9 ms·
Google app analyzes selfies and finds doppelgangers in museum portraits (2018)
- Proven 6y agoThat's some creepy shit for fools without appreciation for privacy (who will later call on their government to prevent Google from rightfully using data they willingly provided). I wish I could downvote this post.
- holografix 6y ago... feeds high res biometric data of your face to the highest bidder.
- rektide 6y agoI just spent the weekend re-reading Virtual Light & Idoru, Gibson's Bridge series, and I think both books mentioned this theory that the most effective way to ask if anyone's seen a person is to try to identify what celebrity a person looks like, & go around asking if anyone had seen a John Malkovich looking dude around the area, or who-have-you. Both books mention that there is a part of the brain literally trained on looking for/at celebrities, special purpose function for it, and declaring that it's the most likely system to register for most people, our common node/vibration where we'll find matches. Kind of getting a weird celebrity-recognition-part-of-the-brain vibe of that off this story? Except it's AI's dreaming instead of human people?
- caf 6y agoPre-pandemic, my wife and I would sometimes play a game at events with large numbers of people we didn't know, where one of us would mention a celebrity lookalike and the other had to spot them in the crowd.
- StavrosK 6y agoTo everyone remarking about privacy, the app does say that the photo won't be sent anywhere. I wonder if everything is done on-device.
- motoboi 6y agoFace has landmarks which can form a unique signature, like a fingerprint. Mobile extracts facial landmarks “signature” and sends to server only numbers. Server runs an algorithm to find “neighbors” to those facial landmarks based on signatures collected from museum photos and present the closest match. Actually pretty simple nowadays. Challenge off course is to collect the museum photos, what google spent some millions doing for other project.
- tsewlliw 6y agoConceptually, this is rather like lossy compression with a shared dictionary. In practice your likeness has still been trasmitted. Jpeg files are after all just numbers too, merely less specialized.
- motoboi 6y agoAgreed. But you’ll only be able to reconstruct the original face if you have a photo from it. Which google probably has anyway.
- securingsincity 6y agoCan we add 2018 to this? this article is from January 18 2018. It will prompt me to see if the one they associated with me then has been updated.
- SomewhatLikely 6y agoMaybe tag this with 2018? I wasn't sure if this was an update to the algo I read about back then.
- throwitaway1235 6y agoI find more value in the historical doppelganger than free email. FAANG should come up with more parlor tricks for my data, instead of boring office products lol.
- rosege 6y agoThe best it could find came back at only 61%. I like the virtual tours on the app but I would like (maybe someone has done this already?) an AR app/option where when you are in a place like the Acropolis you hold it up and it shows you what it looked like in its heyday.
- baby 6y agoLink: https://apps.apple.com/us/app/google-arts-culture/id1050970557 https://apps.apple.com/us/app/google-arts-culture/id10509705...
- isaacimagine 6y agoAndroid: https://play.google.com/store/apps/details?id=com.google.android.apps.cultural&hl=en https://play.google.com/store/apps/details?id=com.google.and...
- oh_sigh 6y ago(2018), and this doesn't work in Illinois because of their anti-biometrics law: https://www.chicagotribune.com/business/ct-biz-google-art-selfies-20180116-story.html https://www.chicagotribune.com/business/ct-biz-google-art-se... I assume this is more of Google not wanting to put the effort into researching and complying with the law for a one-off arts and culture project, sort of like how the website for Mom & Pops deli in upstate NY started banning European visitors because of fear/misunderstanding of the GDPR.
- Jkvngt 6y agoNew trick to get people to upload an image of themselves?
- cynoclast 6y agoYes.
- judge2020 6y agoThey already had one of those[0,1], and received enough data from it to discontinue the free aspect[2]. 0: https://www.howtogeek.com/227084/how-to-use-google-photos-to-store-an-unlimited-amount-of-photos/ https://www.howtogeek.com/227084/how-to-use-google-photos-to... 1: https://support.google.com/photos/answer/9547620?hl=en-GB#gbwa:~:text=Help%20improve%20machine%20learning%20technology%20for%20Google%20Photos https://support.google.com/photos/answer/9547620?hl=en-GB#gb... 2: https://blog.google/products/photos/storage-changes/ https://blog.google/products/photos/storage-changes/
- Cthulhu_ 6y agoWell, their social network (Google+) didn't really take off.
- j_barbossa 6y agoSorry for the rant but Google has time and money to build things like this and yet they don't have customer support for people whose account got suspended for no reason?
- scrollaway 6y agoYou have time to make this comment and yet you don't have time to volunteer at your local pet shelter? (I'll buy you a beer if you convincingly explain to me that these two analogies are meaningfully different) PS, not defending google, they generally suck.
- oreille 6y agoProviding customer support is a part of a having a business and should be prioritized over adding random features no one asked for to your product. Volunteering at a pet shelter is not part of the business of being an angry anonymous internet person.
- elcomet 6y agoNon-paying users are not customers, thus google doesn't really earn them support for their products. For paying users, that's a different story.
- merb 6y agothey banned somebody that bought stuff inside the app store.
- nolok 6y agoThough Google is merely a storefront en payment intermediary of what the banned user bought (apps and movies), it could be argued the client of google per se is the companies selling the product, they're the one paying Google a share of their sales. Not that I disagree with you here just making the argument. Like parent said, I found that google has decent customer service once you clearly become their client (which I have been on adwords, analytics and on gsuite).
- codemusings 6y agoIt's ReCaptcha all over again. They're training their neural nets for free and as an incentive tell you which painting is the closest.
- blackbrokkoli 6y agoThis is the kind of stuff that people in 300 years will shake their head in disbelief about when reading about it in the footnote of an history book. Up there with the people X-raying their feet to find the perfect shoe, the people who deciding to illegally tap into a natural gas line at the overfunded New London High School and the people happily unearthing lost church books to hand over to the Nazis just for the thrill of the puzzle. You have this minuscule upside of going "heh, I am in a painting" and this giant array of really, really, really fucking bad second order effects looming on the other side. This really summarizes the problem with the current society and Big Tech perfectly.
- ButterWashed 6y agoI understand what you're saying and I'm not having a go at you but "really, really, really fucking bad second order effects" are what exactly? Telling anyone that people in 300 years will judge them negatively isn't really going to motivate them to change. I'm personally at the point where there needs to be a set of horror stories written in non-tech language about data people share directly impacting their life. But what are those stories?
- blackbrokkoli 6y ago* Almost all genocides (and extermination wars), especially those in one's country, were heavily dependent on data availability and the technology to crunch it. Most prominent maybe the Holocaust: Doing a census so thorough that the asinine concept of "Quarter-Jews" and what not could be calculated was estimated to take ten years to infinity. Until IBM offered to custom-design their new punch card tech design for the Nazis; they did in three months. The data? Openly accessible church books and forms people filled out, mostly about themselves. You could really have a field day with the data lying around today plus the one people readily give you in exchange for some gimmicks. Arguably, this is already ongoing in various parts of the world. * Terrorism so dangerous that it actually makes a dent in the yearly death count. There is an article about a hacked water plant right now on HN. Just scale that up to, say, all the water plants. For every wacky "hey, we could do this with this data" idea there is somebody else with an equally but way more fucked up "hey, we could do this with this data" idea. * There will be a 2030 study about mass discrimination by AI. Maybe it will be "This hiring AI was used to reject 1.3 billion applications over the last 10 years. Apparently it really dislikes Pakistani men for some reason. Oops" or "There was a weird error in our visa pipeline software. Our condolences to the families of the 43,561 people that were rejected and subsequently killed because their last name contained the letters 'rt'", I don't know. If that sounds absurd to you, go check the HN thread about the alleged murderer, where people argue whether "IP protection law" and "Move Fast and Break Things" are good frameworks to discuss software that literally decides about live and death.
- asssa3000 6y agoSomeone is gathering images with metadata to sell it later to interested parties
- acd 6y agoWho is the best doppelganger most eigenface like face of Monalisa?
- stunlocked 6y agoBeing german i had a very hard time figuring out what this comment is supposed to mean. Could you rephrase it so one might be able to answer your Question?
- HNSAXU 6y agoGoogle should do deeper analysis on them self. They could be shocked how censurating and biased corporation they are.
- mikro2nd 6y agoSo a dark trick to get people to upload pictures of themselves to feed the Google surveillance machinery...?
- lurquer 6y agoGoogle already knows what you look like. And, more importantly, what you look at.
- stemlord 6y ago"I don't need to protect my privacy because they already know everything about me" is a form of privacy nihilism. You still have plenty of privacy worth protecting
- asdfasgasdgasdg 6y ago> You still have plenty of privacy worth protecting But, not this particular bit. If they already know the value of a given bit there is minimal possible harm in disclosing it to them again.
- stemlord 6y agoYes, this particular bit because that's speculation. How do you know they already have it and that there's no added value to them in doing it again?
- dimitrios1 6y agoThe Framers detested dragnet searches and preventing dragnets was the cornerstone of the Fourth Amendment. Data mining threatens to bring back the dragnet search in digital form.
- lurquer 6y agoThe Fourth Amendment is worthless. All the information is held by private companies and the gov’t gets the info from them voluntarily. Ditto for the First Amendment. Worthless. Instead of the govt prohibiting speech, the govt just has private companies do it for them. How many times a week on HackerNews do you see this kind of exchange: A: Facebook censored this or that and thus infringes on free speech. B: It’s a private company; it can do what it wants. A: But, but, there’s something wrong with that! B: Don’t like it, make your own Facebook. Hur dur... Etc. Ad nauseum. Clearly, both (A) and (B) above are trying to reason through something within an outdated and outmoded framework; it’s why both sides sound silly. Anyway, this is the same thing... but with the 4th amendment. Same problem. A private company that holds your data can do with it what it wishes. And, like the first amendment, if the govt makes it clear to these private companies that they ought ‘voluntarily’ cooperate with the govt “if they know what’s good for them,” then then Bill of Rights is worthless. The 1st and 4th were meant to protect us from a tyrannical govt; not a tyrannical corporation. Scrap the entire ball of wax and start over. The Constitution — at this point — is an utter joke.