Slashdot Mirror

← Back to Stories (view on slashdot.org)

3D-Printed Deep Learning Neural Network Uses Light Instead of Electrons (newatlas.com)

Posted by BeauHD on from the proof-of-concept dept.

Matt Kennedy from New Atlas reports of an all-optical Diffractive Deep Neural Network (D2NN) architecture that uses light diffracted through numerous plates instead of electrons. It was developed by Dr. Aydogan Ozcan and his team of researchers at the Chancellor's Professor of electrical and computer engineering at UCLA. From the report: The setup uses 3D-printed translucent sheets, each with thousands of raised pixels, which deflect light through each panel in order to perform set tasks. By the way, these tasks are performed without the use of any power, except for the input light beam. The UCLA team's all-optical deep neural network -- which looks like the guts of a solid gold car battery -- literally operates at the speed of light, and will find applications in image analysis, feature detection and object classification. Researchers on the team also envisage possibilities for D2NN architectures performing specialized tasks in cameras. Perhaps your next DSLR might identify your subjects on the fly and post the tagged image to your Facebook timeline. For now though, this is a proof of concept, but it shines a light on some unique opportunities for the machine learning industry. The research has been published in the journal Science.

76 comments

  1. Perhaps by Daralantan · · Score: 2

    Perhaps your next DSLR might identify your subjects on the fly and post the tagged image to your Facebook timeline.

    No thank you.

    1. Re:Perhaps by Anonymous Coward · · Score: 1, Insightful

      These people have a fundamentally different value system, so they just don't get it.

    2. Re: Perhaps by volodymyrbiryuk · · Score: 3, Interesting

      Exactly this. They have bleeding edge technology on their hands and want to use it to post on FB? Get the fuck out of here.

      --
      sudo rm -r -f --no-preserve-root /
    3. Re: Perhaps by Anonymous Coward · · Score: 1

      It's a silly marketing tactic they're employing by saying that. They're just trying to make it relevant and useful to John Q Idiot.

    4. Re: Perhaps by Anonymous Coward · · Score: 0

      which one - there are 7.5B of them and I think the movie depicting the situation in a bit exaggerated way that is: idiocracy' is not really exaggerating after all.

    5. Re: Perhaps by Potor · · Score: 1

      No - that explanation is to make the tech sound innocuous. The real use is government and police surveillance.

    6. Re: Perhaps by mikael · · Score: 1

      Analytics is the biggest market for big data processing. All those juicy camera photographs sent to Facebook - Identifying people, places, associations, parties, events, whatever...

      --
      Vintage computer adverts: http://www.vintageadbrowser.com/computers-and-software-ads
    7. Re: Perhaps by Anonymous Coward · · Score: 0

      Yup. Because currently, it isn't. That would be because the technology doesn't actually exist. At least they had the decency to acknowledge it's just a proof of concept. If posting to Facebook will be the extent of its utility, it should probably stay that way.

  2. Product without purpose shows great promise by skovnymfe · · Score: 0

    "If we build it, they will come" and then they'll buy the patent for billions of dollars. We're going to be RICH guys. RICH!

    Meanwhile, in the real world, no one really cares.

    1. Re:Product without purpose shows great promise by Anonymous Coward · · Score: 0

      Meanwhile, in the real world, no one really cares.

      Speak for yourself. Instantaneous pattern recognition has the potential to be a game changer.

      If this can scale to complex patterns, you could have real-time identification of objects in an image (or video) at essentially zero processing cost.

      Instead of having an algorithm churning through data, eating up your batteries and heating up your device, you'd simply have an array of sensors that, when hit by light in a specific pattern, could immediately indicate what objects are in a scene, and where.

      This could see applications in AR, automated vehicles (think of a car that can instantly detect a person walking into traffic, with virtually no power cost), factory automation...

      That is all, of course, IF it scales. That might be a pretty big if.

    2. Re:Product without purpose shows great promise by CptLoRes · · Score: 4, Insightful

      The concept of learning algorithms was developed in the 60's and Deep Learning was coined in the late 80's. Either had hardly any practical application at the time they where introduced. Fast forward and companies are now poring billions into this kind of technology. Point being nobody cared about most things, before they suddenly became the next big thing.

    3. Re: Product without purpose shows great promise by ceoyoyo · · Score: 1

      Iâ(TM)m sure it will scale. But will it scale better than an ASIC?

      You could probably design some kind of adaptive optical element that could change its local diffractive properties on command, which would let you train something like this, or adapt it to different situations. But at the moment, and probably for some time, that sounds quite a bit more complicated than just reprogramming an FPGA or loading different values into one of googleâ(TM)s TPUs or tweaking some resistances in an analog (electrical) computer.

      You might be able to come up with some very specialised applications where a fixed filter thatâ(TM)s faster than four or five analog elements in series is useful, but Iâ(TM)m having troubling thinking of one.

      You can do a Fourier transform with light too, but nobody ever does because itâ(TM)s much easier to use a dsp.

    4. Re:Product without purpose shows great promise by mikael · · Score: 2

      There wasn't enough computing power back in the late 1980's. For a desktop PC with an 66MHz x86, graphics boards with i860's, TMS34020's and some TMS320x0's were the most performance you could get. Even then, those boards were around $1200, and you were lucky to get a C compiler let alone Fortran or C++. Even then rendering a single frame of the Mandelbrot set would still take minutes.

      Today, you can buy a PC or server with custom DNN hardware, able to process Terabytes of image data.

      --
      Vintage computer adverts: http://www.vintageadbrowser.com/computers-and-software-ads
    5. Re:Product without purpose shows great promise by captbollocks · · Score: 1

      Either had hardly any practical application at the time they where introduced. Fast forward and companies are now poring billions into this kind of technology. Point being nobody cared about most things, before they suddenly became the next big thing.

      Bullshit. There just wasn't enough computing power nor big enough data sets to do much with the technology.

      If someone had invented an analogue analogue to the neuron as a component may be in IC form factor, the world would be a very different place today.

    6. Re: Product without purpose shows great promise by phantomfive · · Score: 2

      Check out the perceptron, it is almost exactly what you describe, made a long time ago

      --
      "First they came for the slanderers and i said nothing."
    7. Re: Product without purpose shows great promise by Anonymous Coward · · Score: 0

      I don't recall deep learning being used in the late 80s, and I started in the area in 1990. Multilayer neural networks certainly aren't new, though, as we used them back then as you had to given that wide and shallow wasn't plausible on processors back then. Even then, to do multilayer efficiently , you often needed to use custom hardware (FPGA) , and use pipelining for good overall throughput. And people are still doing both of these.

    8. Re:Product without purpose shows great promise by cwsumner · · Score: 1

      The concept of learning algorithms was developed in the 60's and Deep Learning was coined in the late 80's. Either had hardly any practical application at the time they where introduced. Fast forward and companies are now poring billions into this kind of technology. Point being nobody cared about most things, before they suddenly became the next big thing.

      Making a machine learning network does not require any of these new things. The ones first used in the 1960's were much slower, but just as good at learning.
      https://www.youtube.com/watch?...

      Also, Neural networks (that are not really AI at all) only need to be used for the learning. Once done they can be reduced to only the active logic and constructed as code or hard circuits. But they are useful for detecting logic that no one had recognized before that.

    9. Re: Product without purpose shows great promise by captbollocks · · Score: 1

      I don't think there was an analogue hardware implementation of the perceptron in the 80s or 90s even?

    10. Re:Product without purpose shows great promise by InterGuru · · Score: 1

      Back in the 60's lasers were considered a solution without a problem.

    11. Re: Product without purpose shows great promise by phantomfive · · Score: 1

      I'm pretty sure the original was https://en.m.wikipedia.org/wik... but maybe the particular aspect you are looking for was not.

      --
      "First they came for the slanderers and i said nothing."
    12. Re:Product without purpose shows great promise by Anonymous Coward · · Score: 0

      > Bullshit. There just wasn't enough computing power nor big enough data sets to do much with the technology.

      yep. in the early 1990s we were experimenting with nn and wavelets to correlate startigraphy/seismic with oil/gas yields. the hurdle to production was performance.
      our hardware was cray/pyramid, running a whopping gig or so of ram.

  3. Avoid the paywall by Anonymous Coward · · Score: 5, Informative

    and the "magazine" article, get the research paper straight from the horse's mouth for free.

    http://innovate.ee.ucla.edu/wp...

  4. Almost all good by Anonymous Coward · · Score: 0

    Then you mentioned the goddamn thing interfacing with bullshit like facebook and I immediately wanted to smash it with a hammer into pieces.

    Do your civic duty, kill the rich

    1. Re:Almost all good by Anonymous Coward · · Score: 1

      Kill the rich? Starting with you?

      Oh, you don't think you're rich. That's funny. You'll be classified that way once you're no longer useful.

      Remember the kulaks.

  5. It doesn't learn by viperidaenz · · Score: 2

    They develop the network, then 3D print it. Once it's printed it performs a static task.

    1. Re:It doesn't learn by bickerdyke · · Score: 1

      as regular "Machine Learning" does during normal operations if you don't do training cycles.

      --
      bickerdyke
    2. Re:It doesn't learn by Hentes · · Score: 1

      Yep. From the paper:

      We introduce an all-optical deep learning framework, where the neural network is physically formed by multiple layers of diffractive surfaces that work in collaboration to optically perform an arbitrary function that the network can statistically learn. While the inference/prediction of the physical network is all-optical, the learning part that leads to its design is done through a computer.

      Now this is still a really cool tech with a number of potential applications, I don't understand why the authors felt the need to give it such a misleading name.

    3. Re:It doesn't learn by Viol8 · · Score: 1

      Except you with a normal network you have the option to continue training it if you want to and even if you don't , downloading a new one is simple. When this needs to be updated it simply goes in the bin and a new one has to be fabricated. Not terribly efficient.

    4. Re:It doesn't learn by Theaetetus · · Score: 1

      Except you with a normal network you have the option to continue training it if you want to and even if you don't , downloading a new one is simple. When this needs to be updated it simply goes in the bin and a new one has to be fabricated. Not terribly efficient.

      ... for learning new tasks, but potentially very efficient for utilizing already learned tasks. Ignore the click-bait about machine learning, and instead focus on the fact that they created an OCR system that uses no power for processing other than the initial light input.

    5. Re:It doesn't learn by mysidia · · Score: 1

      When this needs to be updated it simply goes in the bin and a new one has to be fabricated. Not terribly efficient.

      You don't mass print it until you are delivering --- it is the same way when you Lithography is being used to manufacture CPUs and other integrated circuits: Once produced there are no further updates without building a new one.

      This is an Optical ASIC that takes a complex input and spits out a result requiring no additional use
      of electrical power, and no stateful units such as Memory or transistors that could fail or respond to EM interference like solar flares to induce bit errors ---- so it is about as efficient as is possible!

    6. Re:It doesn't learn by viperidaenz · · Score: 1

      No EM interference to worry about..
      Just dust, moisture, temperature changes, vibration. Basically anything that can interfere with the optical path or bend/misalign the filters.

    7. Re:It doesn't learn by mysidia · · Score: 1

      Just dust, moisture, temperature changes, vibration.

      It would take some rather extreme conditions to bend layered sheets of plastic, especially inside a decent enclosure.
      The interface points could be a weak spot, but there are a number of ways they could protected, such as putting the detectors flush with the media and filling the space with epoxy.

    8. Re:It doesn't learn by viperidaenz · · Score: 1

      dust and moisture interfere with the light.
      Dust gets in the way, moisture changes the refractive index.

      Temperature makes things expand and contract. Different materials do s at different rates.

    9. Re:It doesn't learn by mysidia · · Score: 1

      The medium is a 3D printed plastic plate --- It seems you are totally ignorant to the basic properties of plastics.

      Dust gets in the way, moisture changes the refractive index.

      Dust and Moisture cannot penetrate plastics. It's trivial to make sure the environment is clean when first 3D printing the media. You can bury a plastic plate/block in a pile of sand or a mound of dirt, dig it up, and spray it off --- not a single spec of dirt or sand or drop of water will penetrate into the material. A plastic plate/block can be submerged to a significant depth with no water or gas able to penetrate through the layers, let-alone dust.

      If during manufacture the detectors and projection units are epoxied to the plate, there will be a continuous layer of plastics, and a Zero percent chance of any dust or moisture ever entering in the first place.

      Temperature makes things expand and contract. Different materials do s at different rates.

      Plastics experience very little or no thermal expansion, some can survive very high temperatures with no change in state. Also, the entire printed sheets are plastic with different bumps printed in different locations on different layers: the plates/units are not a combination of differing kinds of materials.

    10. Re:It doesn't learn by viperidaenz · · Score: 1

      Plastic expands and contracts with temperature just like everything else Having many separate layers that need to be precisely aligned means that the expansion of one plate is going to have an impact on the entire system.

      If you've every done 3D printing, you'll know that PLA, ABS, Nylon, etc, all shrink when they cool down, leading to potential warping and having to print your ABS design 1.5 to 2.5% larger than you want it to be when it's cooled down. That's a 2.5% shrinkage with a 200 degree temperature difference.

      Plastic has in general a higher thermal expansion that most metals.
      Have a look for yourself: https://www.engineeringtoolbox... the higher the number, the more it expands. Glass is around 4, plastics range between 40 and 120.

      If you want to 3D print with nylon, you need to make sure you dry out the filament first, as it's quite hygroscopic and will absorb moisture from the air.

      You seem to be totally ignorant as to how refraction works. How much the light bends depends on the ratio between the refractive index of lens and the medium it is in. Water and air have different indexes. Water on the surface of a lens causes distortion, regardless of the material the lens is made from.

      Moisture can penetrate plastics too. controlling moisture content in plastic is a huge part of reliable production.

  6. Cool! by bickerdyke · · Score: 2

    That sounds like one of the coolest magic tricks ever performed.

    Set up a lamp shining on a screen and have someone paint a number on a glass plate. Put it in front of the lamp to project it on the screen. Add a stack of plates and instead of the number you will see a bright spot in a field on screen representing the number. And that without any hidden active equipment making a "decision" of any kind. Just a combination of refracting patterns.

    --
    bickerdyke
    1. Re:Cool! by Anonymous Coward · · Score: 0

      With light it is easier and faster to recognize patterns!
      Just overlap 2 images and get statistics of the coincidences if they are high...

  7. Well, nice but bulky by Chatterton · · Score: 1

    As long as it take more space and more weight than an IC performing an equivalent task, this will stay a nice research subject.
    We replaced hardware radio receptor by software ones, I don't see why we would replace software neural network by hardware ones... Plus, they miss the 'plasticity' of software ones and a scratch would make it 'dead' :(

    1. Re:Well, nice but bulky by bickerdyke · · Score: 1

      As long as it take more space and more weight than an IC performing an equivalent task

      But it doesn't use power and should have a high radiation/EMP robustness. And is literally as fast as lightning. Any of these factory may mitigate space or weight

      I already wrote in a different post that it's really cool because it mixes hi-tech with absolute low-tech.

      --
      bickerdyke
    2. Re:Well, nice but bulky by ShanghaiBill · · Score: 1

      I don't see why we would replace software neural network by hardware ones.

      1. Speed (it is inherently parallel).
      2. Energy consumption.

    3. Re:Well, nice but bulky by Anonymous Coward · · Score: 1

      Using light, it is much faster than even a purpose-built IC. You train some network to 'perfection', then you have very low production cost in the form of perforated sheets. Useful, if you have a static but high-speed task. Such as recognising stuff on a production line. If the task ever changes, you retrain on a computer and replace the perforated sheets.

    4. Re: Well, nice but bulky by Anonymous Coward · · Score: 0

      How will this affect important things to slashdot readers, like notches and bezels on phones?

    5. Re:Well, nice but bulky by fish_in_the_c · · Score: 1

      I'm guessing the only value of such a thing, beyond research, would be if it is super fast for specialized applications.
      Say you wanted to implement voice recognition with specific keyword capture across all the conversations in the united states at the same time in real time. Something like this 'might' be useful , assuming you can't perform the same task on normal hardware, which i wouldn't know without testing.

      --
      âoeTolerance applies only to persons, but never to truth. Intolerance applies only to truth, but never to persons.
    6. Re:Well, nice but bulky by mysidia · · Score: 1

      As long as it take more space and more weight than an IC performing an equivalent task, this will stay a nice research subject

      Space and weight are mainly only a concern for portable devices, and to a lesser extent, personal electronics.

      Weight isn't that important, and there is PLENTY of cheap land available in the US for industrial applications.
      So imagine you take a picture of your mobile phone and want send it to the cloud to figure out what that picture is, tag all the objects, and categorize it appropriately.

      If your task is common enough, and a neural network gives great results, then the cloud can process millions of images per second through one of these in order to get the result/outcome much faster and at a much lower requirement in terms of total CPU consumption and heat release.....

      The average person might not be able to own a stack of printed neural network modules in their house,
      that may be a waste, but if the processing is more efficient, then I see it being used by central providers with bulk requirements.

    7. Re:Well, nice but bulky by captbollocks · · Score: 1

      ideal for weapon systems that need to identify targets very quickly. Might be able to help an aircraft carrier survive for another 10 minutes in an all-out war with Russia or China.

  8. Can someone confirm... by Anonymous Coward · · Score: 1

    So basically this is like 3D printing a "machine" that is capable of instantaneously evaluating a R^2 -> R^2 function for some degree of precision (specially in the detectors side) and assuming you have already found a good enough NN representing the function.

    Disclaimer: I have not read the paper.

  9. That's nothing by Anonymous Coward · · Score: 0

    Call me when they start using heavy in stead of light !!
    That will be the ultimate game changer.

  10. This is how.. by shyamrox · · Score: 1

    ....the Matrix started.

    --
    I SURVIVED THE GREAT SLASHDOT BLACKOUT OF 2002!
  11. Nope ... by Anonymous Coward · · Score: 0

    Perhaps your next DSLR might identify your subjects on the fly and post the tagged image to your Facebook timeline.

    Avid photographer here ... no fucking way I'd pay for a fucking camera which connects to Facebook, has the capability to connect to Facebook, and has built in knowledge of Facebook. Because the camera itself would be completely fucking un-trustworthy.

    Mark goddamned fucking Zuckerberg can kiss my goddamned ass.

    Enough people, stop expressing everything in terms of Facebook ... many of us don't give a fuck about that stupid thing.

  12. Almost there... by Anonymous Coward · · Score: 0

    A 3d-printed optical neural net? If we use it to mine bitcoin we will have reached peak hipster.

    1. Re:Almost there... by Anonymous Coward · · Score: 0

      I wonder if you could train a NN to attack hash functions via perhaps undiscovered vulnerabilities.

  13. What's old is new again by Eager+Newbie · · Score: 1

    So this is the modern version of punch cards?

    --
    "Your most unhappy customers are your greatest source of learning." Bill Gates Yeah Right!
    1. Re:What's old is new again by Theaetetus · · Score: 1

      So this is the modern version of punch cards?

      More like those old physical video game DRM systems with a colored filter you placed over an obfuscated image to then read the password, but with multiple layers of filters.

  14. Add "VR RGB LED App" on the title by Tsolias · · Score: 1

    and you have the most 2018 article.

  15. 3D printed neural network? by Anonymous Coward · · Score: 1

    Unless it's also an IoT device that mines bitcoins in the cloud I'm not interested.

    1. Re:3D printed neural network? by Anne+Thwacks · · Score: 1
      Unless it's also an IoT device that mines bitcoins in the cloud I'm not interested.

      No, it just posts all your Bitcoins to Zuckerberg

      --
      Sent from my ASR33 using ASCII
  16. More about printing than DL by itamblyn · · Score: 3, Informative

    The impressive part of the work is the fact that they were able to print materials that control light so well, not the actual network itself. The weight optimization and topology design were still done using standard computing hardware - this is just a physical realization of a trained network. You could build something equivalent out of water and tubes (though it would be slower and wetter obviously). The cool part is the optical control which is now possible, not the fact that MNIST works.

  17. Oh is 3D printing still a "thing"? by Anonymous Coward · · Score: 0

    It's like a zombie technology, it won't go away!

  18. holy ... by cascadingstylesheet · · Score: 2

    ... that's some serious buzzword bingo there.

    1. Re:holy ... by Anonymous Coward · · Score: 0

      Agreed, even on the first skim through:
      -3d printing
      -machine learning
      -optical computing
      -car analogy
      -facebook
      -paper is paywalled

      Add some blockchain and carbon nanotubes and it might be the most stereotypical /. article ever

    2. Re:holy ... by cwsumner · · Score: 1

      ... that's some serious buzzword bingo there.

      That's what happens when the salesmen get to write the press release... 8-P

  19. Thanks Dr. Buzzword by Anonymous Coward · · Score: 0

    Did this guy throw darts at a bunch of headline grabby buzzwords and make that the focus of the research as well as the title of his papers?

    1. Re:Thanks Dr. Buzzword by Anne+Thwacks · · Score: 1
      Did this guy throw darts at a bunch of headline grabby buzzwords and make that the focus of the research as well as the title of his papers?

      No - he employs a neural net to throw the darts.

      --
      Sent from my ASR33 using ASCII
    2. Re:Thanks Dr. Buzzword by Anonymous Coward · · Score: 0

      Did this guy throw darts at a bunch of headline grabby buzzwords and make that the focus of the research as well as the title of his papers?

      No - he employs a neural net to throw the darts.

      We could employ a neural net to takeover your continuous, worthless white-noise posting. And then as soon as it does, we could murder it and enjoy the silence.

  20. Someday by jbmartin6 · · Score: 2

    I hope I live to see the day when we just say "manufactured" instead of "3D-printed"

    --
    This posting is provided 'AS IS' without warranty of any kind, implied or otherwise.
    1. Re:Someday by drinkypoo · · Score: 3, Insightful

      We still say "machined" when we could just say something was manufactured via machining. And we will continue to machine forged billets for the foreseeable future, because forged metals have a superior internal structure as compared to sintered. Additive manufacturing is likely to never completely replace other forms of manufacturing, at least not until we develop true nanotechnology and can build our own structures at that level. And it is highly doubtful that will happen in any of our lifetimes.

      --
      "You're right," Fisheye says. "I should have set it on 'whip' or 'chop.'"
    2. Re:Someday by quanminoan · · Score: 4, Interesting

      Additive manufacturing will never replace common manufacturing techniques like casting or forging. This isn't just due to structure, but energy costs as well. Rastering layer by layer, constantly heating and cooling, is far more energy intensive than a "melt and pour". Batch injection molding 1000 parts will always be faster than build by layer.

      Not to shit on AM however - complexity becomes free. With AM you can make a gadget as optimum as you want with intricate bio-inspired structural conmponents, and so long as production volumes are relatively low, say 10,000, AM might be able to compete based on the combination of better performance and material cost savings. Things like rocket engine components are ideal for this.

      Not necessarily a revolution, but a very welcome new tool. CNC machining probably had a greater impact, but went largely unnoticed by the general public. Advanced automation (think robotics mixed with machine learning and AI) is probably poised to make a larger contribution than AM, IMO.

  21. Slow Slashdot by Anonymous Coward · · Score: 0

    I'm getting all my stories ahead of you - and a) I'm not trying b) I'm sure I'm not getting the best tech stories out there.

  22. 400 GHz "light" by dvase · · Score: 1

    While the slashdot summary uses the term "light", the paper states that they used a 0.4 THz source — not the frequency/wavelength most people think of when they hear the word "light".

    1. Re:400 GHz "light" by Anonymous Coward · · Score: 0

      uhh, 400 ghz = 749nm, or just barely in the visible red range.

    2. Re:400 GHz "light" by Anonymous Coward · · Score: 0

      No.
      Frequency of red light starts at 400THz you nutsack, your math is wrong.

  23. Buzzword Bingo! by Zorpheus · · Score: 1

    This must be at the top of many search requests, but does it actually make sense?

  24. HAL 9000 by Anonymous Coward · · Score: 0

    This is the first step to making HAL 9000.

    I for one welcome our new robotic overlords.

  25. holograms by Anonymous Coward · · Score: 0

    Could a series of computer-generated holograms be used instead?

  26. BINGO! by Anonymous Coward · · Score: 0

    My clickbait bingo card filled up from the headline alone.

  27. It's the next level of robotic brains! by bensch128 · · Score: 1

    It reminds me of the intro sequence from an Ian Banks book (I can't remember which one) where a robot is on a ship which is attacked and is forced to eject different levels of consciousness each of which uses a different underlying technology. The photonic brain was the third most powerful level of consciousness it had. It eventually has to resort to it's "primitive" silicon brain because everything else has been corrupted. Maybe the book was by Vinge Verge. I can't remember...