the archive
episodes
Every episode is a thought experiment. We summon the philosophers, artists, and anatomists of the Old World and ask them to debug the present one.
// latest episode
EP. 052
Philosophy + AIWhat Would Socrates Do With a Language Model?
The dialectic method meets the transformer. We ask whether the Socratic tradition of relentless questioning is the perfect antidote to AI hallucination — and whether the hemlock was just a bad prompt.
EP. 051
Music + MachinesThe Renaissance Composer and the Generative Model
Palestrina wrote in strict counterpoint. So does GPT-4 when you ask nicely. A study in pattern, constraint, and beauty.
EP. 050
Anatomy + ImagingDa Vinci's Notebooks Were a Training Set
Leonardo catalogued the body with obsessive precision. Modern medical AI does the same. What is lost when the curiosity is automated?
EP. 049
Architecture + FormVitruvius, Palladio, and the Parametric Turn
Classical proportion as algorithm. From the Pantheon to parametric design: how generative form inherits a very old argument about beauty.
EP. 048
MathematicsFibonacci's Sequence and the Attention Mechanism
The golden ratio shows up in sunflowers, nautilus shells, and — arguably — the weight matrices of transformer models. A meditation on nature's favourite algorithm.
EP. 047
Natural SciencesVesalius, Dissection, and the Ethics of Training Data
Andreas Vesalius revolutionised anatomy by looking directly at bodies rather than trusting inherited texts. AI researchers face the same tension: observe the world, or trust the corpus?
EP. 046
PhilosophyPascal's Wager and the AI Alignment Problem
Blaise Pascal placed a bet on God under conditions of radical uncertainty. AI safety researchers are placing a similar bet. We examine the odds.
EP. 045
Music + AICounterpoint, Cadence, and the Creative Machine
Can an AI compose music that moves you? We test four models against Bach and ask a composer to grade the results. The findings are humbling — for everyone.
EP. 044
Navigation + SystemsDead Reckoning and the Hallucinating Model
Before GPS, sailors estimated position using speed, heading, and time. Large language models do something similar — with similarly unpredictable drift. A story about navigating without ground truth.