I went to a concert the other night, and they where scanning driver licenses and running facial recognition on everyone entering. It was very disconcerting. It is clear to me that we need new laws to protect against the misuse of this technology. Laws that apply to everyone without exceptions for law enforcement. If we don't I fear we are headed towards a 1984 type of future.
I played around with Rekognition a few months ago and frankly wasn’t that impressed with what it could do. As far as I can remember it could sort of compare if the same person was in two photos (definitely false negatives with various photos of myself I compared), basic OCR of text off images, and not very accurate recognition of certain things/objects.
Based on OP's incident and your expereience, I think that AWS Rekognition is currently in its consumption/learning stage. Give it a few months to adequately digest that data and then try!
Woo, better facial recognition for all kinds of nefarious purposes!
Edit: Dead serious, how will Rekognition positively affect the common person? All I see is a myriad of possible ways government, corporations and individuals have abused camera footage today, and emboldened with better technology these problems are likely to get much worse.
I work in tele-medicine and we use rekognition to see if users are following the instructions correctly. E.g. you have to be at a certain distance from the screen, close an eye, take your glasses off, etc.
Also, things like virtual try-on for glasses use facial recognition tech. This allows you to map frames on your face so you can buy them online.
These applications don't use facial recognition to identify users, but to be smart digital assistants. They can function anonymously and don't have to store sensitive information.
Off the top of my head, there will be positive usecases in assistive technology for the visually impaired. Also, if the computer knows where I am looking, it may enable some productivity tools.
That said, I agree with you that these potential positive uses do not outweigh the negative ones you allude to.
I'm sure it will be used for nefarious things too however,
Programs involving photography, yes the big dogs have their own solutions already but this lowers the bar for us hobbyists.
Photo affects (Blurring the background and not the face), sorting images by labels (search for trees and get your photos with trees), searching by people.
Potentially things traffic monitoring, foot traffic in a mall, security systems (record a photo of all customers in the case of theft), automating tracking objects in videos.
There are tons of benign/helpful use cases for face ID. Your car can recognize you and automatically set your preferred radio stations, seat position etc. Your computer can log you in without a password. Border control can be "scan your passport here and keep walking". An airplane boarding card can be "take a picture with your webcam during check-in and just walk on the plane." Your phone can automatically "press the shutter" for a family picture when everyone is smiling, instead of guessing at that timer thingie. Your makeup mirror can turn on special lights when a face is in front of it. You can pre-set mood lighting for your apartment for when you bring a new date home vs football night. Your front door can unlock itself for a handful of trusted people, even when you're out of town and forgot to feed the cat. Or it can set an alarm when your mother in law comes over. You can search your entire media library for just the right picture: "with Phil, but not Gina, where we're having fun and in cat costumes."
Face identification has become the lightest weight authentication method available. So anywhere you would otherwise use something you know (password) or something you have (a transit pass), you could consider using something you are (your face) instead. It's not right for EVERY authentication scenario, but it should be considered for many of them.
Not everything AI is creepy or privacy invading. Especially when you consider running it locally.
That said, Rekognition as a product specifically targets law enforcement and federal agencies, and most of their use cases are NOT good IMO. We do need a strong product preference for open source, or products with AI ethics oversight.
Bias: I work for Microsoft helping big companies implement latest edge tech like this. MS has an ethics board filtering all its AI work, the Azure face ID product doesn't keep user data, and we often implement local-only face id in what I consider non-creepy scenarios. So that defines a lot of how I think "responsible implementation" looks.
It’s quite hard to do this randomly. You need a substantial data set to train the machine to recognise your targets face. You can’t just upload 5 or 6 blurry social media photos.
Seems like a stretch to invent hypotheticals for how a common person might benefit. I am curious what the usage stats are for it and what reflections that has on society.
If the governments using these these sorts of technologies are democratic, responsive, self regulating, have an informed populace, and lack of incentive problems. This is fine, but that’s a long chain of ifs.
Is this something that is considered easy to build locally? When I last looked at the open source options a few years ago it was basically all researchware grade stuff. Amazon is offering this is a production service with SLAs.
[+] [-] bhhaskin|7 years ago|reply
[+] [-] tintor|7 years ago|reply
[+] [-] btown|7 years ago|reply
[+] [-] tictok4|7 years ago|reply
[+] [-] dawhizkid|7 years ago|reply
[+] [-] pupdogg|7 years ago|reply
[+] [-] aviv|7 years ago|reply
[+] [-] sys_64738|7 years ago|reply
Legally, this is the only way to destroy these big brother obsessions that creepy companies like Amazon have.
[+] [-] m1sta_|7 years ago|reply
[+] [-] metildaa|7 years ago|reply
Edit: Dead serious, how will Rekognition positively affect the common person? All I see is a myriad of possible ways government, corporations and individuals have abused camera footage today, and emboldened with better technology these problems are likely to get much worse.
[+] [-] jtwaleson|7 years ago|reply
Also, things like virtual try-on for glasses use facial recognition tech. This allows you to map frames on your face so you can buy them online.
These applications don't use facial recognition to identify users, but to be smart digital assistants. They can function anonymously and don't have to store sensitive information.
There are many good applications.
[+] [-] milaresearcher|7 years ago|reply
That said, I agree with you that these potential positive uses do not outweigh the negative ones you allude to.
[+] [-] penagwin|7 years ago|reply
Programs involving photography, yes the big dogs have their own solutions already but this lowers the bar for us hobbyists.
Photo affects (Blurring the background and not the face), sorting images by labels (search for trees and get your photos with trees), searching by people.
Potentially things traffic monitoring, foot traffic in a mall, security systems (record a photo of all customers in the case of theft), automating tracking objects in videos.
[+] [-] ohthehugemanate|7 years ago|reply
Face identification has become the lightest weight authentication method available. So anywhere you would otherwise use something you know (password) or something you have (a transit pass), you could consider using something you are (your face) instead. It's not right for EVERY authentication scenario, but it should be considered for many of them.
Not everything AI is creepy or privacy invading. Especially when you consider running it locally.
That said, Rekognition as a product specifically targets law enforcement and federal agencies, and most of their use cases are NOT good IMO. We do need a strong product preference for open source, or products with AI ethics oversight.
Bias: I work for Microsoft helping big companies implement latest edge tech like this. MS has an ethics board filtering all its AI work, the Azure face ID product doesn't keep user data, and we often implement local-only face id in what I consider non-creepy scenarios. So that defines a lot of how I think "responsible implementation" looks.
[+] [-] darkport|7 years ago|reply
[+] [-] JDWolf|7 years ago|reply
[+] [-] danols|7 years ago|reply
[+] [-] raverbashing|7 years ago|reply
[+] [-] simple_man|7 years ago|reply
[+] [-] foolfoolz|7 years ago|reply
[+] [-] oh-kumudo|7 years ago|reply
Facial recognition service is very easy to build these days, provided that sufficient data is presented, and government has plenty of such data.
[+] [-] kjar|7 years ago|reply
[+] [-] murukesh_s|7 years ago|reply
[+] [-] awirth|7 years ago|reply
[+] [-] mohi13|7 years ago|reply
https://news.ycombinator.com/item?id=15903490