training data

Benn Jordan investigates infrasound and low-frequency noise from data centers, including xAI's Colossus and a crypto mine, and reports his own experiments on the effects. The video covers the health and community impacts on nearby residents.

Eager Space examines the practicality of putting data centers in orbit, walking through three scenarios with cost calculations and the launch cost reductions that would be required.

Scott Manley examines why SpaceX, Google and others are looking at orbital data centers. He walks through the economics, heat rejection and solar power arguments, with links to the Lumen/Starcloud and Google Suncatcher papers.

Scott Manley explains the thermal balance of spacecraft and how a satellite could be configured to radiate away the heat from datacenter-scale computing in space. His rough numbers suggest it is feasible.

Scott Manley analyzes the idea of putting data centers in orbit, prompted by Lumen Orbit's funding. He examines the claimed advantage of constant solar power against launch cost, cooling and mass.

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.

minutephysics, in collaboration with the US Census Bureau, explains differential privacy and why the 2020 census adds noise to protect individuals' data. It walks through how statistics can leak private information.

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

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.

Musician Benn Jordan uses AI to detect AI-generated music and digs into music industry data, including streaming payouts and sample library copyright problems. He covers ways to tell whether a track is human-made.

Hans Rosling uses augmented-reality animation to show 200 countries over 200 years in four minutes, built from 120,000 data points. It tracks life expectancy and income to show the world's development.

Roy Beck's short gumball presentation using World Bank and Census data to argue that letting people immigrate to the US cannot solve world poverty because the numbers are so large. It is made by NumbersUSA, an immigration-restriction advocacy group.

A Kurzgesagt explainer on why the human population exploded and why growth is slowing, drawing on Our World in Data. It argues that birth rates fall as countries develop, easing overpopulation fears.

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.

Kurzgesagt examines who is responsible for climate change, comparing historical, current and per-capita emissions by country. It argues that individual footprints are a distraction from larger responsibility.

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.

Benn Jordan's video about storing a PNG image in a living bird, a starling that learns to reproduce encoded sounds. It mixes ornithology, audio engineering, and data encoding.

Matt Parker examines claims that Biden's 2020 vote counts violate Benford's Law, explaining the law and why it often does not apply to precinct-level election data.

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.

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.

Scott Manley tells the story of a computerized telemetry system built to sit at the heart of a nuclear test device in place of the plutonium pit, designed to transmit data in the microseconds before it was destroyed.

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

CGP Grey explains the history of the US Social Security card and number, and why they became a de facto national ID despite never being meant for it. He also covers why America resists a proper identity card.

Everyday Astronaut's Tim Dodd runs an hour-long, data-driven analysis of what rockets actually emit and how that compares with airliners and other industries. It breaks down exhaust by propellant type, per-passenger and total emissions, and ways launches could be cleaner.

FortNine looks at the data on why newer pickup trucks with tall, blunt grilles kill so many pedestrians, cyclists and motorcyclists. It argues that lowering hood height would substantially cut fatalities.

CGP Grey explains that postal codes are routing instructions for mail rather than geographic regions. He looks at the quirks and problems this causes.

Mick West analyzes the Navy "Go Fast" UFO footage and shows with simple trigonometry and the data overlay that the object is slow and high up, consistent with a balloon. He explains how parallax makes it look fast and argues the original TV claims were wrong.

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

Matt Parker of Stand-up Maths asks whether a country's reported land area accounts for hills and mountains, working through how surface area depends on the resolution of elevation data. It follows up on a question from a viewer and work by researchers who measured it.

Technology Connections examines an old feature-rich microwave and shows how it uses a humidity or steam sensor and lookup data to cook food well, arguing it beats many modern ones.

Aging Wheels tests how towing trailers affects the range of electric trucks, including a Silverado EV. It shows that trailer aerodynamics matters far more than added weight.

A three-minute animated video that explains sexual consent through the analogy of offering someone a cup of tea. It is based on a blog post by Emmeline May.

Geekerwan reviews smartphone CPUs by power efficiency rather than benchmark scores, with test methodology and chapters on Qualcomm, Apple, Samsung and Google chips. It explains why high benchmark scores can coexist with heat and poor battery life.

Michael Wesch's five-minute video about Web 2.0, showing how the read/write web, XML and user-generated content change how people create and share information.

John Green explains why the US spends far more on health care than other countries, going through insurers, drug companies, hospitals and lawyers. He argues the causes are complicated rather than a simple blame game.

Branch Education explains how Starlink works, from the ground dish and satellites to phased-array beam steering and data routing. Animations cover how a beam stays connected to a satellite moving at 27,000 km/h.

Veritasium explains how investigators caught the Golden State Killer using forensic genealogy, matching crime scene DNA to distant relatives in public genealogy databases. It also covers the privacy implications of DNA databases.

A short 3Blue1Brown primer on exponential and logistic growth, using early COVID-19 case data to show how epidemics grow and eventually level off. It explains why small changes in growth rate make enormous differences in outcomes.

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