If you have a garden of rocks, it is possible that no rock has the same weight as the average. ~Carl Jung
AI models train with a huge number of initially random parameters (you can include architecture search). The information contained in the outputs of an untrained (random) network must be miniscule relative to the dimensionality (number of paths) of the search space. That means that the models are very low complexity even though the number of parameters is high.
Our hands have many degrees of freedom, nodal and distributed.
I am doing research into how we can use our hands like an abacus.
Addition, Subtraction, Multiplication and Division Practice
Another Derivation of the Quadratic Formula
Generalization of the Pigeonhole Principle
Fuel Loop Problem Proof Sketch (with extension to applied problem)
Research Programs (see Imre Lakatos)
Game Integration / Game Theory Alternatives
Example of a Classical Game (can be a collaborative game)
Statistics
There is a common misconception that data typically is normally distributed.
Quantification Problem 2 with Safety Applications
Multiple Sensor Averaging can Reduce Error
(As a note, sometimes too much input to an algorithm can make a problem harder to solve. For example, if too many people in a car are giving the driver instructions, that can make driving more difficult.)
Exponential Linear Region (for quick risk calculations given pressure)
Group Theory
Number Theory
Real Number Subtraction Theorem
Multiplicative Commutativity (inductive proof and non axiomatic proof)
Basis Representation Theorem optional material
Largest Public Prime Number code reference