(Thread IKs:
Platystemon)
|
makeup and strange hairspray architecture come off too unlike below-the-nose sunglasses, it doesn't come off on demand, isn't forced off in low-light contitions, isn't a major safety risk in situations like protests, and there don't currently exist laws and regulations calling for it to be removed on demand
|
#
?
May 25, 2018 23:23
|
|
|
|
| # ? May 29, 2024 16:28 |
|
|
Another huge issue with it is that it's a defense against an attack that doesn't even exist. It should only be considered an artistic/political statement IMO
|
|
#
?
May 25, 2018 23:25
|
|
|
people seem to forget that all this poo poo inspired by the coolio hacker dudez from neuromancer and snow crash and poo poo was used by people who were criminals and fugitives already. like sure, maybe if you are a dataless ghost you wear that stuff out, but if you're fully wired into the capitalist tech machine this is only going to draw more attention. "hey, this dude has purchase history and life insurance and all that but we never see his dot, what's up with that? oh dazzle paint? outlaw it and say it's for public safety or kidnapping of w/e lol"
|
|
#
?
May 26, 2018 01:12
|
|
|
https://www.youtube.com/watch?v=CA3ix3zt9bo
|
|
#
?
May 26, 2018 01:31
|
|
|
lets all just become gas mask weirdos
|
|
#
?
May 26, 2018 01:51
|
|
|
just wear those hospital masks. those can’t realistically be outlawed and you can just say “I’m sick” throw in a normal pair of sunglasses, maybe a hat, and BOO(M) you’re a ghost
|
|
#
?
May 26, 2018 02:39
|
|
|
General Dog posted:Pictured: the world's least recognizable man in russia we had a joke about "Uncatchable Joe" who could not be caught because no one was looking for him
|
|
#
?
May 26, 2018 03:42
|
|
|
Using a few LEDs, a hat, and a and a bit of computer savvy you can train up an adversarial model that makes you appear as whoever you want to be. https://arxiv.org/pdf/1803.04683.pdf  
|
|
#
?
May 26, 2018 06:33
|
|
|
lmao I don't speak computer but I assume those work bc they don't look very different to me
|
|
#
?
May 26, 2018 07:01
|
|
|
DragQueenofAngmar posted:lmao I don't speak computer but I assume those work bc they don't look very different to me Literally the same Asian guy is making a facial recognition algorithm think he is variously Moby/Hoi-chang/etc, just by having LEDs highlight different areas of his face. Adversarial models can be created for all neural nets (that you are able to run tests against), it is a fundamental problem with their nature.
|
|
#
?
May 26, 2018 07:39
|
|
|
Vincent Van Goatse posted:those sleep apnea photos are people wearing a thing that monitors their breathing and EEG during the night they remind me of the nightmare fuel scene at the end of Superman 3
|
|
#
?
May 26, 2018 09:25
|
|
|
Krankenstyle posted:Literally the same Asian guy is making a facial recognition algorithm think he is variously Moby/Hoi-chang/etc, just by having LEDs highlight different areas of his face. one of my favorites: https://twitter.com/daniel_bilar/status/925541876484231168
|
|
#
?
May 26, 2018 13:32
|
|
|
turtles and rifles... comrade Jeb! will save us!
|
|
#
?
May 26, 2018 13:41
|
|
|
Crusader posted:one of my favorites: yeah same, it owns I think someone on here called adversarial models "optical illusions for computer vision". Carthag Tuek has issued a correction as of 14:35 on May 26, 2018 |
|
#
?
May 26, 2018 14:32
|
|
|
Crusader posted:one of my favorites: This would probably work on cops.
|
|
#
?
May 26, 2018 15:24
|
|
|
|
|
#
?
May 26, 2018 16:22
|
|
|
https://www.youtube.com/watch?v=OUJe0Wmfo5g
|
|
#
?
May 26, 2018 22:44
|
|
|
Crusader posted:one of my favorites: Jeb! is not a gun
|
|
#
?
May 27, 2018 02:42
|
|
|
etalian posted:Jeb! is not a gun https://twitter.com/jebbush/status/699706718419345408
|
|
#
?
May 27, 2018 02:45
|
|
|
Strangely this is categorized as a turtle
|
|
#
?
May 27, 2018 02:51
|
|
|
hobbesmaster posted:Strangely this is categorized as a turtle 
|
|
#
?
May 27, 2018 02:58
|
|
|
https://twitter.com/grinfie/status/1001277675485396998
|
|
#
?
May 29, 2018 02:54
|
|
|
https://twitter.com/RealRonHoward/status/1000853152856399872
|
|
#
?
May 29, 2018 03:26
|
|
|
https://www.youtube.com/watch?v=i1sp4X57TL4 I'm not a rebel, I'm a toaster https://gizmodo.com/this-simple-sticker-can-trick-neural-networks-into-thin-1821735479
|
|
#
?
May 29, 2018 03:28
|
|
|
|
|
#
?
May 30, 2018 02:07
|
|
|
Improved geofencing means you can now target ads at people in emergency rooms. https://www.npr.org/sections/health-shots/2018/05/25/613127311/digital-ambulance-chasers-law-firms-send-ads-to-patients-phones-inside-ers
|
|
#
?
May 30, 2018 14:09
|
|
|
https://twitter.com/fouroctets/status/1000067640121946112
|
|
#
?
May 30, 2018 17:20
|
|
|
lmao
|
|
#
?
May 30, 2018 20:34
|
|
|
Blockade posted:Using a few LEDs, a hat, and a and a bit of computer savvy you can train up an adversarial model that makes you appear as whoever you want to be. my pisstape hat grows mightier
|
|
#
?
May 30, 2018 23:57
|
|
|
Blockade posted:Using a few LEDs, a hat, and a and a bit of computer savvy you can train up an adversarial model that makes you appear as whoever you want to be. Sadly as the paper points out, you have to have an exact copy of the facial recognition model as used by the surveilling facial recognition system, including the exact same training data
|
|
#
?
May 31, 2018 00:37
|
|
|
https://twitter.com/NBCNews/status/1001953665652051968 Thanks for the update, faceless multinational pharma corp.
|
|
#
?
May 31, 2018 00:38
|
|
|
but those 1000s of ods? yeah that was us
|
|
#
?
May 31, 2018 01:21
|
|
|
Trabisnikof posted:Sadly as the paper points out, you have to have an exact copy of the facial recognition model as used by the surveilling facial recognition system, including the exact same training data Believe it or not its not too difficult to do this, many facial recognition systems only use one of a handful of commercially available models, and even if they have something custom you can train against it as long as you can get a pass/fail response in some way.
|
|
#
?
May 31, 2018 02:32
|
|
|
mrmcd posted:https://twitter.com/NBCNews/status/1001953665652051968 woke pharma trying just tryna help you sleep
|
|
#
?
May 31, 2018 02:43
|
|
|
Blockade posted:Believe it or not its not too difficult to do this, many facial recognition systems only use one of a handful of commercially available models, and even if they have something custom you can train against it as long as you can get a pass/fail response in some way. That's certainly true for most of the commercial off the shelf stuff, but since you need the exact same training set, most of the actual interesting targets will either have training data you don't have access to or use a proprietary algorithm. Take for example, Rekognition, which offers facial recognition services for US police department camera feeds. I'm sure its just off the shelf components stitched together into saas, Bad example. But, as per our shttiest possible cyberpunk dystopia the Rekoginition FAQ has this gem: quote:Q: What is a face collection and how do I create one? Trabisnikof has issued a correction as of 02:59 on May 31, 2018 |
|
#
?
May 31, 2018 02:48
|
|
|
in recent papers on adversarial examples (the academic term for a specially-constructed input to fool a computer vision system) there is evidence that they can transfer across systems, and if you can construct one of these examples that can fool a bunch of different computer vision systems (differences in eg the specific neural network structure it has, the dataset it uses, etc) then it could fool one you have no knowledge of because it happens to be particularly robust although most of the datasets people use for academic research have at most on the order of 1000s of things it can classify it is. i dont know enough about what facial recognition neural networks do when they want to classify between tens or hundreds of millions of people. my guess is that they in practice perform like poo poo
|
|
#
?
May 31, 2018 03:24
|
|
|
https://arxiv.org/abs/1705.07386quote:Biometric authentication is important for a large range of systems, including but not limited to consumer electronic devices such as phones. Understanding the limits of and attacks on such systems is therefore crucial. This paper presents an attack on fingerprint recognition system using MasterPrints, synthetic fingerprints that are capable of spoofing multiple people's fingerprints. The method described is the first to generate complete image-level Masterprints, and further exceeds the attack accuracy of previous methods that could not produce complete images. The method, Latent Variable Evolution, is based on training a Generative Adversarial Network on a set of real fingerprint images. Stochastic search in the form of the Covariance Matrix Adaptation Evolution Strategy is then used to search for latent variable (inputs) to the generator network that optimize the number of matches from a fingerprint recognizer. We find MasterPrints that a commercial fingerprint system matches to 23% of all users in a strict security setting, and 77% of all users at a looser security setting. The underlying method is likely to have broad usefulness for security research as well as in aesthetic domains. They work just about as well on different fingerprint readers they weren't trained for as well.
|
|
#
?
May 31, 2018 03:32
|
|
|
holy poo poo these "real life streams" are so dystopian
|
|
#
?
Jun 1, 2018 05:06
|
|
|
people humiliating themselves and creating disturbances for pocket change https://www.youtube.com/watch?v=jgDQ7exm3FU
|
|
#
?
Jun 1, 2018 05:51
|
|
|
|
| # ? May 29, 2024 16:28 |
|
|
https://twitter.com/Reuters_Health/status/1002587602841632769 not cyberpunk but def dystopian
|
|
#
?
Jun 1, 2018 18:21
|
|
