Superintelligence
In Superintelligence, Oxford philosopher Nick Bostrom asks what happens if and when artificial intelligence exceeds human intelligence. He explores the paths by which a superintelligent agent might emerge, the speed at which it could gain power, and the difficulty of controlling a system far smarter than its creators. Bostrom introduces ideas such as the control problem, instrumental convergence, and the risk of a single agent gaining a decisive strategic advantage. He argues that aligning an advanced AI's goals with human values is both extremely difficult and extremely important. Rigorous and methodical, the book draws on philosophy, computer science, and economics. Published in 2014, it helped move AI safety into mainstream discussion and influenced researchers, technologists, and policymakers thinking about long-term risk.
What you might want to know about Superintelligence
The questions readers send us most often, answered without spoilers.
A philosopher examines the risks and possible outcomes of creating machine intelligence that surpasses human capability.
Bostrom argues that machine intelligence surpassing human capability could pose serious risks, and that aligning its goals with human values is difficult but essential.
It was published in 2014 by Oxford University Press.
Superintelligence was written by Nick Bostrom, published in 2014 by Oxford University Press.
Superintelligence is 352 pages in standard print editions, though page counts vary slightly between hardcover, paperback, and large-print formats.
At an average reading pace of about 250 words per minute, Superintelligence takes most readers 5 to 8 hours to finish.
Superintelligence is available in hardcover, paperback, ebook, and audiobook formats from Amazon, Bookshop.org, ThriftBooks, and most major bookstores.