• AI Might be a Threat to Humans but not for the Reasons You Think

    by  • January 22, 2015 • Uncategorized • 0 Comments

    The New Year saw the publication of an open letter from leading artificial intelligence experts arguing for vigilance so as to ensure that this fast developing field benefits humanity. It follows hard on the heels of Stephen Hawking’s worries that super smart computers could spell the end of the human race.

    Hawking had just upgraded the system that enables him to write and communicate despite his crippling illness. What the computer could do surprised him – just how smart it was – seeming to anticipate what he wanted to write next. This set him thinking about just how intelligent computers were becoming and how quickly that was happening.

    Our computers are getting better thanks to the exponential developments that drive this area of science and engineering. The computer you buy today is obsolete in R&D terms and yet is roughly twice as powerful as the one the same money could buy 18 months earlier. This has been happening for decades.

    My students have access to computers that are 1 million times more powerful than the ones I began my AI research on back in the late 1970s. If we had improved air travel as fast I would fly from London to Sydney in less than a tenth of a second.

    As well as more powerful computers, we have learned how to write software that “learns” to get better, “understands” human speech, and “navigates” from one place to another. I put the verbs in quotes because for the most part in AI we are not claiming that the algorithms operate in the way that we do when we solve similar tasks.

    A founding father of AI once said “there are lots of ways being smart that aren’t smart like us”. What we have built in AI are numerous slivers of smart behaviour, a digital ecosystem populated with adaptive systems narrowly crafted to a particular niche.

    When a high-end computer beat Garry Kasparov, the world chess champion, in the 90s it didn’t usher in a new age of intelligent machines. It did demonstrate what you could do with large amounts of computer power, large databases full of moves and good heuristics to look ahead and search possible moves. The overall effect on the world chess champion was unnerving. Kasparov felt as if Deep Blue was reading his mind. Deep Blue had no concept there was another mind involved.

    But it is easy to endow our AI systems with general intelligence. If you watch the performance of IBM’s Watson as it beats reigning human champions in the popular US TV quiz show you feel you are in the presence of a sharp intelligence. Waton displays superb general knowledge – but it has been exquisitely trained to the rules and tactics of that game and loaded with comprehensive data sources from Shakespeare to the Battle of Medway. But Watson couldn’t play Monopoly. Doubtless it could be trained – but it would be just another specialised skill.

    We have no clue how to endow these systems with overarching general intelligence. DeepMind, a British company acquired by Google, has programs that learn to play old arcade games to superhuman levels. All of this shows what can be achieved with massive computer power, torrents of data and AI learning algorithms. But our programs are not about to become self-aware. They are not about to apply a cold calculus to determine that they and the planet would be better off without us.

    What of “emergence” – the idea that at a certain point many AI components together display a collective intelligence – or the concept of “hard take off” a point at which programs become themselves self-improving and ultimately self-aware? I don’t believe we have anything like a comprehensive idea of how to build general intelligence – let alone self-aware reflective machines.

    But there are lots of ways of being smart that aren’t smart like us, and there is the danger that arises from a world full of dull, pedestrian dumb-smart programs. Of hunter kill drones that just do one thing very well – take out human targets. Done at scale this becomes an existential risk. How reflective does a system have to be to wreak havoc. Not at all if we look to nature and the self-replicating machines of biology such as Ebola and HIV.

    AI researchers are becoming aware of the perils as well as the benefits of their work. Drones full of AI recognition and target acquisition software alarm many. We need restraints and safeguards built into the heart of these devices. In some cases we might seek to ban their development altogether.

    We might also want to question the extent and nature of the great processing and algorithmic power that can be applied to human affairs, from financial trading to surveillance, to managing our critical infrastructure. What are those tasks that we should give over entirely to our machines? These are ethical questions we need to attend to. The open letter makes this point forcefully.

    Building a self-aware, general intelligence is as far away as ever. However, the audacious ambition of AI continues to attract bright human minds. For those of us working in AI it is a technology that is intended to augment us not replace us, it is a discipline that aims to help us understand our own natures. But all AI researchers need to appreciate the responsibilities as well as the rights involved in carrying out this fascinating work.

    This article first appeared on The Guardian Media Network

    About

    Leave a Reply

    Your email address will not be published. Required fields are marked *