Meta-Blog

SEARCH QandO

Email:
Jon Henke
Bruce "McQ" McQuain
Dale Franks
Bryan Pick
Billy Hollis
Lance Paddock
MichaelW

BLOGROLL QandO

 
 
Recent Posts
The Ayers Resurrection Tour
Special Friends Get Special Breaks
One Hour
The Hope and Change Express - stalled in the slow lane
Michael Steele New RNC Chairman
Things that make you go "hmmmm"...
Oh yeah, that "rule of law" thing ...
Putting Dollar Signs in Front Of The AGW Hoax
Moving toward a 60 vote majority?
Do As I Say ....
 
 
QandO Newsroom

Newsroom Home Page

US News

US National News
Politics
Business
Science
Technology
Health
Entertainment
Sports
Opinion/Editorial

International News

Top World New
Iraq News
Mideast Conflict

Blogging

Blogpulse Daily Highlights
Daypop Top 40 Links

Regional

Regional News

Publications

News Publications

 
The pace of innovation: more gasoline on the flames
Posted by: Billy Hollis on Tuesday, January 08, 2008

Via Geekpress, I saw this article in The Economist on the growing use of genetic/evolutionary algorithms in innovation and design. One of the consequences of growing computing power is the feasibility of generating improvements through what you might think of as a massive trial-and-error approach. Random variations are introduced into designs, and the results are measured against some metric to see which ones do best. Those best variations are then "cross-bred" with other good variations to see what comes out.

The result can sometimes be dramatic improvement over anything a human designer can come up with. For example:
At the University of Sydney, in Australia, Steve Manos used an evolutionary algorithm to come up with novel patterns in a type of optical fibre that has air holes shot through its length. Normally, these holes are arranged in a hexagonal pattern, but the algorithm generated a bizarre flower-like pattern of holes that no human would have thought of trying. It doubled the fibre's bandwidth.
When I think about the application of this technology, plus the real genetic manipulation going on in biology, and the availability of information on all kinds of innovative ideas from search engines, I think there's a lot of possible cross-reinforcement. Innovation has been accelerating throughout my entire lifetime, and it shows no signs of stopping that acceleration. The very pace of innovation picks up every year.

It's always been hard to predict future innovation, but when we're into realms where it's not even people doing the work, the results are literally impossible to predict. We get situations like this one from the article:
His team at Stanford developed a Wi-Fi antenna for a client who did not want to pay a patent-licence fee to Cisco Systems. The team fed the algorithm as much data as they could from the Cisco patent and told the software to design around it. It succeeded in doing so. The result is a design that does not infringe Cisco's patent—and is more efficient to boot.
So now it is possible in some instances to bypass patents. That should speed some things up. I'm not sure what the net effect is; patent protection is one of the things we've always believed promoted innovation by making it possible to gain returns on the investment required to innovate. Is innovation becoming so cheap and pervasive that this concept no longer applies, or is at least significantly weakened?

I had a few other random thoughts. What if someone uses genetic algorithms to improve the genetic algorithms themselves? Will genetic algorigthms thus become more efficient and flexible? Will our lives someday be managed by a device that uses genetic algorithms to find the best way to satisfy our desires?
 
TrackBacks
Return to Main Blog Page
 
 

Previous Comments to this Post 

Comments
Well that’s a waste of time, EVOLUTION IS BUNK’UM! The Earth is only 6,000 years old....
 
Written By: Joe
URL: http://
The singularity is near!!!

And that’s the main point of a sci-fi story I’m starting. With a couple big twists though...
 
Written By: Keith_Indy
URL: http://asecondhandconjecture.com
It should be noted that applying such algorithms requires the appliers to have a tremendous amount of domain knowledge. Learning machines are still a long way off, especially since Microsoft canned Msoft Bob.
 
Written By: Grimshaw
URL: http://
Is innovation becoming so cheap and pervasive that this concept no longer applies, or is at least significantly weakened?
The only reason they had to use the algorithm was due to the patent, so that patent is very powerful. Anytime a patent makes a competitor do this it is a success. In this case, it succeeded, but in many other cases it may not.

There has been a trend for instance to use design patents to protect the shape of a part in a larger system. That forces people to try to work around the patent, but often the only work around involves added cost.

Also, this technique can only be applied to certain problems. It can’t come up with "original" ideas, like say the laser jet printer. It might be able to help you find out what kind of ink jet nozzle is best though.

Not to mention, once this kind of technology is demonstrated, it will slowly become commonplace. Initially, people seeking patents will run their product through the spin cycle a few times, and make sure to patent the actual best product that comes out. Voila, less chance to use this to work around a patent.

Finally, it will probably end up as an plug-in to your 3D CAD software in the future, and function much like spell check. "Okay, I’ve designed the flange. Now run it through the EvoCheck to see if its the most efficient." At some even more distant date, you could even use it as an expert system and delegate whole design issues for some parts to the machine.

This is probably a good thing. It just created a whole new job: algorithm designer (because you will need a new one for different circumstances) and is part of creative destruction.

Oh,why did I study PoliSci?

 
Written By: Harun
URL: http://
What if someone uses genetic algorithms to improve the genetic algorithms themselves? Will genetic algorigthms thus become more efficient and flexible?
It’s called meta-evolution, and it does work somewhat, but it’s largely still too slow to be currently workable. The design and refinement of a genetic algorithm for a certain problem is largely a black art, and in my personal view, is likely to remain so for quite a while.
Will our lives someday be managed by a device that uses genetic algorithms to find the best way to satisfy our desires?
GA techniques only work if you can measure what the outcome will be given a certain design. There was certainly some simulation of the fiber optic cable you cited that could answer "given air pockets arranged like so, what will the bandwidth be?" To satisfy human desires, you’d need a system that could predict your happiness for different possible situations, and since we’re not very good at doing that for ourselves, even with all our intimate knowledge of our own preferences and experiences, I wouldn’t hold out hope that GAs can do any better.

This also hints at a problem with the clever "EvoChecker" Harun mentions. You wouldn’t just need to provide your design to the checker, but the method you’re using to evaluate designs, as well as specifications about the problem space you’re trying to explore. It seems like it might be easier to just design the widget with a GA in the first place, but maybe someone has a better idea about how to get around these limitations.
 
Written By: Jared
URL: http://
"This is probably a good thing. It just created a whole new job: algorithm designer..."
This job exists. It’s variously called software engineer, electrical engineer, mathematician, etc..
 
Written By: Grimshaw
URL: http://
So, basically, this is an innovation that makes innovation more ... innovative?
 
Written By: Achillea
URL: http://
Hollis:
Will our lives someday be managed by a device that uses genetic algorithms to find the best way to satisfy our desires?
No.
 
Written By: Martin McPhillips
URL: http://mcphillips.blogspot.com/
:-)
Will our lives someday be managed by a device that uses genetic algorithms to find the best way to satisfy our desires?
According to Dale our day to day lives will be managed - but by Sacremento and politicians - certainly not something as intelligent or even as benevolent caring as an algorithm :-)
 
Written By: BIllS
URL: http://bills-opinions.blogspot.com
Genetic algorithms are fun, but they aren’t perfect. Naturally as a good Christian I favor an approach that depends more on intelligent design of experiments. I say that only slightly tongue in cheek. Years ago I was involved with a research group doing resin transfer mold design with GA. We also tried an experimental design approach. The design approach beat optimized GA by almost as much as GA beat unguided trial and error. Perhaps there are better GA algorithms now? Or GA is better for other problem types and scales? Or GA may preform better outside of a well defined and mapped solution space.
patent protection is one of the things we’ve always believed promoted innovation by making it possible to gain returns on the investment required to innovate. Is innovation becoming so cheap and pervasive that this concept no longer applies, or is at least significantly weakened?
Having an idea is relatively cheap. Initial development can also be pretty cheap. However putting them into production often requires a ton of investment which also needs to be protected and recouped in order for that initial cheap idea to be profitable. Thats the trick.
 
Written By: Jeff the Baptist
URL: http://jeffthebaptist.blogspot.com
Billy:

Thanks for your thoughts on genetic algorithms. I hadn’t run across them before in my search for "innovation" but I would have to say that evolutionary approaches to design appear to have a great future. I cross-linked to you post at the Innovators-Network in hopes that some of my readers will visit your site and read your piece.

Anthony Kuhn

 
Written By: Anthony Kuhn
URL: http://blog.innovators-network.org
Any day now, we are about to be visited by intelligent robots from the future.
 
Written By: Jimmy the Dhimmi
URL: www.warning1938alert.ytmnd.com
I read about this sort of thing in Science Fiction books way back in the 1970’s. It is sort of neat to see it coming about in front of my eyes.

What I would like to see is if anyone has ever run one of these programs to come up with a more efficient architecture for solar panels and the solarcells themselves. They work but are notoriously inefficient.
 
Written By: kyleN
URL: http://impudent.blognation.us/blog
Hey Kyle, look here.

The real reason I’m commenting, having nothing to do with GA or its applications in solar energy research, is that your comment reminded me of this, which has lots of potential for enhancing efficient use of solar energy. So you might be interested.

 
Written By: Linda Morgan
URL: http://
Heh, I’m actually working on a genetic algorithm that will optimize itself (though the parts are all hand coded). That’s not the main feature, though. The thing is, there’s probably not a "best" GA. Hopefully, we’ll be able to find operators that are best for specific classes of problems.
 
Written By: Effeminem
URL: http://ethermind.blogspot.com

 
Add Your Comment
  NOTICE: While we don't wish to censor your thoughts, we do blacklist certain terms of profanity or obscenity. This is not to muzzle you, but to ensure that the blog remains work-safe for our readers. If you wish to use profanity, simply insert asterisks (*) where the vowels usually go. Your meaning will still be clear, but our readers will be able to view the blog without worrying that content monitoring will get them in trouble when reading it.
Comments for this entry are closed.
Name:
Email:
URL:
HTML Tools:
Bold Italic Blockquote Hyperlink
Comment:
   
 
Vicious Capitalism

Divider

Buy Dale's Book!
Slackernomics by Dale Franks

Divider

Divider