SolidGoldMagikarp
A Descent into the AI Underworld
Will be published approx. on 27. August 2026
Book
Hardback
368 pages
978-1-3996-3588-2 (ISBN)
Description
When English mathematician Matthew Watkins went to San Francisco to study an early version of ChatGPT, he expected dry technical work. Instead, he uncovered a series of 'rare tokens' that made the system go haywire. Some produced gibberish. Others turned it hostile. One trigger word - petertodd - sent the machine into what looked like full-blown paranoia. Was it a simple bug? A trail of breadcrumbs left by whistleblowers? Or a glimpse into how AI really thinks?
The search for answers pulls Watkins into the secretive world of AI safety research. In this rapidly expanding domain, the line between science and belief blurs: billionaires play god while rogue chatbots win fortunes and groupies. What starts as a computer mystery becomes something much bigger: both a window into the eerie inner worlds of LLMs and a cryptic clue to what might be history's greatest power grab.
The system has since been patched. Even so, before the window closed, we saw inside for a brief moment - and what stared back at us was both stranger and more human than we might like to admit.
The search for answers pulls Watkins into the secretive world of AI safety research. In this rapidly expanding domain, the line between science and belief blurs: billionaires play god while rogue chatbots win fortunes and groupies. What starts as a computer mystery becomes something much bigger: both a window into the eerie inner worlds of LLMs and a cryptic clue to what might be history's greatest power grab.
The system has since been patched. Even so, before the window closed, we saw inside for a brief moment - and what stared back at us was both stranger and more human than we might like to admit.
Reviews / Votes
Top notch. Unfolding like a nerdy occult techno-thriller, Watkins tells the weird tale of how he discovered two ghosts in the LLM machine - 'glitch tokens' whose uncanny behavior makes it clear, at least, that AI will enchant as much as it will disrupt. The deets are important, and Watkins presents them with care and clarity. But equally valuable is his dry Gen X take on certain Bay Area-correlated pockets of the AI alignment scene whose cultural peculiarities continue to shape our visions of the technological future -- ERIK DAVIS, author of HIGH WEIRDNESS Watkins was the sole witness to the weirdest - and perhaps most important - story in AI. His rigorous obsession with this mystery will become yours -- CULLEN HOBACK, director of MONEY ELECTRICMore details
Language
English
Place of publication
London
United Kingdom
Publishing group
Orion Publishing Co
Target group
College/higher education
Professional and scholarly
Product notice
Trade binding
Dimensions
Height: 240 mm
Width: 156 mm
ISBN-13
978-1-3996-3588-2 (9781399635882)
Copyright in bibliographic data and cover images is held by Nielsen Book Services Limited or by the publishers or by their respective licensors: all rights reserved.
Schweitzer Classification
Other editions
Additional editions
E-Book
approx. 08/2026
Weidenfeld and Nicholson
€16.99
Not yet available
Person
Matthew Watkins is a British mathematician and AI safety researcher. While studying language models at SERI MATS in Berkeley and London, he discovered 'a cluster of strange keywords that will break ChatGPT' (Vice). Posts about his findings for the tech-rationalist website LessWrong were among the most upvoted in the site's history. Leading AI critic Eliezer Yudkowsky called his research 'one of the more hopeful processes happening on Earth right now'.