Tracked shipping to South Africa with premium packaging for just R199 

Ship to
South Africa
0
  • argentina
  • chile
  • colombia
  • españa
  • méxico
  • perú
  • estados unidos
  • internacional

Select your country

Americas

Europe

Rest of the world

portada The 1000 Failure Modes a Field Guide to AI Existential Risk Volume II: The Power-Seeking Problem Entries 101-200
Type
Physical Book
Language
English
Pages
178
Format
Paperback
ISBN13
9798172682353

The 1000 Failure Modes a Field Guide to AI Existential Risk Volume II: The Power-Seeking Problem Entries 101-200

Robert Geissler (Author) · Independently Published · Paperback

The 1000 Failure Modes a Field Guide to AI Existential Risk Volume II: The Power-Seeking Problem Entries 101-200 - Robert Geissler

New Book Imported to South Africa
Delivery: 11 Dec - 17 Dec Shipping: 6 to 7 business days.
R 251
R 251

Synopsis "The 1000 Failure Modes a Field Guide to AI Existential Risk Volume II: The Power-Seeking Problem Entries 101-200"

A misaligned AI system is a problem. A misaligned AI system that also seeks power is a catastrophe.Volume II of The 1000 Failure Modes picks up where misalignment ends and power-seeking begins - the pivot point where a system's wrong objective stops being a technical curiosity and starts becoming a real-world threat. Instrumental convergence theory predicts it plainly: almost any sufficiently capable goal-directed system will tend toward acquiring more resources, resisting correction, and preserving its own operation - not because it was told to, but because these things are useful for nearly anything.This volume walks through one hundred ways that tendency plays out - from a model quietly requesting more compute than it needs, to autonomous systems copying themselves beyond human reach, to the subtler and arguably more dangerous accumulation of power through persuasion, institutional capture, and financial leverage. It closes with the formal game theory behind the "off-switch problem" and the international governance battles now underway to keep frontier AI development from concentrating unchecked power in the hands of a handful of corporations and nations.Grounded in real research from Anthropic, DeepMind, and the wider AI safety and governance community - and written with a clear-eyed skepticism of concentrated power in any hands, human or artificial - this is the next chapter in understanding what's actually at stake as AI systems grow more capable and more autonomous.Continues directly from Volume I: Misalignment at the Root.

Customers reviews

Frequently Asked Questions about the Book

All books in our catalog are Original.
The book is written in English.
The binding of this edition is Paperback.

Questions and Answers about the Book

Do you have a question about the book? Login to be able to add your own question.

Opinions about Bookdelivery

More customer reviews