Review of Nick Bostrom's book Superintelligence on the risks of artificial intelligence
Artificial Intelligence (AI)
Software that performs human-like functions is generally regarded as artificial intelligence, until it is so commonplace that no one is impressed any more that something more complicated becomes the next target of AI. There have been several AI winters – where early success in one area failed to develop into wider progress. But recently, narrow AIs have become superhuman in many games including chess, backgammon, scrabble; and more recently Jeopardy! and Go.
I'll leave the philosophical debate of can a machine really be conscious, or is its intelligence really only the output of the human written software, to the academics, but what is definitely coming soon is machines far more powerful than a human brain. Think how many times more powerful a modern day PC is compared to one 20 years ago - eventually that will be the difference between human level intelligence and a super-AI-intelligence. It won't be the difference between someone of average intelligence and Einstein, it will be like a mouse trying to understand quantum physics. We will be the new mouse and won't be able to understand the thoughts and capabilities of this superintelligence.
And we will have no choice but to give those machines control of our lives, for example it won't be humanly possibly to control the increasing air/rail/road traffic so those quick decisions will have to be made by some sort of artificial intelligence. Why it's a question of survival is because we don't know if, Terminator style, the machine will suddenly decide that its existence is more important that its day job.
Or as Nick Bostrom explains in his excellent book, Superintelligence, it may not even be a malevolent AI - but one that tries too hard to please as (its creator) and uses up all the resources on the planet in doing so. Alternatively, if we didn't clearly define the goal for the AI, then it may solve world hunger by killing most of the humans on the planet meaning there was plenty of food to go around.
This is one area that I did disagree with Ray Kurzweil in his book The Singularity Is Near - he assumes that any human created higher intelligence would treat us as a curiosity and care for us. As this isn't how we treat lower animals to us (which we cull, abuse and eat) I don't think we can be sure that we would be treated any differently.
Artificial Intelligence (AI) News
AI holds the promise of longer lives, shorter workweeks, and economic growth
CNBC - 03-Oct-2023
Jpmorgan CEO sees AI transforming healthcare, workforce, and global economyRead more...
Navigating AI in Healthcare: The Opportunities and Challenges of Language Learning Models
JAMA - 27-Mar-2023
Applications and risks of AI tools while considering ethical, legal, and regulatory implicationsRead more...
Research Scholars Programme at Future of Humanity Institute - now open
Future of Humanity Institute (FHI) - 12-Aug-2020
Research long-term consequences in Oxford with world leading thinkers such as Nick BostromRead more...
IBM Watson will be better than other commercial artificial intelligence systems
Independent - 11-Mar-2020
Will artificial intelligence surpass human intelligence one day?Read more...
DPhil Scholarships available at Oxford's Future of Humanity Institute
University of Oxford - 04-Oct-2019
An amazing opportunity to work with some of the world's leading thinkersRead more...
Artificial Intelligence (AI) Resources
Click on resource name for more details.
In Apocalyptic AI, Robert Geraci offers the first serious account of this "cyber-theology" and the people who promote it
This book focuses on the implications of a future society faced with an abundance of human and artificial intelligence written by James D. Miller