When an article is entitled New Technology Portends 'Sentient Computers', its going to catch my eye. Not because I think, "OMG! The robots are taking over!", but because its usually some hack job reporter extrapolating some basic technology's ability to an exponential degree. Not this time:
“Landmark recognition software” can correctly identify many popular landmarks on the Web throughout a range of perspectives and scenarios. This new technology “enables computers to quickly and efficiently identify images of more than 50,000 landmarks from all over the world with 80 percent accuracy,” the vendor claims.
This is working, now. A robot programmed with this tech would, 4 times out of 5, be able to instantly know where it was. And the vendor who created this amazing technology?
Wait for it...
Up until now, computers could “see” what we see only in a limited and imperfect way. But suddenly things are becoming clearer. Google claims to have made substantial progress in endowing computers with image recognition capabilities, often referred to as “computer vision,” a task artificial intelligence (AI) researchers have been wrestling with for 35 years.
Google. Why did it have to be Google?
Computers are good at solving complex mathematical problems, sorting through enormous amounts of data quickly, storing massive amounts of information, etc. As Ray Kurzweil pointed out at 22.28 in a 2006 video on the "Roots of the Matrix": “Machines can remember billions of things accurately, they can do logical analysis at extremely high speed… We are not very good actually at logical or analytical thinking. Computers are already much better than us at considering the logical implications of many different factors.”
They have detailed files.
Visual pattern recognition is something AI researchers have struggled with mimicking in computers. And Google is not the only major entity accelerating in this field. Apple Inc. (Nasdaq: AAPL) is also a player. Face recognition is a component of programs such as iPhoto 09, where accuracy is about 50 percent to start with, and the program learns and the hit rate rises as more and more identifications are confirmed or rejected.
Although Google and Apple’s ambitions may appear narrow and of limited interest at the moment, a more advanced version of this software applied to the Web in general could be an enormous advance. When we add to the mix things that computers excel at -- rapid processing and prodigious memory -- we can begin to envision the potential magnitude of this budding tool, and it is like nothing we have yet imagined.
Actually, we have imagined it. It was made into 3 great movies and one shitty one.
Previous Stories:
I've got a secret I've been hiding under my skin
Google = Skynet, Part 4
Jake's JK: This is not a good sign
Uh oh, Google is at it again
Google's Plan for Worldwide Domination
No comments:
Post a Comment