Thursday, December 22, 2016

Facebook, hmmm

  1. jfieb

    jfiebWell-Known Member


    I will store some important stuff here.

    They should just buy Sensory Inc?


    Applied Research Scientist, Speech Recognition Language Modeling
    (Menlo Park, CA)
    Facebook's mission is to give people the power to share, and make the world more open and connected. Through our growing family of apps and services, we're building a different kind of company that helps billions of people around the world connect and share what matters most to them. Whether we're creating new products or helping a small business expand its reach, people at Facebook are builders at heart. Our global teams are constantly iterating, solving problems, and working together to make the world more open and accessible. Connecting the world takes every one of us—and we're just getting started.
    Facebook is seeking a Speech Recognition Research Scientist to join our research team in Menlo Park. The ideal candidate will have research experience developing speech recognition systems in different languages. Individuals in this role should be experts in language modeling and have experience working on vast quantities of data, advanced recognition systems and perform rigorous performance evaluations. The candidate will help Facebook conduct research that support naturally spoken input in more than 70 languages.
    Responsibilities
    • Create language models from large corpora of text data in different languages using Hadoop/Hive
    • Work closely with other speech recognition experts and researchers on implementing algorithms that power user and developer-facing products
    • Be responsible for measuring and optimizing the quality of your algorithms
    Minimum Qualifications
    • Strong desire to build beautiful, expressive products that delight users in any language
    • M.S. or Ph.D. in Computer Science, Electrical Engineering, Speech/Language Processing or Machine Learning
    • Experience with scripting languages such as Perl, Python, PHP, and shell scripts
    • Experience in C/C++
    • Experience with Hadoop/Hbase/Pig or Mapreduce/Sawzall/Bigtable a plus
    • Fluency in at least 2 natural languages is a plus
    Last edited: 1 minute ago
  2. jfieb

    jfiebWell-Known Member


    and this one




    Typically, neural networks run on large numbers of computer servers packed into data centers on the other side of the Internet—they don’t work unless your phone is online—but with its new app, Facebook takes a different approach. The Picasso filter is driven by a neural network efficient enough to run on the phone itself. “We perceive the world in real-time,” Mehanna says. “Why wouldn’t you want the same thing from your AI?”

    the link

    https://www.wired.com/2016/11/fb-3/

    Really, I said to myself, holy smokes, YES! THis will be better than 10 axis fusion for indoor location..


    Here is some more...


    Already available in Ireland and due soon here in the States, this new Facebook app is another sign that deep neural networks will push beyond the data center and onto phones, cameras, and various other devices spread across the so-called Internet of Things. As Rick points out its just like INTC/Altera only on the other end.....


    Commentary.... ALready we have the BIG dogs of the FPGA world talking bout cloud neural networks...and Facebook talking up Todd Mozers vision...we need more intelligence ON THE DEVICE. And the article only gets better....

    Yes, these tools can operate without an Internet connection. And that points to a future where our smartphone apps can perform a much wider range of tasks while offline. But it also shows we’re moving towards technology that can handle more complex AI tasks with less delay. Ultimately, if you can complete a task without sending a bunch of data across the wire, it will happen quicker.

    Training in the cloud and inference ON THE DEVICE.....

    Facebook’s app doesn’t train its neural networks on your smartphone. That still happens on servers in the data center. But the phone does execute the neural net—without calling back to the data enter. That may seem like a small thing, but building a deep neural net that can so quickly execute on a phone—which offers limited processing power and memory—is no simple task.


    Am blown away this am....best find in a loooong time for 

    As companies like Facebook and Google continue to push neural networks onto smartphones, phone makers will start building hardware into these devices that can run neural networks with even greater speed. 



    QUIK, with is multi yr focus on low power, with its FPGA IP, this starts to flesh it out. QUIK has every bit and piece needed to create a ubiquitous NNLE.( neural network learning engine). 

    A Picasso photo filter won’t change your life. But this one points to big changes in the years to come.


    QUIK's IP just took a jump in value, likeXilinx.


    From the other recent reading, DSPs can do the training very well, FPGAs are vying for the inference part of AI.

    There should be a LOT of interesting talk with Dr Saxe and team,

    Please read this one over every couple days for incremental understanding of what it means....

    I have waited a LOOOOONG time to read something so profound that is good.

    jfieb, happy today



    am so blown away by implications I have to log off and do other stuff and come back l8er to look at it again and see if I got it right.

    Sorry bout the underlining, once I use it for a phrase it does NOT stop.

    Please make comments, ask questions etc. This IS HUGE.?


    Facebook has a LOT of jobs related to this...........

    Last edited: A moment ago
  3. jfieb

    jfiebWell-Known Member

    New

    https://techcrunch.com/2016/11/08/shining-light-on-facebooks-ai-strategy/

    Shining light on Facebook’s AI strategy

    a few snips

    Caffe2Go won’t remain limited to Style Transfer — it holds the key to deploying convolutional neural nets across Facebook’s suite of mobile apps.

    “For anything we build on the server, we now have a vehicle to ship it to mobile devices,” noted Schroepfer.



    Bringing machine intelligence to the smartphone is just step one. For the time being, the world still doesn’t have an answer for training neural networks on mobile devices. The prospect coupled with Facebook’s long-term roadmap make a compelling argument for a future where the average person could design and train a custom neural net on their own smartphone for daily use.


    Regardless of the barriers, Facebook has little choice in prioritizing AI as its competitors pour billions into beating the company to the next great breakthrough — though the mere fact that everyone is all in on the race is perhaps what makes it so interesting.

    While Google first popularized algorithmic search, and Snapchat is now making augmented reality mainstream, it’s Facebook delivering artificial intelligence to the masses. AI isn’t necessarily about communicating with a conscious computer. It’s about sifting through mountains of data to make sense of the chaos. And no one has more data about us than Facebook.


    consider that in the cloud Inference is done on FPGAs. Some Soc makers will have to try it out for themselves.

    Facebook should just buy Sensory and be where it needs to be much much sooner and with really good stuff?

    Dr. Saxe you got this covered for us? :) thanks in advance

No comments:

Post a Comment