 |
| From the cover of Robot Visions by Isaac Asimov |
Your Android phone (or iPhone, if that’s how you roll) is an impressive machine, with computing speeds and storage capacities thousands of times those of desktop PCs from only years ago. If Moore’s Law holds up, your smart watch may outshine today’s phones the way today’s phones eclipse old PCs.
But no matter how powerful these machines become, they may never develop true intelligence if we continue to rely on conventional computing technology. According to the authors of a paper published in the journal Physical Review X last July, however, adding a dash of quantum mechanics could do the trick.
The problem lies in part with the step-by-step processes that limit conventional artificial intelligence learning algorithms. The authors of the paper equate it with classical random walk searches. Random walks are sometimes described as being like the stumbling of a drunk person – each step is about the same size, but the direction of the steps are random. Random walkers can cover a lot of territory, and an artificial intelligence system that explores various problems with random walk learning algorithms can eventually learn new behaviors, but it takes a long time.
Quantum walks, on the other hand, describe a walker who doesn’t exist at one spot at a time, but instead is distributed over many locations with varying probability of being at any one of them. Instead of taking a random step to the left or right for example, the quantum walker has taken both steps. There is some probability that you will find the walker in one place or the other, but until you make a measurement the walker exists in both.
Compared with a random walk, quantum random walks are much, much faster ways to get around. To the extent that learning is like taking a walk, quantum walks are a much faster way to learn.
That’s not to say you’d need to make a full-blown quantum computer to build a truly intelligent machine – only part of an otherwise classical computer would need to be supplemented with a bit of quantum circuitry. That’s good because progress toward developing a stand-alone quantum computer has been about as slow as the progress toward artificial intelligence. Combining artificial intelligence systems with quantum circuitry could be the recipe we need to build the HAL 9000s and R. Daneel Olivaws of the future