get people to work together was they all insisted on using only what they were familiar with. A dyon region is a plasma.  The reason for remembering that it has electric and magnetic field and those are dynamic is to remember that every voxel is important and might make a critical difference.  A dyon can be a 3D basis set of many forms. A dyon remembers the legacy of Schwinger.  It is a useful unique tag on the Internet like #dyon or #dyons.

Since anything with a neutron will have internal electric charge and magnetic gradients, that means pretty much everything, even if you have to excite some of the states by using multiple crossed beams and MegaTesla pulsed fields (high harmonic gain lasers) and radioactive beam sources.  Any moving particle will likely be “dyonic” part of the time. Regions of “binding energy” have their own field, but it has to use standard units.  It is not so locked down and fought over it cannot be adapted and updated.

https://en.wikipedia.org/wiki/Dyon.

Twisted particles in heavy-ion collisions by Alexander J. Silenko, Pengming Zhang, Liping Zou at https://arxiv.org/abs/2101.03620

Equation of spin motion for a particle with electric and magnetic charges and dipole moments by Alexander J. Silenko at https://arxiv.org/abs/2309.04985

Spin rotation as an element of polarization experiments on elastic electron-proton scattering by Leonid M Slad at https://arxiv.org/abs/0904.1671

Logic and numbers related to solar neutrinos by Leonid M Slad at https://arxiv.org/abs/2408.06041# (12 Aug 2024)

Maury Goodman (anl.gov) has a Neutrino newsletter “Long Baseline news” for Aug 2024 the last link is Leonid Slads paper. August is at https://www.hep.anl.gov/ndk/longbnews/2408.html and the Index is at https://www.hep.anl.gov/ndk/longbnews/index.html

I have been telling groups “it is electromagnetism all the way down” and I did not know that Julian Schwinger wanted to go that direction. I know almost exactly what it means for new global industries.
] —-

The fundamental flaw of ALL those GPT “AIs” now is they did not losslessly index their source data, and they did not learn how to program and run algorithms and software. So it is an “open” system in the bad sense – they left out critical data, without which no problem can be solved completely.
 
Memorization (exact, compiled, pre-tested, pre-verified, traceable datasets and algorithms and systems) are the core of human knowledge as it is currently stored. An “AI” that every time tries to figure out how to add or divide two numbers in scientific notation (ChatGPT always fails) is wasting valuable time, and missing critical opportunities.
 
I have spent many 10s of thousands of hours checking how large groups at global scale work out problems in every field. They all use compiled software, perfectly memorized datasets, perfectly memorize fundamental constants and reference data. They do NOT use 3E8 for the speed of light and gravity. They do not round at every step like most every “AI” now.
 
Stop saying “memorization” like it is always a bad thing for human children. “Memorization” also means “lossless and perfect storage and retrieval of information of all sorts”. A human does mostly cannot losslessly store and retrieve images. But a computer with memory and cameras and retrieval software can. (a camera with a computer, a smart camera, an intelligent camera, an intelligent sensor)
 
Groups making sloppy generative AIs that do not memorize their input data, including links to sources of all generated actions and conclusions — should be made to sit in the corner of their place on a small stool and say a billion times “NEVER let the AI do anything unless it has memorized what is in the source data, the statistical index, and what the users are doing”
 
I cannot summarize decades of working out how to make the Internet “lossless”, “perfect memory”, “perfect lossless data sharing”, “able to work in all human languages”, “able to work in all domain specific languages”.
 
Richard Collins, The Internet Foundation
 
Richard K Collins

About: Richard K Collins

The Internet Foundation Internet policies, global issues, global open lossless data, global open collaboration


Leave a Reply

Your email address will not be published. Required fields are marked *