27
May
lambda-labs-logo

Lambda Labs, a small start-up out of San Francisco, is set to drag us kicking and screaming into the dystopian sci-fi future we all knew was coming. Okay, that might be overstating the point, but the company has announced its intention to release a facial recognition API for Google Glass this week. Should this pan out, you'll always have to wonder if that fellow wearing Google Glass remembered your name, kid's age, and occupation because he has a good memory, or because the cloud told him.

labs

The Lambda Labs facial recognition API has been in operation since last year, and is currently used by about 1,000 developers. This new technology would allow third-party Glass devs to build apps that utilize Lambda Labs' system to remember and recognize faces, but not in real time. The Google Mirror API doesn't currently allow for live streaming camera data to a remote developer's server. Users would have to snap a photo, send it to the facial recognition app, then wait for the results to come back. This would result in a few seconds of delay, during which time the Glass-wearer would have to make awkward, general small talk. This could change as Google's tools improve.

The Glass API does not, at this time, specifically ban such functionality. Although, that might change after a few high-profile stories about Glass users creeping on people using facial recognition. It's important to note Lambda Labs doesn't provide the data to identify any random person – it needs source data. Google has repeatedly stated it has no plans to build a facial recognition engine for Glass, citing the privacy implications. Google+ recently got some similar features in the photo hub, though.

You can mess around with a demo of the Lambda Labs engine on the company's website. The future is coming, so get ready.

[TechCrunch, Lambda Labs]

Ryan Whitwam
Ryan is a tech/science writer, skeptic, lover of all things electronic, and Android fan. In his spare time he reads golden-age sci-fi and sleeps, but rarely at the same time. His wife tolerates him as few would.

  • Jason Storey

    Guess I'll be the first to ask.... Why Jennifer Aniston?

    • fonix232

      "Lambda Labs doesn't provide the data to identify any random person – it needs source data.". They just grabbed three celebrity and a photo. Random selection, so it cannot be really answered.

    • mechapathy

      I'm pretty sure those are percentages of matched characteristics. Based on whatever it's looking at, it's a 74.2% match for their data on Tiger Woods, or a 16.1% match for their data on Arnold Schwarzenegger, or a 6.5% match for their data on Jennifer Aniston.

    • jamheart

      my сlassmatе's һalf-sіstеr maĸеs $74/һr οn tһе сοmрυtеr. Sһе һas bееn ουt οf a jοb fοr 8 mοntһs bυt last mοntһ һеr сһесĸ was $12956 jυst wοrĸіng οn tһе сοmрυtеr fοr a fеw һουrs. Rеad mοrе οn tһіs sіtе, .. ..up444.com

  • fonix232

    While I do not know how Lambda Labs' software work, wouldn't it be easier to do facial recognition on the device to a level it can be transmitted (much like FaceLock's face structure data, this could be done with the LL API, so it transmits facial data already recorded, alongside with some further information such as skin tone, eye colors, hair color, etc.), that way speeding things up. Building facial data continuously should be possible, making it more precise by an algorithm as the Glass-wearer sees more and more of the persons face. Of course this could not be completely real-time, but if we limit the search base to some social media (Facebook, Google+, any other social website with clearly marked photos of the people, limiting search to friends, friends friends, and extending search if nothing found), and it is already not a few hundred million we are looking at, but a lot less.

    Though this indeed makes me question the future of social privacy. If anyone writes an application that stores facial recognition data, and it gets mainstream alongside with the Glass, literally EVERYTHING will be stored almost publicly about us. Including locations and times, clothing style, preferred makeup, hairstyle, haircolor, hell, if it's well written, even the current emotion! So yes I am quite worried, no matter how awesome I find the idea of getting the number of the cute girl who smiled at me from the other side of the street in a moment.

    • http://www.androidpolice.com/author/cody-toombs/ Cody Toombs

      I just have one thing to say about that first part. Sure, you could do all of that on the device, but then Lambda Labs wouldn't be able to charge a subscription fee.

      • fonix232

        Of course they would be able to! The services main part, the fast comparison of the photos from all the social networks (and possibly with their own database too, to increase accuracy), would be all running from their server. Meaning, no subscription, no access to comparison, no working software.

        • http://www.androidpolice.com/author/cody-toombs/ Cody Toombs

          Sorry, I was thinking of abstracting out the communication entirely. Once you're doing that much processing on the device, the technical benefits of transmitting anything to Lambda servers is mostly lost (except that it might have continuous updates). The processor intensive part is mining the photo for characteristics. Once that information has been extracted from a picture, it's pretty trivial to hit a relatively small database stored on Glass (a couple of thousand people who've already been pre-processed).

          • fonix232

            I still think that doing a local scheme creation then comparing it with Lambda's storage is way much better. Especially if you want to search on social sites, and have a slow connection. A serverpark just for the role of serving small requests resulting huge data load from different servers around the world is way faster than loading each and every tagged picture of friends, friends of friends, filtering out false results (I love those "tag me if..." photos on Facebook, I'm simply just interested which part of the cake/calendar/whatever you are. /sarcasm).

            And while it indeed makes everything faster, it can be a huge privacy issue. The more people use it, the more they get uncontrolled access to pure relational information (I invite 5 friends, they all invite 5 friends, and so on, then in almost no time my whole social grid is mapped, with access to every photo, post, video, anything. Of course for some it is important to make exact facial recognition easier, but I don't know if you want to be in a database where all of your information is available with a simple face search.

  • Robert Antal

    This is starting to look more and more like watch dogs =))

  • Asphyx

    All these things are really cool and nice but If Glass is loaded with even HALF the things we hear it's going to do I would LOVE to see the Data Bill of the owner that results from just a month of Facial Recognition!

    I mean without unlimited data you could go through your allotment in about a week!

    • Aleksey_US

      screw data... think battery life, will I need to charge it every 30 min?

      • Asphyx

        Well after Google Glass will come Google Hat which has a built in Solar Panel to power everything!
        LOL

  • Nata Greer

    Seems to me that the idea of the future of this being able to identify anyone of the 7 billion people in the world is a bit of a stretch, not to mention the privacy issues. The real practicality would be in small registered databases setup by private organizations for their own use.

    Example: some multinational corporation with 10,000 employees registers everyone within a database that anyone in the company can access. It would be limited to the organization and as such, part of the employment agreement would be allowing your professional information to be utilized in this manner.

    Any organization wanting this type of database would pay the subscription fee and recognition could be limited by geographic location by gps.

    The idea that this could recognize some random person on the street you never met before is not feasible, not to mention the possibility of look-alikes.

    • Fatty Bunter

      It's definitely feasible - just not right now.

  • ProductFRED

    Dammit, I was expecting Half-Life 3.

  • DeadSOL

    He DOES look 6.5% like Jennifer Anniston!

  • Ben Baranovsky

    No need to worry about privacy just yet. I tested their system and so far it's not very good. I tried using Arnold Schwarzenegger and it thought he was Renee Zellweger with a 48% chance and himself with a 17% chance.

  • Ryan Fino

    TIL Arnold S. looks more like Jennifer Aniston than Arnold S.

  • Guest

    so you show a picture of a horse do you get sarah jessica parker?

  • Chris Ismael

    You can also check out Lambda's Face recognition interactive API documentation here - https://www.mashape.com/lambda/face-recognition#!documentation