now browsing by month
An anonymous reader quotes a report from IEEE Spectrum: Deep learning has a DRAM problem. Systems designed to do difficult things in real time, such as telling a cat from a kid in a car’s backup camera video stream, are continuously shuttling the data that makes up the neural network’s guts from memory to the processor. The problem, according to startup Flex Logix, isn’t a lack of storage for that data; it’s a lack of bandwidth between the processor and memory. Some systems need four or even eight DRAM chips to sling the 100s of gigabits to the processor, which adds a lot of space and consumes considerable power. Flex Logix says that the interconnect technology and tile-based architecture it developed for reconfigurable chips will lead to AI systems that need the bandwidth of only a single DRAM chip and consume one-tenth the power.
Mountain View-based Flex Logix had started to commercialize a new architecture for embedded field programmable gate arrays (eFPGAs). But after some exploration, one of the founders, Cheng C. Wang, realized the technology could speed neural networks. A neural network is made up of connections and “weights” that denote how strong those connections are. A good AI chip needs two things, explains the other founder Geoff Tate. One is a lot of circuits that do the critical “inferencing” computation, called multiply and accumulate. “But what’s even harder is that you have to be very good at bringing in all these weights, so that the multipliers always have the data they need in order to do the math that’s required. [Wang] realized that the technology that we have in the interconnect of our FPGA, he could adapt to make an architecture that was extremely good at loading weights rapidly and efficiently, giving high performance and low power.”
of this story at Slashdot.
The upcoming mission will attempt to plumb the depths and hidden mysteries of our nearest planetary neighbor. …read more
Fitbit’s smartwatch play is helping it pivot to be more of a healthcare platform. Garmin, which is much larger than Fitbit, also reported earnings. …read more
Security researcher David Buchanan has found that Twitter image uploads can be polyglot files, meaning they can be valid simultaneously in multiple formats, such as a .jpg, a .rar archive and a .zip archive. From a report: Using some Python code he wrote, he created a thumbnail image of William Shakespeare overlaid with the words, “Unzip Me” and posted it to Twitter. The .jpg image is also a valid .zip file, so if you download it, you can unzip it and extract the contents, a multipart .rar archive of the text of Shakespeare’s plays. […] Twitter performs some processing on uploaded images, which has the potential to mess with the data. But Buchanan found that his multi-format file survived this process. It may be that image itself (excluding the rather bulky metadata) is light enough not to trigger any compression or post-upload processing.
of this story at Slashdot.
Here’s how to take part in Epic’s next live event, which will bring an end to its Halloween celebration. …read more
Voters can bike, scoot, carpool or hitch a ride to go to cast their ballots. …read more
Troll preserve’s images can be used to distribute code, PDFs and other stuff
A picture turns out to be worth much more than a thousand words, at least on Twitter. For security researcher David Buchanan, it amounts at least 884,000, roughly the number words in the complete works of William Shakespeare.…
Engineers are building a prototype of a robotic factory that will create water, oxygen, and fuel on the surface of Mars. From a report: The year is 2038. After 18 months living and working on the surface of Mars, a crew of six explorers boards a deep-space transport rocket and leaves for Earth. No humans are staying behind, but work goes on without them: Autonomous robots will keep running a mining and chemical-synthesis plant they’d started years before this first crewed mission ever set foot on the planet. The plant produces water, oxygen, and rocket fuel using local resources, and it will methodically build up all the necessary supplies for the next Mars mission, set to arrive in another two years. This robot factory isn’t science fiction: It’s being developed jointly by multiple teams across NASA. One of them is the Swamp Works Lab at NASA’s John F. Kennedy Space Center, in Florida, where I am a team lead. Officially, it’s known as an in situ resource utilization (ISRU) system, but we like to call it a dust-to-thrust factory, because it turns simple dust into rocket fuel. This technology will one day allow humans to live and work on Mars — and return to Earth to tell the story. But why synthesize stuff on Mars instead of just shipping it there from Earth? NASA invokes the “gear-ratio problem.” By some estimates, to ship a single kilogram of fuel from Earth to Mars, today’s rockets need to burn 225 kilograms of fuel in transit — launching into low Earth orbit, shooting off toward Mars, slowing down to get into Mars orbit, and finally slowing to a safe landing on the surface of Mars. We’d start with 226 kg and end with 1 kg, which makes for a 226:1 gear ratio. And the ratio stays the same no matter what we ship. We would need 225 tons of fuel to send a ton of water, a ton of oxygen, or a ton of machinery. The only way to get around that harsh arithmetic is by making our water, oxygen, and fuel on-site. Different research and engineering groups at NASA have been working on different parts of this problem. More recently, our Swamp Works team began integrating many separate working modules in order to demonstrate the entire closed-loop system. It’s still just a prototype, but it shows all the pieces that are necessary to make our dust-to-thrust factory a reality. And although the long-term plan is going to Mars, as an intermediate step NASA is focusing its attention on the moon. Most of the equipment will be tried out and fine-tuned on the lunar surface first, helping to reduce the risk over sending it all straight to Mars.
of this story at Slashdot.
When you create a Thing for the Internet of Things, you’ve made a little computer that does a simple job and which probably has a minimal interface. But minimal interfaces leave little room for configuration, such as entering WiFi details. Perhaps if you made the Thing yourself you’ve hard-coded your WiFi credentials in your code, but that hardly translates to multiple instances. So, how to put end-user WiFi credentials easily on more than one Thing? Perhaps [Rob Dobson] has the answer with his technique of sending them as a sequence of audible tones.
Of course, this is nothing new, as any owner of an 8-bit machine that had a cassette interface will tell you. And on the face of it it’s much easier than those awkward impromptu hotspots with a web interface to which you connect and pass on your credentials. But while we quite like the convenience, we can’t help wondering whether expressing the credentials in audible free space might be a bit too insecure for many readers. The technique however remains valid, and we’re sure that other less sensitive applications might be found for it. Meanwhile we hope he hasn’t inadvertently shared his WiFi password in the video below the break.
As it turns out, crime pays incredibly well for some
The infamous GandCrab malware infection has netted its operators an estimated nine-figure payout from targeting large, high-value corporate systems.…