When software gets smarter, the first effect is to empower the already powerful. The fantastic powers available now to Google and Facebook, which are now in practice the publishers of most of what appears on the public internet, is one example. More sinister is the power of nation states to spy on us, to manipulate their own citizens, and to disrupt the workings of their enemies. But these advantages cannot last. Soon they have to be reinforced by law, and ultimately force, as the techniques behind them spread and hardware grows cheaper and more plentiful.
The speed of technological progress, and the ease with which ideas can now spread, mean that few techniques can long remain the preserve of large firms or entities. Every advance in power and convenience available to the ordinary consumer will soon be available to criminals too. Illegal commerce, whether in drugs, forged documents, stolen credit cards or emails, is nearly as slick and well organised as the legal sort. So are the criminal world’s labour exchanges: hiring someone to hack a website, or to boost your Twitter account with fake followers, is easily done. So is renting a botnet of suborned devices to knock an enemy’s website off the net. Last year large chunks of the consumer internet in the US were knocked out for hours, apparently by an assault launched from subverted home security cameras.
We are on the brink of an explosion of devices that do not look like computers but will all be connected to the net and all potentially hackable. The more complex, useful and intelligent they are, the greater the harm they could do. Even if they are not easily hacked when they are installed, they will be more or less impossible to keep secure as new vulnerabilities are discovered. Software, like everything else we build, must constantly be repaired if it is not to decay. This is not a technological problem, or at least it is not a problem with a technological fix: it will need coordinated political, social and bureaucratic action over a period of decades.
As Professor Ross Anderson told a conference on the future of artificial intelligence last week, this is not a matter of regulating the technology in itself; it is the devices in which the software is embedded, and the firms which make and sell them, that need regulation. The spread of artificial intelligence (AI) downwards and outwards from the few large firms that now deploy it will pose further problems. Some of the most impressive recent advances in the field – such as a program which can beat even the word’s best Go players – owed a lot to the use of adversarial learning, whereby different programs were trained by competing with each other inside the computer.
If a program can be built to beat the best human Go players, it should certainly be possible to build some to beat most security experts, and soon this will require no exceptional skills. The benefits of ubiquitous networked intelligence are real enough. But we are wrong to think that the only real danger comes from its capture by a few large companies or states. There will also be hundreds of smaller and more purely malevolent groupings using technology against us for their own purposes. The street finds its own uses for things, as William Gibson noted in the short story Burning Chrome, and governments, police and private citizens must all be prepared for what is coming when the street finds its own uses for AI.
from Artificial intelligence (AI) | The Guardian http://ift.tt/2uITfdG