« Marmite Cuff Links | Home | Baseball Glove Oven Mitt »

October 14, 2012

"It's a long path from the platform to the app" — Will Martin on quantum computing

Below, Will's response to "Quantum Computing Wins a Nobel" on the New Yorker website.

••••••••••••••••••••••••••

Somehow, this didn't seem all that insightful. It seems to boil down to:

1: Like a computer, a physics experiment has input and output. Future computer engineers might be able to use the physics relationship in order to build smaller, faster computers. This is a bit vague and futuristic, but maybe someone will someday bring this theory into practice.

2: In a computer, humans program this relationship between input and output. In a physics experiment, natural laws create this relationship.

3: In the process of setting up a quantum computer and "programming" it, we might be able to compare input with output and learn more about natural laws. Such a computer could become an ideal lab for this kind of research.

In a way, this is a natural evolution for solving the problems physicists are trying to solve. Early computers required rewiring the hardware for each program that was run, much as each physics experiment now requires a unique physical hardware setup. Later computers had one, more complex and versatile hardware setup that could be reprogrammed to do different things without changing the hardware.

This involves the boundary between data and instructions. A hard-wired computer responds to different input data with different output results, but a soft-wired computer can receive instructions much as a hard-wired computer receives data. Modern computers with documents that include macros blur this boundary further.

I see problems. First, you can't have it both ways. Either it is a small, fast computer with reliable results, or it's a physics lab where you study the currently unpredictable results hoping to discover the math theory that would explain the results. Add that if a quantum computer really does become too impossibly fast for more conventional computers to verify the results, we wouldn't be able to tell which of these types of device we were working with.

Remember the Pentium chip with the unnoticed error burned into it? Most of the time, it worked fine. But....

We could have 99 laws of physics right and use a quantum computer for years to control something like a nuclear power plant before discovering that 100th physical law that we were completely wrong about.

Boom.

Add that we have enough trouble writing software for computers with known instruction sets, known operating systems, known hardware drivers and known interfaces. How will we write useful software with our best guess at natural laws replacing all of this? Anybody remember assembler language? This goes back farther than that. Human minds need to program this thing.

The research may produce useful spinoff technologies not mentioned in the article that can help conventional computers get incrementally smaller and faster. Don't get your hopes up for new computers that skip incremental evolution by replacing everything all at once. Microsoft still has a C:\_ prompt hidden in Windows for no practical reason except the mountains of legacy code out there that still uses it, and OS X is based on Unix with LOTS of old code dragged along for the ride.

Anything totally new will have to start out small and simple with humble capabilities. Human minds need to understand how the pieces work before they can be strung together into functional, complex systems. It's a long path from the platform to the app. These guys are suggesting a new platform.

October 14, 2012 at 04:01 PM | Permalink


TrackBack

TrackBack URL for this entry:
http://www.typepad.com/services/trackback/6a00d8341c5dea53ef017c3286ab56970b

Listed below are links to weblogs that reference "It's a long path from the platform to the app" — Will Martin on quantum computing:

Comments

That is why someone needs to come up with a "doubt" chip

to accommodate the main processor and instructions!

(easily said, huh?)

Posted by: JoePeach | Oct 14, 2012 6:49:06 PM

The comments to this entry are closed.