Robotic Nation Evidence

It's too bad that visual recognition can not be put into a JavaScript AI Mind.
 

Vision systems wil most likely develop very quickly, but Manna and Robotic Nation only cover the corporate side of robots. Hobbyist robotics may have a considerable impact. Also, clanking replicators could reduce the cost of production, and therefefore potentialy the cost of entry into an industry.
 

NEWSFLASH!

Scientists and Engineers have been working on vision systems for about 2 decades now. Interestingly, they have made very little progress in making them practical without hooking them up to a supercomputer. Quick -- hold your breath for science to figure out the vision problem and i'll bury your bodies tonight!
 

This comment has been removed by a blog administrator.
 

Yeah, but I'll be able to make progress, because I'm the best.
 

Researchers Identify Brain Cells Used to Categorize Images

 

Intel is working on tera-scale processors for 2016.
 

where's my shovel?
 

I've got the perfect open field for you. I'll bring extra shovels and a bandsaw.

Vision systems will continue at their same sluggish pace for decades to come. There is no pressing need or cry from consumers for anything like this.

I dub thee, "D.O.A."
 

Kurzweil
 

Actualy, intrest in vision systems reasarch will probably increase once it has publicly visible commerical applications. One reason there's such a difference between intrest in graphics research and in other areas is that the whole point of graphics it is to be visible, for poeple to notice it. They see it. They get interested in it. They make art with it. They play games with it. They buy it. They get jobs making it. Vision systems didn't have that advantage, however, I beileive that vision sytems will become workable in many applications long before they reach the point of matching human vision, and that once this occurs, visoin systems will become more prominent, people will become more interested in them, more people will perform research on the subject, and hopefully, the technology will advance faster.
 

i don't have a problem with that perspective as a matter of principle, mdude. I have a HUGE problem however, with a comparison between the development of image synthesis and image analysis hardware. The problems couldn't be more different, both in structure and complexity. Image synthesis is, essentially, a geometric problem. People develop fancy lighting models but all of that is physics which was largely understood 100 years ago. The task of building it into hardware is a matter of what consumer markets are willing to pay for.

Image recognition, on the other hand, is a vastly parallel process requiring many apriori models of the world. Even if we did understand how the brain processes images, complex tasks are NOT well suited to computer hardware. That is not to say that it wont happen but it will be a long time before it is practical (i.e. low cost).

The truth is that computer vision is already needed in many applications. But it doesn't work. So engineers find other ways around it. Look at the next entry in this blog -- the DARPA Grand Challenge. The winning team did the vast majority of their motion planning and driving with a GPS system and one laser. Only certain advanced obstacle tasks were done with full-frame vision hardware and they were hard pressed to make that work well. The entire field of robotics has been at a standstill because of lack of understanding of the vision problem.

... And to think, that Minsky once assigned the understanding of vision to one of his grad. students as a summer project. Hillarious.
 

Kurzweil said he has a product for the blind that can identify objects. He made it by analyzing Google images.