Group items tagged

Strange and exotic cityscapes. Desolate wilderness areas. Dogs that look like wookies. Flickr, one of the world's largest photo sharing services, sees it all. And, now, Flickr's image recognition technology can categorize more than 11 billion photos like these. And it does it automatically. It's called "Magic View." Magical deep learning! Buzzword attack!

You're forgetting power usage. GPUs need 1000 hamster wheels worth of power while FPGAs can run on a potato. Since space applications are highly power limited, putting any kind of GPU monster in orbit or on a rover is failed idea from the start. Also in FPGAs if a gate burns out from radiation you can just reprogram around it. Looking for seals offline in high res images is indeed definitely a GPU task.... for now.

The discussion of how to make FPGA hardware acceleration solutions easier to use for the 'layman' is starting btw http://reconfigurablecomputing4themasses.net/.

An interesting research centre in California! Focus areas: Business Services Electronic Materials, Devices, & Systems Information & Communication Technologies Biomedical Systems Cleantech

XEROX PARC was definitely heavily involved in computer development: eg. mouse, GUI, ethernet, OO programming, all came out of PARC, and all that without focusing on computers but printers...

aaah its the XEROX centre, didn't know. Yep they made the mouse and then handed it over nicely to Apple after IBM thought it was useless

It seems we cannot add a publication automatically (eg with bibtex) if it's not on pubmed. Bit useless for non medical sciences...

what a pity ...

no registration, you just send an email to post@posterous.com and automatically creates a blog with your link, check my test one e.g. http://nikolis1.posterous.com/ Read more: "Posterous - The place to post everything. Just email us. Dead simple blog by email." - http://posterous.com/#ixzz0EYUFIaWX&A

bullshit

KI tool that automatically organizes files, emails, etc.

interesting story with many juicy details on how they proceed ... (similarly interesting nickname for the "operation" chosen by our british friends) "The spies used the IP addresses they had associated with the engineers as search terms to sift through their surveillance troves, and were quickly able to find what they needed to confirm the employees' identities and target them individually with malware. The confirmation came in the form of Google, Yahoo, and LinkedIn "cookies," tiny unique files that are automatically placed on computers to identify and sometimes track people browsing the Internet, often for advertising purposes. GCHQ maintains a huge repository named MUTANT BROTH that stores billions of these intercepted cookies, which it uses to correlate with IP addresses to determine the identity of a person. GCHQ refers to cookies internally as "target detection identifiers." Top-secret GCHQ documents name three male Belgacom engineers who were identified as targets to attack. The Intercept has confirmed the identities of the men, and contacted each of them prior to the publication of this story; all three declined comment and requested that their identities not be disclosed. GCHQ monitored the browsing habits of the engineers, and geared up to enter the most important and sensitive phase of the secret operation. The agency planned to perform a so-called "Quantum Insert" attack, which involves redirecting people targeted for surveillance to a malicious website that infects their computers with malware at a lightning pace. In this case, the documents indicate that GCHQ set up a malicious page that looked like LinkedIn to trick the Belgacom engineers. (The NSA also uses Quantum Inserts to target people, as The Intercept has previously reported.) A GCHQ document reviewing operations conducted between January and March 2011 noted that the hack on Belgacom was successful, and stated that the agency had obtained access to the company's

I knew I wasn't using TOR often enough...

Cool! It seems that after all it is best to restrict employees' internet access only to work-critical areas... @Paul TOR works on network level, so it would not help here much as cookies (application level) were exploited.

love this one ... it seems to take physicist to explain to the AI crowd what they are actually doing ... Deep learning is a broad set of techniques that uses multiple layers of representation to automatically learn relevant features directly from structured data. Recently, such techniques have yielded record-breaking results on a diverse set of difficult machine learning tasks in computer vision, speech recognition, and natural language processing. Despite the enormous success of deep learning, relatively little is understood theoretically about why these techniques are so successful at feature learning and compression. Here, we show that deep learning is intimately related to one of the most important and successful techniques in theoretical physics, the renormalization group (RG). RG is an iterative coarse-graining scheme that allows for the extraction of relevant features (i.e. operators) as a physical system is examined at different length scales. We construct an exact mapping from the variational renormalization group, first introduced by Kadanoff, and deep learning architectures based on Restricted Boltzmann Machines (RBMs). We illustrate these ideas using the nearest-neighbor Ising Model in one and two-dimensions. Our results suggests that deep learning algorithms may be employing a generalized RG-like scheme to learn relevant features from data.

asm.js: it's really fast, backwards compatible, and now in the release version of Firefox. "By doing this the result is highly optimized and can be converted directly from this Asm.js syntax directly into ASSEMBLY without having to interpret it, as one would normally have to do with JavaScript" Here is great introduction to asm.js and C++ to js compilation: http://ejohn.org/blog/asmjs-javascript-compile-target

We discussed a few times about whether it is possible to determine the quality of a paper by extracting visual features from the paper and then learning a mapping to a measure of quality such as the number of citations etc. This paper circulated at CVPR 2010, and does exactly that, mapping visual features to estimate whether it has been accepted for the main conference or the workshops.

We made it to Slashdot's frontpage !!! :)

Congratulations, gentlemen!

Zivile can you order this for me?

I will get one as soon as they put it in a spherical dome like that of Darth Vader's...

@Johannes: The Japanese version will definitely have a toilet included, of course with health analysis, automatic appointment making with the doctors, life time estimation, diet planning, etc.

I'm sure Francesco will love it... perhaps of use for ACT's Google calendar/docs?

is there an easy way (easy for Francesco I mean) to retrieve the citation number of papers in google scholar automatically (e.g. for act papers)?

It seems like google scholar is not supported yet.

make you wonder when the FLPP will actually lead to something flying !

I did not even know that the airforce continued the X37 project when it was cancelled at NASA ... interesting stuff indeed ...

One more nice citizen scientist example as I try to find for space since some time. Any good ideas this mit inspire you to?

why don't you programme one and get rich?

Would love to but this requires *TIME* :-( Automatic image-based mushroom recognition... perhaps could do as a FP7 study... it may save lives actually!

Via Slashdot. Mandatory implementation

impressive: use of human body for communication between electronic devices

i saw once a similar system that would automatically transfer your business card when you shake hands ... funny ..

A handy feature that allows linking of over 300 online apps. For example, generating Trello cards from calendar events automatically (and vice versa).

Dataset automatically constructed from public images uploaded by users on FB/Instagram with hashtags used as labels! They refer to this as 'weakly supervised learning'. Then neural net fine-tuned for ImageNet and achieved record 85.4% accuracy.