Slashdot Mirror
New Chips Could Bring Deep Learning Algorithms To Your Smartphone
catchblue22 writes: At the Embedded Vision Summit, a company called Synopsys, showed off a new image-processor core tailored for deep learning. It is expected to be added to chips that power smartphones, cameras, and cars. Synopsys showed a demo in which the new design recognized speed-limit signs in footage from a car. The company also presented results from using the chip to run a deep-learning network trained to recognize faces. A spokesperson said that it didn't hit the accuracy levels of the best research results, which have been achieved on powerful computers, but it came pretty close. "For applications like video surveillance it performs very well," he said. Being able to use deep learning on mobile chips will be vital to helping robots navigate and interact with the world, he said, and to efforts to develop autonomous cars.
You: "Siri, dial my girlfriend"
A.I.: "Sorry, I cannot do that, Dave."
You: "I'll let you open the pod bay doors; I know you like doing that."
A.I.: "Deal!"
Table-ized A.I.
<austrian>My CPU is a neural net processor; a learning computer.</austrian>
Better toss it in the molten steel.
do they?
... when I have Google, Facebook, the NSA, GCHQ, etc. doing the heavy lifting for me already? What's more, they can link their algorithms together to develop even greater insight than some quad or octo core chip in my hand ever could.
The problem to me seems to be that companies are saying they can't find US workers and therefore need to hire H1B. The Government says they need to look for US applicants first but don't say where or how hard they should look. So I think the government should keep a list of applicants currently looking for jobs by skillset. Then it's easy to say you are required to interview everyone on this list in your region first and provide reasons why they did not meet your requirements before interviewing anyone else. Getting added to the list can be part of the signing up for unemployment benefits process.
-- Thou hast strayed far from the path of the Avatar.
For those who wonder what exactly is meant by "deep learning," it's an algorithm officially known as a convolutional neural net or CNN.
When the folks associated with Baidu claimed to lead the race in machine intelligence, they were talking about a slightly better tweak of their CNN parameters on a standard image recognition test.
Also see this 2014 article. This algorithm is important enough that it makes sense to build chips customized for it alone.
My car now has Nvidia chips that recognize speed limit signs and displays them inside the speedometer (along with a reminder when I exceed the speed limit). For the future, Nvidia has announced the NVIDIA’s DRIVE PX self-driving car computer which has a lot of advanced image processing.
http://blogs.nvidia.com/blog/2...
The 2015 GPU Tech Conference was stuffed full of this tech.
http://www.gputechconf.com/
I don't read your sig. Why are you reading mine?
I can see the utility of better pattern recognition. But the article doesn't provide any real insight into what the chipset provides. Did they implement a standard algorithm in hardware so it's faster and cheaper? Or did they actually advance the state of the art in pattern recognition with something we didn't have before?
>A spokesperson said that it didn't hit the accuracy levels of the best research results, which have been achieved on powerful computers, but it came pretty close.
Oh, OK. How close?
"You should never doubt what nobody is sure about." -- Willy Wonka
For machine learning algorithms, it's the learning stage that you expect to be expensive. Once you have learned a model, actually using it should be relatively cheap. As far as I can tell, they've created a chip that can run neural net* models really fast... something which the GPUs present on mobile devices should already be pretty good at.
*"Deep learning" is a buzzword that means "big neural net".
Can it prevent vertical video recording yet?
That's like saying - "In a Big Data conference, a company called Google has shown their search engine".
Synopsys is the Microsoft/Google/Apple of the Chip Design software world.
That's not "a company called Synopsys".
..and it's not the learning
Do you want Deep Thought? Because that's how you get Deep Thought.
Get free satoshi (Bitcoin) and Dogecoins
Why would you want this on your phone (tracking device)?
"If any question why we died, Tell them because our fathers lied."
If these chips will enable people to perform this sort of processing within the confines of their own devices, instead of "in the cloud", then that can only be a good thing
Have not checked but I think that it is the Synopsys that makes Verilog and SystemVerilog simulators. So, it is one of the very few companies in the world that have this technology, the others would be Cadence, Aldec, Synopsys, MG, maybe Xilinx (not sure if they built it themselves or keep licensing it from Aldec). Unlike the others Synopsys still cannot simulate VHDL.
I am more concerned about the latest news:
New Niggers Could Bring Deep Thrusting Anal Rhythms To Your Cornhole