Profile cover photo
Profile photo
Gary Leeming
560 followers
560 followers
About
Gary's posts

Post has attachment
Check out this video on YouTube:

Post has shared content
"...technique could be used as a method for machines to talk to each other while evading human detection."
> Deep neural networks (DNNs) have recently been achieving state-of-the-art performance on a variety of pattern-recognition tasks, most notably visual classification problems. Given that DNNs are now able to classify objects in images with near-human-level performance, questions naturally arise as to what differences remain between computer and human vision. A recent study revealed that changing an image (e.g. of a lion) in a way imperceptible to humans can cause a DNN to label the image as something else entirely (e.g. mislabeling a lion a library). Here we show a related result: it is easy to produce images that are completely unrecognizable to humans, but that state-of-theart DNNs believe to be recognizable objects with 99.99% confidence (e.g. labeling with certainty that white noise static is a lion). Specifically, we take convolutional neural networks trained to perform well on either the ImageNet or MNIST datasets and then find images with evolutionary algorithms or gradient ascent that DNNs label with high confidence as belonging to each dataset class. It is possible to produce images totally unrecognizable to human eyes that DNNs believe with near certainty are familiar objects. Our results shed light on interesting differences between human vision and current DNNs, and raise questions about the generality of DNN computer vision.

Full paper: http://arxiv.org/abs/1412.1897

via +Bruno Gonçalves 

// This is clever. It's funny that they draw a lesson about how easy it is to fool these DNNs, since the same technique could be used as a method for machines to talk to each other while evading human detection.
Photo

Post has attachment

Post has shared content
It's amazing to me that CGI, the company that brought us the healthcare.gov fiasco, has the chutzpah to protest the ways that the UK GDS has cut the cost of government IT services while making those services far better for citizens. These guys should be laughed out of the room. The IT vendors need to work to improve their own services, not spread FUD about those who are showing how to do it better!

Post has shared content
IBM has built a new brain-like computer chip that's the size of a postage stamp, boasts 1 million neurons and uses as little electricity as a hearing aid. http://ibm.co/1pEyAfr
Photo

Post has shared content
Twelve Tomorrows Takes Tangible Science and Imagines the Future
http://adafru.it/dMX

Science Fiction Authors have a long history of forecasting developments in technology. Twelve Tomorrows, MIT Technology Review‘s Science Fiction Anthology, features some of today’s best writers in the genre. It’s coming out soon (sometime this August). Sign-up here to be notified when its out!

Inspired by the real-life breakthroughs covered in the pages of MIT Technology Review, renowned writers Pat Cadigan, Cory Doctorow, and Christopher Brown join the hottest emerging authors from around the world to envision the future of the Internet, biotechnology, computing, and more.

Read More
http://adafru.it/dMX
Photo

Post has attachment
Browsing for superhero pictures for Thea's birthday party
Photo

Post has attachment

Post has attachment
Wait while more posts are being loaded