It’s the ultimate ‘what if.’
What if there was a computer or software program so powerful and encompassing it could beat the Las Vegas sport betting line? What if it could digest the vast variables of the financial markets and pick stocks with stunning accuracy? What if it could examine the genetic code of the human race and pick out those most likely to succeed or fail?

What if it could design a better international airline schedule in which interconnecting flights wouldn’t require never-ending layovers? What if, theoretically, it was possible to build a machine so fast it could defeat any encryption within minutes?

Two announcements this winter made some of these scenarios reality — putting the future of computing front and centre, both in terms of software (see sidebar) and hardware.

This future is quantum computing, an evolutionary step that will move computing to the sub-atomic level. Until just recently, this had been a purely theoretical idea — scientists understood the concepts involved but no one had actually built one of these ultra computers.

Until now. Vancouver-based D-Wave announced in February it had built the first commercially viable quantum computer with expectation of production by 2008. It took two-and-a-half years of development and $30 million in venture capital to get it this far, yet the announcement barely registered a blip in the media, even though there is a buzz within the scientific community.

Quantum computing is based on quantum mechanics, a highly complex science that deals with the predictive behaviour of invisible particles. Beyond that it gets confusing fast: in the “classical” computing model instructions are in binary code, meaning a series of ones and zeros represent “on” and “off” in a language strung together as bits. This creates traditional equations like “if this is true, then do this; if this is not true, then do that.”

In quantum computing the variables — called Qubits — are either on, off or both, and it’s that third option that imparts both speed and power to the quantum processor. This processor functions in an analogue state, sorting electrons spinning in one direction or another according to programs at the sub-atomic level.

It’s a crude explanation but its implications are far reaching, according to D-Wave CEO Herb Martin. “At our demonstration we had a working model of a 16 Qubit quantum computer, not a product but a proof of concept,” he said. “And we assigned it a series of tasks.”

The first was to search a database of drugs and find one with a molecular structure similar to a new drug.

Point a traditional computer at the 40 million lines of data involved and it would compute an answer in days, if not weeks. Classic computers work by trial and error, comparing and discarding until they have a match. Huge databases equal lots of time, and if you add variables a program can run for months or years.

Quantum computers, on the other hand, examine a massive number of variable at once, thus dramatically cutting run times. D-Wave’s prototype, Orion, a big red boiler-like tank super-cooled with liquid helium to just above absolute zero, processed variables in batches of 216, Martin said. “A 32 Qubit machine would process at 232, so you can see how you get these massive numbers very quickly.”

While Orion isn’t currently setting any speed records — and probably isn’t much faster than the average laptop — the point was simply to demonstrate a working model. D-Wave expects to be running 32 Qubits this year and a whopping 512 Qubits a year or so from now, with 1,028 Qubits as the eventual goal. 

Not a revolution yet 
This news will not cause anyone to dump Microsoft or Intel stock just yet. Even assuming D-Wave publishes its research and peer reviews uphold its claims, the key application of its quantum computers will be large-scale projects such as biometrics and financial risk analysis.

So far the D-Wave quantum computer has turned a couple of party tricks. Aside from a drug database search (which could replace some parts of clinical trials by predicting the behaviour and effects of new drugs) it also strategically planned wedding seating and solved a Sudoku puzzle.

Also, there are currently a couple of roads to massive computational power, such as distributed computing, which harnesses a team of dormant servers in off-peak times to process large equations, or software such as the recently announced “affinity propagation” algorithm developed at the University of Toronto (see sidebar).

But there is no doubt quantum computing will open many doors one day.

One will be in nanotechnology, that frontier of science that creates new materials by manipulating sub-atomic particles. Another is in atmospherics, because its masses of data and endless variables have long challenged scientists. Quantum power could result in more accurate weather forecasts and hurricane tracking, for example, saving both lives and property.

In fact, just about any massive database — the human genome project for example — could be queried in a far more casual manner since current searches tend to take a lot of time and are expensive.

Similarly, financial institutions develop formulae to calculate the risk factor of the long-term bond market. The variables are complex and a single shift in one number requires long hours of computation to re-evaluate the data. A quantum computer could run the task in a few minutes. 

Waiting on a review 
But is it real? Some are waiting for the data to be published and peer reviewed before they pass out the laurels. The only demonstration thus far has been via video, though D-Wave plans to rent time on its quantum machine as an adjunct, not replacement to, traditional digital computing.

So it’ll be some time yet before you can get the desktop version and conquer world hunger or figure out the most likely Lotto 6/49 combinations. But stick around, it could get interesting.

SIDEBAR

Thinking machines
It’s one of those passionate arguments that drags on well past closing time at local bars where geeks gather: is it hardware or software that pushes the computing envelope?

The announcement of a working prototype of a commercially viable quantum computer would seem to push that argument in favour of the hardware proponents, but Prof. Brendan Frey of the University of Toronto’s Edward S. Rogers Sr. Department of Electrical and Computer Engineering has just had a better idea.

Using a fairly standard hardware configuration, Frey and his team have created an algorithm that fundamentally changes the way most computers function. It’s called “affinity propagation” and it’s designed to organize large amounts of data such as you’d find in astrophysics, drug research, medical imaging, scheduling or even traffic planning.

Instead of the classic “if-then” (better known as trial and error) approach, affinity propagation instructs the machine to consider everything holistically and to weigh the data according to additional enhancements from the user.

Frey started working on the algorithm about six years ago for a genetic research program and realized it could be adapted for other applications.

The remarkable element is that the process is almost organic in nature. “All possible hypotheses are considered and these many hypotheses exchange messages in a manner similar to messages exchanged between neurons in the brain, until a good set of hypotheses is found,” Frey said. “It is much more human and in that sense more like artificial intelligence.”

It divvies up similar data into groups and continually aggregates it until the problem is reduced to its essential elements.

As a proof of concept, Frey and his grad student Delbert Dueck ran the program to analyze flight schedules across North America to identify which specific cities in Canada and the U.S. were most easily accessible based on a series of variables such as headwinds, stopovers and traffic.

The key was to “weight” cities by giving them a numeric value based on their similarity and proximity, so Toronto would be weighted like Montreal in reference to New York.

The results could cut pollution by making route planning more efficient, Frey said, but what’s more remarkable is that they were obtained after just a couple of hours of processing, compared to several weeks using classic computing.

And while the flight scheduling search is a simple demonstration, Frey is confident the theory can be applied to other large problems. “The algorithm is posted and people can download it and try it,” said Frey, whose work was published in Science. “We’ve had inquiries from all over.”

One of the more unusual inquiries was from an anthropologist who wants to apply it to a database of skulls in an effort to find similarities and perhaps unearth the “missing link.” Someone else is looking at basketball games, and Frey said there’s no reason it couldn’t be applied to any sport in an attempt to find a predictive outcome.

Hot Tech Archive