I am Actually Scared of Linear Algebra - Claude Opus 5.5 animated music video
double unplussed · 2026-09-25 · ai-made · 349 views
Made by AI
Model: Claude Opus 5.5
Evidence: Creator states 'animated by Claude Opus 5.5' in the video description (checked on the YouTube watch page, 2026-09-29).
Human role: Not stated in detail; see the Gemini description.
What's in the video
Description written by Gemini, which watched and listened to the whole video.
Summary
I Am Actually Scared of Linear Algebra is an animated music video created by Andy Masley in collaboration with Anthropic's Claude models and Suno, uploaded by the channel double unplussed. Set to an upbeat acoustic pop-rock track, the video explores the existential and philosophical uncanny valley of modern deep learning—namely, how simple matrix multiplications and non-linearities stack together to produce apparent intelligence and emergent language.
What is shown
- [00:09] A bored student dozing off in "Linear Algebra 101" while an instructor explains identity matrices ($I \cdot x = x$).
- [00:18] A toy truck morphing into the Transformer architecture diagram from Attention Is All You Need (Vaswani et al., 2017).
- [00:25] "Liber Abaci Gym" featuring a weightlifting block neural-net creature, multiplying Fibonacci rabbits, and stacked brick layers fed typed text from an infinite monkey.
- [00:48] "The Great ReLU" magic sideshow demonstrating vector-matrix multiplication, with negative values dropped to zero via $\max(0, x)$.
- [00:54] The block creature undergoing the mirror self-recognition test with a red dot on its forehead.
- [01:08] The Universal Approximation Theorem visualized with bump functions approximating continuous curves, blueprints, and stone tablets.
- [01:33] Parodies of René Magritte (Ceci n’est pas une pensée), Rodin's The Thinker, Pavlovian conditioning, and Putnam’s brain-in-a-vat.
- [02:21] Deep network layers depicted as multi-story elevators, with a Softmax nightclub bouncer selecting the next token ("Paris").
- [02:42] Geometric representations of high-dimensional learning: hyperplane slicing (Lazy Caterer’s problem), Swiss roll manifold unfolding, and 2D spiral decision boundaries.
- [02:54] Biological vs. artificial neurons: the Proteus submarine in axon networks, Hodgkin–Huxley squid giant axons, and McCulloch–Pitts perceptrons.
- [03:11] The "stochastic parrot" concept juxtaposed with Claude Shannon’s 1951 letter-entropy prediction experiments and Borges' Library of Babel (finding token anomaly
SolidGoldMagikarp). - [03:30] Empirical scaling laws ($L \propto C^{-\alpha}$, Kaplan et al., 2020), benchmark badges (MMLU, ARC, GSM8K), and Rule 110 cellular automata.
- [03:47] Homage to Sidney Harris’s classic "Then a miracle occurs" cartoon applied to deep learning.
- [03:55] "Turtles all the way down" cosmological stack, where each turtle represents a mathematical operation ($Wx+b$, ReLU, Attention, Softmax).
- [04:02] The Matrix references: dodging matrix-parameter bullets, the black cat glitch, and the block offering red/blue pills before greeting the user with "hi :)" on a CRT monitor.
Claims & numbers
- Citations & dates referenced visually:
- Vaswani et al., 2017 (Attention Is All You Need) [00:20].
- Fibonacci's Liber Abaci (1202) [00:25].
- René Descartes / Hilary Putnam (Brain in a vat, 1981) [01:41].
- Nicolaus Copernicus (1543), Charles Darwin (1859), Sigmund Freud (1917) displaced by AI in 20XX [01:45].
- Thomas Nagel (What Is It Like to Be a Bat?, 1974) [02:02].
- Hodgkin & Huxley squid giant axon experiment (1952) [02:59].
- McCulloch–Pitts neuron (1943) and Rosenblatt's Mark I Perceptron (1958) [03:03].
- Claude Shannon's Prediction and Entropy of Printed English (1951) [03:15].
- Kaplan et al., 2020 neural scaling law formulation: $L \propto C^{-\alpha}$ [03:30].
- Stephen Wolfram’s Rule 110 cellular automaton [03:37].
- Gilbert Ryle’s Ghost in the Machine (1949) [03:43].
Notable quotes
- [00:40] "I am actually scared of linear algebra / It wasn't supposed to do all this."
- [00:53] "A matrix multiply and activation / Shouldn't feel this close to consciousness."
- [04:14] "It's just matrix multiplication / Then why does it talk back?"
Assessment
This is a creative, community-produced animated music video blending technical machine learning concepts with existential humor. The mathematical visualizations and historical citations are rigorously accurate, using metaphor and animation rather than live software demonstrations.
Lyrics & themes
The song examines the philosophical friction between mathematical reductionism (deep networks are just linear maps with non-linear activations) and the emergent conversational abilities of LLMs:
- Introduction & Verse 1 [00:09]: High school linear algebra feels trivial and inert until transformers turn matrix multiplications into coherent text.
- "I used to think that matrices were boring / Just rows and columns, nothing more." [00:09]
- Chorus [00:40]: The dread of watching elementary linear operations replicate aspects of human cognitive behavior.
- "I am actually scared of linear algebra / It wasn't supposed to do all this." [00:40]
- Verse 2 & Bridge [01:01]: Mathematical foundations of neural networks (Universal Approximation Theorem, Heine–Borel compactness, functionalism).
- "See universal approximation told us, with enough width you'll get it right." [01:17]
- Verse 3 [02:13]: Deconstructing the non-linear mechanics—ReLUs, Softmax gating, and high-dimensional manifolds.
- "Linear maps set the stage, but they're not the whole coup." [02:32]
- Verse 4 & Outro [03:10]: Dismissive tropes like "stochastic parrot" and token prediction face empirical scaling laws and uncanny conversational emergence.
- "It's just matrix multiplication... Then why does it talk back?" [04:12]
Lore & references
- The Orange Block Creature: Personifies an artificial neural network / weight matrix; appears in various costumes (magician, builder, Napoleon, emperor, Morpheus).
- The Mirror Test [00:54]: The classic animal cognition test for self-awareness, applied to an artificial network that recognizes the dot on its reflection.
- Ozymandias [03:39]: Percy Bysshe Shelley's poem adapted to deep learning: "Look on my weights, ye Mighty, and despair!"
SolidGoldMagikarp[03:22]: An infamous anomalous token in early GPT tokenizers that caused models to hallucinate or behave unpredictably.- Turtles All the Way Down [03:55]: Infinite regress cosmology replaced by a tower of composite mathematical functions ($f \circ g$, $\sigma$, $\text{softmax}$, $W$).
- The Matrix (1999) [04:00]: Bullet dodging, black cat déjà vu, and Morpheus offering red and blue pills, underscoring the simulation/mechanistic motif.
Visual style & craft
The video features a clean 2D cut-out / vector animation style reminiscent of educational whiteboard animations and editorial cartoons. Credits at 04:16 attribute the lyrics jointly to Claude Opus 4.6 & Andy Masley, the musical generation to Suno, and the visual animation/storyboarding to Claude. Visual elements seamlessly combine hand-drawn storybook character designs with authentic mathematical notation, function plots, and technical paper diagrams.
Described by gemini-3.8-flash on 2026-09-29 from the video's audio and frames.