llm training

A short explainer on the history and mechanics of public key cryptography and the Diffie-Hellman key exchange. It uses a mixing-paint-colors analogy to show how two parties can agree on a shared secret in public.

A short Art of the Problem clip explaining Diffie-Hellman key exchange, using a paint mixing analogy for one-way functions and touching on modular arithmetic and discrete logarithms.

3Blue1Brown's visual explanation of how transformers and GPT-style language models work, tracing data through tokens, embeddings and attention.

Yosh trains a reinforcement learning agent to drive Trackmania until he can no longer beat it. The video explains how the training works and the design choices involved.

Rick Beato argues that adults cannot develop true perfect pitch, which he ties to critical-period brain plasticity and language acquisition. He says it can only be developed in young children, and discusses the research behind this.

Robert Miles explains reward modelling, in which an AI learns what humans want from their preference comparisons instead of a hand-written reward function. He walks through the 2017 paper on deep reinforcement learning from human preferences.

FortNine's tongue-in-cheek guide to why drivers fail to see motorcyclists and how riders can make themselves more visible. It touches on perception psychology, including the role of saccades.

A short cartoon explaining how language models like ChatGPT work, from autocomplete and conditional probabilities to neural networks, gradient descent and backpropagation. It shows how networks act as universal function approximators.

Kurzgesagt examines how AI-generated content is flooding the internet, from video and music to news and books. It looks at how this makes verifying what is real harder.

Alex O'Connor shows that ChatGPT cannot draw a wine glass filled to the brim, and ties the failure to David Hume's missing shade of blue. It asks what the failure says about how image models learn.

Rob Miles explains "glitch" tokens, odd strings in large language model vocabularies that make the model produce gibberish. He covers how they likely arise from tokenizer training data, such as Reddit usernames, that never appeared in the model's own training.

Chubbyemu presents a case study of a man who followed AI health advice, replaced salt with sodium bromide and developed bromism with severe neurological effects. It is told as a dramatized medical case by a doctor.

Computerphile video in which Mike Pound explains a University of Maryland paper on watermarking LLM output. The method biases token choices so generated text can later be detected statistically.

3Blue1Brown's visual walkthrough of the attention mechanism inside transformers, covering self-attention, multiple heads and cross-attention. It is chapter 6 of his deep learning series.

Adam Neely explains what perfect (absolute) pitch is, how it differs from relative and quasi-absolute pitch, and why he would not want it. He covers its drawbacks, including drift with age.

3Blue1Brown's second neural network chapter explains cost functions and gradient descent, showing how a network learns by minimizing error during training.

Veritasium explores why the number 37 turns up so often, drawing on Tom Magliery's collection of 37 sightings and a poll of about 200,000 people asked to pick a random number. It covers how humans choose numbers and why 37 stands out.

Musician Benn Jordan explains and demonstrates adversarial noise that poison-pills music files against AI training and scraping.

A 1951 Bell System training film showing how telephone calls are routed through a step-by-step (Strowger) switching office.

Robert Miles covers a paper where researchers ran real versions of his earlier mesa-optimizer thought experiments. They show inner misalignment, where an AI learns the wrong goal in training and misbehaves when deployed.

Practical Engineering explains how continuous welded rail copes with thermal expansion by restraining it, using the relationship between temperature, stress and strain. The video is part of a series on railways.

A SAAMI-produced training video for firefighters explaining what happens when sporting ammunition burns, including that unchambered cartridges lack the energy to penetrate much. It includes test footage of ammunition in fires.