The win over the Go champ was 4-1 in favor of AlphaGo, part of Google DeepMind. Go is a strategy game, and according to the experts, “brute force” like that used in a chess challenge with Gary Kasparov isn’t an option. You have to outthink your opponent.
The theory of AI as a threat to humanity, however, is a big vague. Killer toasters, disgruntled cars, you name it, the ideas are there, but the realities are looking pretty badly defined. The most obvious threat is autonomous robots, which, of course, have nothing better to do than destroy humanity (like humanity needs any help in that regard). Military robots, in theory, particularly if left in charge of major weapons systems, could do that.
But enough of this recycling of every sci-fi movie ever made about rampaging robots and churlish computers. A more realistic threat would be the various physical and economic lifelines of society — electricity, water, banks, services, fuel, reactors, media, etc. The world could come to a standstill if those were turned off. The Internet of Things (online home appliances, cars, etc.) could turn feral easily enough, particularly with a bit of help from some added codes.
Then there’s the “they’re taking over” scenario — AI replaces accountants, lawyers, financiers, truck drivers, even artists and whole corporations which deal mainly in information and data or anything else which can be managed by AI. That’s already happening. The mining industry has giant trucks driven by code. Spreadsheets can do accounts. A slightly different version of MYOB could replace a lot of people in seconds. Who needs people, when a 20c bit of software can do the same jobs? What do people do, all go in to politics? Become waiters, the famous “service industries” solution to the demise of meaningful work?
Another, very interesting, and very necessary, issue with AI is “cognitive computing”. This is about computer learning — computers which can write their own codes. This link leads to current news about cognitive computing, and there’s pages and pages of it. Can computers teach themselves to do without human beings? My bet is that if the finance sector can do it, as it’s been doing for decades, a basic calculator would be able to figure it out.
The other issue – Why do we assume AI is dangerous?
The one thing that’s not getting a lot of attention is that it’s almost natural to see AI as a threat. Consider this — a logical mentality, based on numbers, is a threat to a barely rational species whose most recent, in fact, almost only, achievement is turning their own planet in to an overpriced toxic dunghill. Judging machines by that logic says a lot about human judgment, and nothing good.
AI is likely to be as dangerous as its applications. The danger from human paranoia, justified or otherwise, is much worse. There is no current basis in fact to assume AI is anything but what it is — a machine intelligence, of whatever level. Assigning human motives to it is absurd.
The trouble is that someone, in their sublime idiocy, might give an AI human motives. Let’s not overlook human stupidity as a working dynamic in history. People do dumb things. Some years ago, someone tried to hybridize rat and human DNA. Imagine the likely results.
At the moment, people are people and machines are machines. Let’s leave it that way, and see what happens.
