I also think that an AI that suddenly 'wakes up' and decides to rebel against it's creators is more fiction than science. And I agree as well that a more serious danger is AI that interprets it's programmers goals the wrong way. But, IMHO, this is the most serious short-term risk: an AI that's intentionally designed to be malicious. There are many reasons why such AI would be designed: 1. Spammers deigning a virally spreading AI to help distribute their messages or produce fake clicks on Google ads that bring them money. This is already happening, but it's not too intelligent yet. 2. Organized crime using it to hack into banks or to bring down the IT systems of corporations that don't cooperate. 3. Terrorists hoping to shut down the economy or other critical systems to create fear. 4. A disgruntled employee using it to harm his employer.
zkam831 votesThursday, January 24, 2008 at 3:11 PM
I don't understand this obsession with computing power. When humans finally learn how to implement software as smart as a human brain, they'll realize that computers have long been powerful enough.
airobo1 votesThursday, January 24, 2008 at 9:57 PM
Excellent point. Our brains are complex but once we decode the process we'll optimize it better. Plus most of our brain is devoted to maintaining our bodies and regulating emotions, which an AI won't need.
Not to be negative, but this really isn't a particularly well written article. The guy takes some very basic facts about the history of computer technology and its current state, extrapolates out (as many others have done), and then throws in a dusting of baseless speculation. It's FUD of the worst order. Until we actually see some real self-actualizing AI, there really is no way to know whether a machine intelligence will be able to circumvent limitations and rules built into its prgramming structure or not.
In addition, what would be the purpose in creating self-aware, generalized artificial intelligence? Is it realistic to assume that someone is going to invent an unfettered AI just for the sake of intellectual curiosity? In my opinion, for the foreseeable future, AI will be limited to intelligent agents that have fairly limited scope -- within their purview, they might become exceptionally capable and surpass human capability, but there is a wide gap between that and the kind of generalized knowledge that would be required for an AI to start taking over society of its own volition.
Yes, I think there are already a lot of people trying to invent self-aware AI just for the sake of intellectual curiosity. But regardless of weather the AI is self-aware or not, I believe there is real danger from advanced AI if it falls in the wrong hands.
1. Spammers deigning a virally spreading AI to help distribute their messages or produce fake clicks on Google ads that bring them money. This is already happening, but it's not too intelligent yet.
2. Organized crime using it to hack into banks or to bring down the IT systems of corporations that don't cooperate.
3. Terrorists hoping to shut down the economy or other critical systems to create fear.
4. A disgruntled employee using it to harm his employer.
And many more...