Slashdot Mirror
Google: Our New System For Recognizing Faces Is the Best
schwit1 writes Last week, a trio of Google researchers published a paper on a new artificial intelligence system dubbed FaceNet that it claims represents the most accurate approach yet to recognizing human faces. FaceNet achieved nearly 100-percent accuracy on a popular facial-recognition dataset called Labeled Faces in the Wild, which includes more than 13,000 pictures of faces from across the web. Trained on a massive 260-million-image dataset, FaceNet performed with better than 86 percent accuracy.
The approach Google's researchers took goes beyond simply verifying whether two faces are the same. Its system can also put a name to a face—classic facial recognition—and even present collections of faces that look the most similar or the most distinct. Every advance in facial recognition makes me think of Paul Theroux's dystopian Ozone.
The approach Google's researchers took goes beyond simply verifying whether two faces are the same. Its system can also put a name to a face—classic facial recognition—and even present collections of faces that look the most similar or the most distinct. Every advance in facial recognition makes me think of Paul Theroux's dystopian Ozone.
The NSA and CIA must love the direction this company has taken.
FaceNet achieved nearly 100-percent accuracy...
" performed with better than 86 percent accuracy. "
I'm not able to parse these numbers, or I have a misunderstanding as to what nearly means.
I'm a satanic clam.
on really BAD ideas and freedom-killing ideas?
you don't think this will be mostly used against us, one way or another?
do no evil? yeah, right. I have this bridge here I can sell you...
yet another example of 'lets plow ahead and not care about social blowback from our research'. scientists and engineers really need to go back to school and take an ETHICS course or two. maybe they'd realize that 'because you can, does not mean you should'.
I see nothing good coming from this. nothing at all. just pure evil to be used against people.
it takes a wise person to realize that some things should not be done. of course, google has geniuses but those geniuses have no idea at all how they are being played and how their work will be used to reduce freedom and privacy. sad that smart people can be conned into working against their own best interests ;(
--
"It is now safe to switch off your computer."
Yup, that's a face. I'd recognize one of those things anywhere.
"Have you ever thought about just turning off the TV, sitting down with your kids, and hitting them?"
But how will they ever tag neckbeards? There's too much data interference to get a good analysis of the face for the neural network. Neckbeards will take over us all and be immune to surveillance. We'll have to go back to reading license plates of cars at Taco Bell at 3am.
I'm a satanic clam.
SkyNet approved.
how about instead, when Google face recognition blinks green, it provides the SSN and path to school for Sergey or Larry, instead of me. doesn't that sound appropriate?
if this is supposed to be a new economy, how come they still want my old fashioned money?
I so frequently see people who remind me of someone else. Similar facial features/hair/skin coloring/etc. It would be interesting to have it run thru millions of faces and group them by similarity to see if there is a sort of finite number of different face "types". Also, it would be fun to give it a photo of my face and find my doppelgangers.
I wouldn't mind getting my hand on their API. I am almost done with SkyNet. I just needed better face matching... Too many accidents.
If something is so important that you feel the need to post it on the internet... It probably isn't that important.
And so do "invisible" glasses and scatter light hoodies.
But that's internal face recoginition stats the NSA and other mil agencies don't want you to know.
Just dress up for cosplay and game events (foot whatever) and they have a really really hard time tracking and ID'ing you. Especially if you do social camoflage like changing clothing and group and walk.
-- Tigger warning: This post may contain tiggers! --
Considering that they took the "do no evil" and remove the "no"
Now do you understand Google+'s initial policy on real names?
Pete Boyd
Hmm... you're right, we have to get that under control.
Get the PR dept to slap together a press kit. What we need is footage of some murderers and generally insane people with neckbeards. Make sure that you add a few shots of those islamist loonies, they're full of face wool, too. Tell them to not waste all their powder on one shot, have them release it slowly so it can sink in with the targets. Try to get that Stallman guy in there too, somehow, that way we might get a shot at that open source stuff where we can't sensibly include our backdoors. At least with the boomers it should stick, they know jack about computers but trust the TV.
Networks should gobble this up without asking, it's free news, that's all they care about.
For the millennials, try to get the story into one of those "10 things you didn't know" pages. Slap something together about 10 things you don't know about neckbeards, 10 most heinous murderers (of course you pick the neckbeard fraction), get creative! They'll love it.
Find celebrities and make sure they hate beards. Beards have to become "uncool". Only clean shaven guys get the chicks. And guys, let's not forget the fags.
Get back to me when this is rolling, maybe we don't even have to do more.
We used to have a Bill of Rights. Now, with the rights gone, all we have left is the bill.
Facial recognition and object recognition was always thought of in the AI community as a "pattern interpretation" skill, and we suspected that human brains have special magic gears for "effortlessly" succeeding at these sorts of tasks, while AI coders struggled to emulate our success.
Now we're seriously talking about computers already being better at these tasks than we are. This is one of those milestones in AI research when we have to cross off another item from the list of "things that keep AIs from matching or exceeding human intelligence". For now, there are still many items on that list, but I wonder which ones will be crossed of next, and how soon.
This would help people like me, who can't put names to faces or fail to notice when a face is the same as the one I saw last year... Perfect Glass application.
Just thinking about the accuracy level, there are old:middle:young, male:female, black:white:Asian etc
Can this system even tell the difference between two young white male people or two old female black people? What is the accuracy then?
Sounds pretty rubbish to me.
Waterfox - a Firefox fork with legacy extension support, security updates and better privacy by default.
Services and users high quality facial recognition can be plugged into: CIA, FBI, NSA, state police, local police, bill collectors, "bounty hunters", stalkers, paparazzi, computer security, club membership and virtually every other kind of venue/door lock, real or virtual, sorting your own photos, detecting use of your face in otherwise un-discoverable venues, plus basically about any kind of surveillance you can think of.
See, here's the thing. If they have your face, and they're hunting for a match, that's pretty bad already in terms of your privacy, but it requires that they already have your face and a specific existing concern, so therefore a reason to go looking for you.
But a sufficiently powerful system that can identify you generally -- see your face, associate your identity -- can track you (along with everyone else) when they're not looking for you. The implications for privacy and government misbehavior are extremely negative.
Underlying this is that our system of law is riddled with bad law, for bad reasons, with bad consequences. Copyright, drug use, sexuality, all serve as good examples of issues that the government sees one way, and significant numbers of citizens see somewhat differently. When your private, consensual behavior is unobserved, this causes far less (although certainly not zero) problems. When you are observed all the time... I think we can look for serious negative social impacts.
I've fallen off your lawn, and I can't get up.
That sounds good. The problem is, and will always be, that you may have the right ethics handled, but somewhere else, someone doesn't, and they'll just create and use the tech against you in an environment where you've not looked at it in a real world sense, and properly compensated for it to the extent that is possible.
I've fallen off your lawn, and I can't get up.
So Google has discovered Deep Learning.
I wonder if they'll publish some material claiming they invented the theory behind it in spite of all the previous work, like with their other achievements?
Scientists and engineers are by definition not supposed to be ethical.
"I just invented the bomb. I didn't drop it."
--Brice, Max Headroom Episode 1 "Blipverts", 1987
Reference (in particular, the third video clip): http://www.avclub.com/article/...
Back then that line was meant as tongue-in-cheek humor, funny because of its ridiculousness Depressing that we've degenerated so far that you've actually said the equivalent with all seriousness. (The same could be said for many things in that once funny, now prophetic series.)
As engineers and scientists we do NOT check our humanity at the door, or our ethics. At least, good engineers and scientists do not.
The Future of Human Evolution: Autonomy
Not just neckbeards, creative hairstyles and makeup can easily break facial recognition algorithms. You can already find some nice examples and tutorials for this online. I sort of like the idea of a future where everyone is wearing ziggy stardust kind of makeup.
Google,
can I get API access please so I can find my dopplegangers and partner(s?!)
A blog I run for the wealth
I'll believe it when my phone can reliably recognize me.