robert miles

Robert Miles introduces the first of several videos on the paper 'Concrete Problems in AI Safety', focusing on how AI systems can cause accidental negative side effects. It explains the problem in accessible terms.

Robert Miles explains instrumental convergence: why an AGI with almost any goal would tend to pursue sub-goals like self-preservation and resource acquisition. It builds on Omohundro's basic AI drives and his orthogonality thesis video.

Robert Miles explains the orthogonality thesis: how an agent can be highly intelligent while pursuing goals that seem stupid or alien to humans. He separates intelligence from goals and touches on the stamp-collector example.

Robert Miles asks whether corporations can be thought of as artificial general superintelligences and how far the analogy holds. He argues the differences, such as being limited by employees' capabilities, matter for AI safety.

Robert Miles examines whether the case for AI safety is a Pascal's mugging, a tiny probability of an enormous payoff. He argues it is not, because AI risk does not depend on vanishingly small probabilities.

Robert Miles goes through ten common reasons people give for dismissing AI safety, adapted from Stuart Russell's list, and answers each one. It is aimed at skeptics and newcomers.

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.

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.

CGP Grey walks through the odd history and quirks of the US-Canada border, including how it was surveyed and its irregularities such as exclaves.

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 Computerphile video in which Rob Miles warns against assuming a general AI will think like a human. He uses the thought experiment of a stamp-collecting AI to show how a goal-driven system could pursue its aim with dangerous results.

Rob Miles explains on Computerphile why implementing an off switch for a general AI is hard. A capable agent would resist being stopped because shutdown prevents it from achieving its goal.

Rob Miles explains for Computerphile why simple rule sets like Asimov's Three Laws cannot make robots and AI safe. Terms like harm and human are too ambiguous to specify precisely.

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.

Sabine Hossenfelder explains how close nuclear fusion really is to break-even, and why headlines about ITER and lab milestones are misleading. She distinguishes plasma gain (Q plasma) from whole-system gain (Q total) and shows the latter is still far below one.

The first part of Extra History's animated miniseries on the South Sea Bubble. It explains how England's government debt led Robert Harley to turn to financier John Blunt, whose schemes began inflating artificial demand for shares.

A 20-minute S3 episode touring Terraform Industries, a startup making synthetic natural gas from air and solar power. It walks through the reactor, injection system, direct air capture and electrolyzer, plus the company's cost goals and next milestone.

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.

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 explains the nuclear salt water rocket, Robert Zubrin's concept of a rocket fueled by a rapid nuclear reaction in uranium salt dissolved in water. He covers its extraordinary performance and the dangers of managing it.

Tom Scott visits Peachtree City, Georgia, which has about a hundred miles of dedicated golf cart paths that serve as an alternative to driving. He explains how the network works as a quasi-transit system that keeps cars off the roads.

Physics Girl compares hydrogen fuel cell cars and battery electric cars after driving 1800 miles in a Toyota Mirai. The video was part of a series sponsored by Toyota.