model s

Tom Scott tells how John Reber's plan to dam San Francisco Bay was disproved by the Army Corps of Engineers' giant hydraulic Bay Model in Sausalito. The model showed the plan would have been dangerously wrong.

Rob Miles explains on Computerphile how language models generate plausible text and how transformers improved on recurrent networks, with a look at what comes next.

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.

Rob Miles discusses GPT-3 on Computerphile, covering how scaling parameters kept improving results over GPT-2 and how the model nearly learned arithmetic without anyone knowing exactly how.

An engineering explainer on why Tesla abandoned the induction motor for the Model 3 and used an IPM-SynRM design. It covers the reluctance torque and permanent magnet principles behind the choice.

A Wendover Productions video on how dollar store chains expanded across America and why they displace grocery stores in small towns. It covers their business model and supplier leverage.

Sabine Hossenfelder explains climate sensitivity and why recent observations suggest climate models may have underestimated warming. She says this changed her view on how worried to be.

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.

A Wendover Productions explainer arguing that major airlines make much of their profit from loyalty programs and co-branded credit cards rather than flying. It frames airlines as quasi-banks that sell points in bulk, using United's pandemic-era loan as a hook.

PBS Space Time explains the Standard Model Lagrangian, the single equation encapsulating the particle physics theory, and visualizes what its terms mean.

Munro Live's Model Y episode 36 has Sandy Munro walk through the disassembled Tesla Octovalve and heat pump system. He explains how the components are made and what each one does.

Munro Live's team breaks down the thermal management system of the Ford Mustang Mach-E and compares it with the Tesla Model Y. They look at parts count, coolant loops and heating approaches.

Linus Tech Tips breaks down the company's income streams, including sponsors, YouTube Premium, merch and other products. It gives percentages for each revenue source.

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

A guest video by Welch Labs on 3Blue1Brown explaining how text-to-image and video models work. It covers CLIP, shared embedding spaces, diffusion models (DDPM, DDIM), conditioning, guidance and negative prompts.

Veritasium's Derek Muller argues that Facebook's business model works against pages and creators, because it charges them to reach the audiences they built. He explains how filtering and paid promotion undermine the value of likes.

3Blue1Brown builds toy SIR models to simulate how an epidemic spreads and tests scenarios like social distancing, travel restrictions, central gathering places and quarantine. It was made during the early COVID-19 outbreak.

A short Al Jazeera animation narrated by Amy Goodman presenting Noam Chomsky's five filters of the mass media: ownership, advertising, media elite sourcing, flak and a common enemy. It is part of the MediaTheorised series.

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.

A Welch Labs explainer on neural scaling laws and the compute-efficient frontier, asking why model loss seems to hit a line it cannot cross. It draws an analogy to an ideal gas law for AI.

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.

Tom Scott's long explainer on YouTube's copyright system, Content ID and the DMCA. He argues the platform's system is a workaround for a broader copyright system that does not fit the modern world.

Veritasium explores planned obsolescence, from the Phoebus lightbulb cartel to printers and phones. It explains how companies deliberately shortened product lifetimes and why products aren't better.

Real Engineering explains why SpaceX is building Starlink, including the economics of satellite internet. It covers the latency advantage of light in vacuum over fiber for long-distance links.

Wendover Productions explains how UK rail privatization worked and why it failed. It covers the fragmented operator and infrastructure structure and why market forces apply poorly to rail.

RethinkX summarizes its report arguing that mainstream LCOE estimates are inflating a bubble in coal, gas, nuclear and hydro assets. It claims solar, wind and batteries will strand conventional energy investments.

Marques Brownlee explains why iMessage's blue and green bubbles matter, covering the technical, commercial and social side of Apple's messaging lock-in. It also discusses RCS and what could be done about it.

Part one of Technology Connections' account of the VHS versus Betamax format war. It argues Sony was caught off guard by JVC's cheaper, more practical VHS system.

Extra Credits explains from a game developer's perspective why AAA games have stayed at $60 despite inflation, and how that pressure drives DLC, microtransactions and loot boxes.

Asianometry explains why India, a chip design powerhouse, cannot fabricate chips, tracing the 1980s government attempt to build a national semiconductor champion and how it collapsed. It includes the sabotage at the SCL fab.

Asianometry examines why Japan's software industry has lagged despite its hardware strengths. It looks at dependence on US software and the low prestige of programming.

Tom Scott's video whose title shows its own live view count, and the story of how he made that work. It explains why open APIs were once common and why they largely disappeared.

Asianometry traces the 30-year rise and peak of Japan's semiconductor industry from the 1950s to the 1980s, covering how a postwar country gained competence in an emerging technology.

Tom Scott visits IEX, a stock exchange that slows down trading with a long coil of fiber. The coil removes the speed advantage of high-frequency traders.

Tom Scott explains the famous Dutch flower auction clock at Royal FloraHolland in Aalsmeer, a literal Dutch auction, and how the internet changed it. It is a short look at how the system worked then and now.

Engineering Explained follows up on Tesla's claim that a Cybertruck towing a Porsche 911 could beat the 911 alone over a quarter mile. He walks through the MotorTrend drag race and explains why the claim did not hold up.

Potholer54 presents the climate science case and a market-oriented, conservative-friendly approach to addressing global warming, with sources listed. It is part one of a two-part series.

Yannic Kilcher walks through the Attention Is All You Need paper, which introduced the Transformer, a sequence model based solely on attention without recurrence or convolutions. He reviews the architecture and the paper's translation results.

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.

Cody's Lab builds a 200-billionth-scale model to show how far away the nearest stars are, using parallax and light speed. A short astronomy explainer on interstellar distances.