One of the many things that’s easy for (adult) humans and very difficult for computers is image recognition. Take twenty thousand pictures of a street, half with cars and half without. A human being will separate car pictures from non-car pictures with about 100% accuracy (if not falling asleep from boredom). Not so for computers, at least for computers as we generally know them. Even with special image recognition software and hours of ‘training’, the best a standard computer will do is around 60%, and that’s under ideal conditions where the pictures are large and clear. Is this more than a technical problem? Sure is. With the advent of digital photography the world produces billions of photographs. Collections of photos routinely run to millions. Right now, these pictures cannot be effectively searched based on the image content.
The example of choosing between 20,000 pictures with and without cars was recently demonstrated by – no surprise – Google. What was something of a surprise, the computer Google used was not a conventional computer, but one that incorporates at least some of the principles and functionality of quantum computing. Partly because of its enormous Google Images collection and partly out of interest for anything that pertains to digital searching, Google has been pursuing ultra-fast computing for some time.
That Google would investigate quantum computing seems natural. The big promise of quantum computers is that they provide an alternative to traditional computers (i.e. the microcomputer and all such), an alternative potentially much faster at certain kinds of computation. Quantum computing takes advantage of the behavior of atomic particles at or very near absolute zero temperature. The quantum behavior allows storage and computation based not on the traditional notion of ‘bits,’ represented by 0’s and 1’s; but of ‘qubits’ (quantum bits) represented by 0’s, 1’s and, in a sense, both 0’s and 1’s (known as quantum superposition). Qubits can operate simultaneously, holding different ‘states’ (in a sense, values). Three qubits can hold a superposition of 8 states, whereas a traditional computer using three bits can be in only one of those states at a time.
Yes, quantum computation is difficult to comprehend, sometimes even for the practitioners; but it works. At the moment, scientists are struggling with the physical implementation of qubits (i.e. a kind of chip expressed in quantum gates) and in figuring out what kinds of problems can be solved (best) by quantum computing.
The computational system used by Google is manufactured by D-Wave (Burnaby, British Columbia, Canada: D-Wave). I won’t go into the details of what D-Wave calls adiabatic quantum computing, it’s quantum oriented processor (chip), nor the skepticism surrounding the hardware and techniques developed by D-Wave. The controversy is interesting, if only because it illustrates the difficult boundaries between for-profit companies and their desire to patent their findings, and the desire of science (and scientists) to know how leading-edge work is being done and whether it’s verifiable, or not. It appears, however, that whatever the theoretical limitations of D-Wave’s computational system – which appears to be a clever hybrid of traditional computing with sub-routines (algorithms) that use some form of quantum computing – Google is satisfied that it is getting significant results.
According to Google, the 20,000 pictures with/without cars, can be solved by the D-Wave system much faster than by any of the other computer systems at Google’s disposal.
It’s important to keep in mind that this and at least for the moment any other implementation of ‘quantum computing’ is experimental. The complexity of working with quantum anything is formidable – and always a potential for error is high. Much of the work is done with excruciatingly advanced mathematics and very sophisticated models – not, it should be noted – with actual physical systems, although like D-Wave there are even commercial attempts at harnessing a physical reality for quantum computing.
As you might guess, quantum computing development is expensive, neither the specialists nor the required equipment come cheaply. So the presence of numerous academic programs and the efforts of behemoths such as Google, IBM, and Microsoft indicate that the payoff may be great. In the area of image processing along, there are literally hundreds of potential applications (sorting pictures, catching terrorists, reading traffic images…).
Quantum computing and image recognition
One of the many things that’s easy for (adult) humans and very difficult for computers is image recognition. Take twenty thousand pictures of a street, half with cars and half without. A human being will separate car pictures from non-car pictures with about 100% accuracy (if not falling asleep from boredom). Not so for computers, at least for computers as we generally know them. Even with special image recognition software and hours of ‘training’, the best a standard computer will do is around 60%, and that’s under ideal conditions where the pictures are large and clear. Is this more than a technical problem? Sure is. With the advent of digital photography the world produces billions of photographs. Collections of photos routinely run to millions. Right now, these pictures cannot be effectively searched based on the image content.
The example of choosing between 20,000 pictures with and without cars was recently demonstrated by – no surprise – Google. What was something of a surprise, the computer Google used was not a conventional computer, but one that incorporates at least some of the principles and functionality of quantum computing. Partly because of its enormous Google Images collection and partly out of interest for anything that pertains to digital searching, Google has been pursuing ultra-fast computing for some time.
That Google would investigate quantum computing seems natural. The big promise of quantum computers is that they provide an alternative to traditional computers (i.e. the microcomputer and all such), an alternative potentially much faster at certain kinds of computation. Quantum computing takes advantage of the behavior of atomic particles at or very near absolute zero temperature. The quantum behavior allows storage and computation based not on the traditional notion of ‘bits,’ represented by 0’s and 1’s; but of ‘qubits’ (quantum bits) represented by 0’s, 1’s and, in a sense, both 0’s and 1’s (known as quantum superposition). Qubits can operate simultaneously, holding different ‘states’ (in a sense, values). Three qubits can hold a superposition of 8 states, whereas a traditional computer using three bits can be in only one of those states at a time.
Yes, quantum computation is difficult to comprehend, sometimes even for the practitioners; but it works. At the moment, scientists are struggling with the physical implementation of qubits (i.e. a kind of chip expressed in quantum gates) and in figuring out what kinds of problems can be solved (best) by quantum computing.
The computational system used by Google is manufactured by D-Wave (Burnaby, British Columbia, Canada: D-Wave). I won’t go into the details of what D-Wave calls adiabatic quantum computing, it’s quantum oriented processor (chip), nor the skepticism surrounding the hardware and techniques developed by D-Wave. The controversy is interesting, if only because it illustrates the difficult boundaries between for-profit companies and their desire to patent their findings, and the desire of science (and scientists) to know how leading-edge work is being done and whether it’s verifiable, or not. It appears, however, that whatever the theoretical limitations of D-Wave’s computational system – which appears to be a clever hybrid of traditional computing with sub-routines (algorithms) that use some form of quantum computing – Google is satisfied that it is getting significant results.
According to Google, the 20,000 pictures with/without cars, can be solved by the D-Wave system much faster than by any of the other computer systems at Google’s disposal.
It’s important to keep in mind that this and at least for the moment any other implementation of ‘quantum computing’ is experimental. The complexity of working with quantum anything is formidable – and always a potential for error is high. Much of the work is done with excruciatingly advanced mathematics and very sophisticated models – not, it should be noted – with actual physical systems, although like D-Wave there are even commercial attempts at harnessing a physical reality for quantum computing.
As you might guess, quantum computing development is expensive, neither the specialists nor the required equipment come cheaply. So the presence of numerous academic programs and the efforts of behemoths such as Google, IBM, and Microsoft indicate that the payoff may be great. In the area of image processing along, there are literally hundreds of potential applications (sorting pictures, catching terrorists, reading traffic images…).