Posts

Showing posts from November, 2026

Grasshopper

Image
 

Calypso and the Secret of Ogygia

  Calypso and the Secret of Ogygia A CCSLM Odyssey Far beyond the Pillars of Hercules, where the Mediterranean gives way to the endless blue, there lay an island of extraordinary beauty. It was called Ogygia. Its trees bore golden fruit, its streams ran with crystal water, and its meadows were carpeted with flowers that never faded. The goddess Calypso lived there, immortal and beautiful, surrounded by every luxury imaginable. But Ogygia had a secret. For Calypso was not merely a goddess of beauty. She was a mistress of transformations. She had spent thousands of years studying the nature of the universe, and had developed a peculiar theory. She believed that everything in existence could be understood through the repeated application of complicated nonlinear transformations. Her palace was filled with enormous machines, each designed to bend, stretch, twist and distort the geometry of whatever entered it. She called them her Neural Engines. Unfortunately, despite centuries of effo...

Sinbad and the Seven Seas of Linear Algebra

  Sinbad and the Seven Seas of Linear Algebra A tale from the Arabian Nights, as retold in the age of CCSLM. Long ago, in the great city of Baghdad, there lived a merchant sailor named Sinbad who had grown tired of transporting spices, silk, and precious stones. He had heard rumours of a strange new treasure, said to be more valuable than all the gold in the Sultan's treasury. It was called Intelligence. And so Sinbad set sail. The First Voyage: The Island of Nonlinearity After many days at sea, Sinbad's ship came upon an extraordinary island. Its mountains twisted, its valleys folded into themselves, and its forests grew in directions that seemed to change whenever one looked away. On the shore stood a company of learned mathematicians. "Welcome, traveller!" cried their leader. "You have arrived at the Island of Neural Networks. Here we create intelligence through nonlinear transformations." Sinbad was impressed. "How do you make these transformations?...

The Priory Dispute III

  The Priory Dispute III The Abbot's Emergency Meeting A confidential gathering of the higher orders. Setting: The Chapter House. Midnight. All senior monks have been summoned. Two people have specifically NOT been invited: Terence Tao and Jürgen Schmidhuber. The Abbot enters, looking deeply troubled. The Abbot: Brothers, I fear we have a serious problem. For decades we have imagined neural networks as collections of advanced nonlinear functions, each performing its own little piece of geometric gymnastics on the data. But have we stopped to consider what happens when we compose all these elaborate transformations? We could end up with hopeless geometric chaos! At the very least, how would anything generalise? Brother Anselm: Surely the training process sorts it all out? The Abbot: That is not an explanation. That is a prayer! Brother Matthias: Perhaps we have been looking at the wrong thing. Consider linear mappings. They perform relatively simple geometric operations: rotati...

The Priory Dispute II

  The Priory Dispute II The Case of the Missing Geometry A mathematical farce in one act. Setting: The Great Hall of the Priory of Applied Mathematics. A large geometric diagram has been pinned to the monastery wall. It shows a cloud of points gradually rotating, stretching, folding and changing shape through a sequence of linear transformations. Underneath is written: GLOBAL GEOMETRIC GRADUALISM Terence Tao and Jürgen Schmidhuber are standing on opposite sides of the diagram. The Abbot enters, visibly exhausted. The Abbot: Gentlemen! I have been informed that there is another dispute. Tao: A serious matter, Your Reverence. The phrase Global Geometric Gradualism is mathematically misleading unless properly qualified. Schmidhuber: That's not the dispute. He claims I didn't think of it first. Tao: I claim no such thing. I merely observe that the underlying mathematics has been known for considerably longer than either of us. Schmidhuber: Knowing the mathematics is not the ...

Global Geometric Gradualism

Image
  From Local Nonlinearity to Global Geometric Gradualism A different way to picture what happens inside a neural network When people imagine what happens inside a neural network, they often picture a succession of nonlinear functions, each performing its own little piece of mathematical gymnastics on the data. These functions bend, reshape, clip and otherwise manipulate the results, with the overall intelligence emerging from their coordinated activity. But there is another, rather different way to understand the process. Instead of imagining countless little nonlinear operations doing complicated things to the data, we can view a neural network as progressively transforming the geometry of its data through a sequence of relatively simple linear mappings. This is the idea of Geometric Gradualism . Complexity through accumulation A linear mapping is geometrically straightforward. It can rotate, reflect, stretch, compress, shear or collapse dimensions. These are familiar operations f...

The Name of the Rose

The Priory Dispute: A Tragedy in Several Linear Mappings A fictional mathematical comedy in one act. Setting: A medieval priory. Two distinguished scholars have been summoned before the Abbot to settle a dispute concerning the ownership of a particularly interesting neural network architecture. On the table lies a manuscript entitled: Concatenated Context-Selected Linear Transformations: A New Foundation for Intelligence. Tao: I believe we should begin by establishing precisely what is being claimed. Schmidhuber: Naturally. The idea is that a sequence of linear mappings, selected according to context, can produce arbitrarily sophisticated nonlinear behaviour. Tao: Yes, but that is rather obvious once one understands the composition of functions. Schmidhuber: Composition is not the point. The point is that the mappings are selected by an intelligent mechanism. Tao: Then you have introduced a nonlinear selection mechanism. Schmidhuber: I have introduced context. Tao: Context is a...

Funktionieren

Image
 

Centralizing Neural Networks on Linear Algebra

Image
1. Executive Summary Modern deep learning relies heavily on stacked non-linear activation functions, which often obscure the underlying geometric transformations of a network and complicate interpretability. Concatenated Context Selected Linear Mappings (CCSLM) reframes neural computation by separating routing decisions from data transformations . By utilizing binary context signals to dynamically select or assemble linear mappings, CCSLM restores pure linear algebra as the central operational paradigm of neural architectures. This structural clarity unlocks explicit Context Engineering , Linear Mapping Engineering , and direct spectral analysis via Singular Value Decomposition (SVD) . 2. Core Architectural Framework In a CCSLM framework, a layer's computation is decoupled into two distinct mechanisms: context generation and linear transformation. y=M(c)x where x is the input vector, c represents a discrete context signal, and M(c) is the context-dependent linear operator appl...

The nonlinear defense

Image
 The nonlinear defense "Why do you need nonlinear functions?" Tao was asked. "Because intelligence is nonlinear." "Why?" "Because linear functions cannot do nonlinear things." "But what if the linear mappings change according to context?" "That's nonlinear." "Yes. But the actual data transformation is linear." "That's cheating." "Isn't that what intelligence is supposed to do?"