|
DONT THREAD ON ME posted:i'd only do it if there were a way to prevent people from using it to deep fake other people. your third visitor on day one would upload a child's face, and your host would drop you about four billiseconds later
|
#
?
May 29, 2019 13:48
|
|
|
|
| # ? Apr 25, 2024 10:06 |
|
|
flakeloaf posted:your third visitor on day one would upload a child's face, and your host would drop you about four billiseconds later and you would make a quadrillion bitcoins in your international waters seastead
|
|
#
?
May 29, 2019 16:16
|
|
|
ah but you’re forgetting that the ML version will have unpredictable results handling unexpected input!
|
|
#
?
May 29, 2019 17:10
|
|
|
is there hdcp but for webcams? just require the user's pics come from an hdcp-compliant 3d webcam. much harder to just upload a 3d scan of a rando celeb's head that way. still very easy for them to just hold up their hostage child or whatever tho
|
|
#
?
May 29, 2019 17:22
|
|
|
Bloody posted:is there hdcp but for webcams? just require the user's pics come from an hdcp-compliant 3d webcam. much harder to just upload a 3d scan of a rando celeb's head that way. still very easy for them to just hold up their hostage child or whatever tho or just, like, hold up a picture of a child pointsofdata posted:we replaced a vendors "ml" process with an afternoon writing regexes and got better result. seems like a rip off op, I wouldn't buy again. 
|
|
#
?
May 29, 2019 20:34
|
|
|
eschaton posted:ah but you’re forgetting that the ML version will have unpredictable results handling unexpected input! god damnit i wanted this porno to be bi-assed, not biased!
|
|
#
?
May 29, 2019 20:43
|
|
|
animist posted:or just, like, hold up a picture of a child holding up a picture doesnt work because of the hdcp 3d webcam
|
|
#
?
May 29, 2019 21:53
|
|
|
you could just make them go to a mall kiosk to get their face/dick scanned. workers being paid minimum wage and treated like garbage will do a really good job making sure nobody sticks a kid in the machine or make it the apple porno service and charge thousands of dollars and make them go to a salon to get scanned. that's probably the way to go. more money and you can still pay the workers minimum wage.
|
|
#
?
May 29, 2019 22:29
|
|
|
just build a machine learning system to ensure none of the pictures are of kids and by "machine learning system" i mean hire a bunch of human content moderators in whatever country has the cheapest humans
|
|
#
?
May 29, 2019 22:36
|
|
|
i read like 50% of the op before work today will get to the rest soon can you use ml to predict where the s&p 500 is going to be in 3 months and let me know? would like to buy some options thanks
|
|
#
?
May 29, 2019 22:40
|
|
|
the year is 2023. i walk into a mall kiosk labeled PORN-O-SCAN. "one scan please," i say. "i want to watch porn with my face in it."
|
|
#
?
May 30, 2019 01:07
|
|
|
animist posted:the year is 2023. i walk into a mall kiosk labeled PORN-O-SCAN. "one scan please," i say. "i want to watch porn with my face in it." deepfaked porn where "you" are having sex with "natalie portman" is a logical step on the way to futurama-style lucy liu sex bots
|
|
#
?
May 30, 2019 01:12
|
|
|
Bloody posted:holding up a picture doesnt work because of the hdcp 3d webcam oh i meant hold up a sculpture of a child
|
|
#
?
May 30, 2019 01:17
|
|
|
making a startup for advanced fonts called machine kerning. don’t tip your waitstaff I’ll be here until my self driving Tesla arrives to run me into the nearest fire truck
|
|
#
?
May 30, 2019 01:35
|
|
|
animist posted:i can effortpost about deep learning interpretability research if people are interested, that's my area. i wanna read that post
|
|
#
?
May 30, 2019 04:33
|
|
|
Sagebrush posted:deepfaked porn where "you" are having sex with "natalie portman" is a logical step on the way to futurama-style lucy liu sex bots She would be so lucky
|
|
#
?
May 30, 2019 04:58
|
|
|
for those of you who haven’t worked with machine learning, here’s how it’s works: * you take a bunch of pictures of shoes or horses or whatever the gently caress you want to recognize. * then reduce those images to some ridiculous low resolution thing * then you cram those images through some matrices until you get some matrix that happens to tell you horse or not horse. * nobody knows why this matrix is horse, but it seems to work so you’re like, cool, I made a horse model. now you take your horse model and set out to detect some horses. you take some pics and run them through your model and it’s all here a horse, there a horse, no horse there. but then there’s one picture of a horse and your model says no, not horse. and you have no idea why it’s not horse. the matrix is just a bunch of 8-bit values, it doesn’t even look like a horse. the answer, which you will never discover, is that the picture has a reflection of the sun in the water under the horse, which is not where the sun goes and your model couldn’t deal with that. and then you realize that your model doesn’t understand poo poo, it hasn’t learned anything. it is incapable of learning, it’s a loving filter.
|
|
#
?
May 30, 2019 06:12
|
|
|
To me the most damning example is those papers with adversarial techniques where they apply a really subtle filter to the image and it tricks the algorithm. Like you start with an image of a turtle, and the machine recognizes it as a turtle, and then you apply a tiny convolution that affects 10% of the pixels in an almost undetectable way, and the algorithm is now certain that it's a picture of a gun. But it still looks like a turtle to you. So the takeaway is that whatever the algorithm is triggering on, it's absolutely not what humans think of as turtle-like features. It doesn't see things the way that humans do, doesn't have the same "mental" model of the world. That means you can't make any assumptions about its behavior that are based on how humans would treat a given situation. yet that's exactly what people do when they start pressing these systems into service in things like self-driving cars.
|
|
#
?
May 30, 2019 06:32
|
|
|
Sagebrush posted:To me the most damning example is those papers with adversarial techniques where they apply a really subtle filter to the image and it tricks the algorithm. Like you start with an image of a turtle, and the machine recognizes it as a turtle, and then you apply a tiny convolution that affects 10% of the pixels in an almost undetectable way, and the algorithm is now certain that it's a picture of a gun. But it still looks like a turtle to you. wow, if this is true, you should contact the media. could be a big story
|
|
#
?
May 30, 2019 07:45
|
|
|
fart simpson posted:wow, if this is true, you should contact the media. could be a big story unfortunately tech journalists don't want stories that crush their doe-eyed naivety
|
|
#
?
May 30, 2019 08:20
|
|
|
tech journalism has been entirely displaced by tech bloggers. there is no tech journalism anymore.
|
|
#
?
May 30, 2019 08:28
|
|
|
Helianthus Annuus posted:i wanna read that post okay, so, first thing you gotta understand is what people mean by "deep neural networks". a neural network is made up of "neurons", which are functions of weighted sums. thats it. here's a "neuron": code: inputs get multiplied by weights ([10, 300, -.5]), then summed, and passed through a nonlinear function. the function we're using is called ReLU, defined relu(x) = (x if x > 0 else 0). you might think thats a lovely function but it's super common in deep learning for some loving reason of course, neural network people hate labeling things, so theyd actually draw it like this:  which looks more convincing. now, what if you don't know what weights to use? that's easy. pick some random weights. then, steal a dataset of car specs + prices from somewhere. pick a random car and feed its specs to your function. your function will return a price, which will be wrong. so, tweak your weights to make them more correct. this is easy to do, because you know how much the output will change if you change your weights. if the car you're looking at has a wheel size of 10, changing W[0] by 1.5 will change the output by 15. capische? so just tweak all your weights a little so your function's output is a little closer to the actual price. now do this a bazillion times. if you're lucky, your network will now give good estimates for car prices. if it doesn't, you can always add more neurons:  the later ones are sums of the earlier ones, see. what do the ones in the middle mean? idk, but now your network can express more functions. the training algorithm still works the same way, divide-and-conquer style. but that's childs play. thats barely any neurons at all. that's the sort of poo poo you'd see in a neural network paper from the 1980s. weve got gpus now. you can throw as many neurons as you want around, in giant 3d blocks of numbers, each made up of the sums of other giant 3d blocks of numbers. just go hogwild:       these are all state-of-the-art networks. if you take a long time and learn a bazillion tricks to train them correctly, you can get these to give pretty good accuracies for benchmark problems. they were all discovered by, basically, loving around. there's barely any theoretical basis for any of this. machine learning! this post got fuckoff long so im not gonna even post about interpretability. just think about picking some numbers from the middle of those giant networks and trying to decide if they're racist. now imagine your career hinging on getting good results from that. welcome to my grad program animist fucked around with this message at 10:47 on May 30, 2019 |
|
#
?
May 30, 2019 10:28
|
|
|
thanks for the post it helps my imposter syndrome to know that everyone does the same stupid bullshit and not just me op
|
|
#
?
May 30, 2019 10:42
|
|
|
poty posted:thanks for the post it helps my imposter syndrome to know that everyone does the same stupid bullshit and not just me op we are all students of the ancient practice of loving around until something works 
|
|
#
?
May 30, 2019 10:58
|
|
|
poty posted:i read like 50% of the op before work today will get to the rest soon no but if you say you can convincingly enough you'll make a lot of money regardless.
|
|
#
?
May 30, 2019 13:56
|
|
|
btw, ML processors that are the next magical computing wave are just parallel arithmetic engines that operate on matrices. many of them only support 8-bit arithmetic. the point is, whenever you read some article hyperventilating about TPUs and ML chips, you can mentally replace them with floating point units or vector instructions and see if the article still makes sense.
|
|
#
?
May 30, 2019 15:39
|
|
|
everything created by humanity is irredeemably tainted by our biases, be it algorithms or data or maybe machine learning is telling us that racism is a fundamental law of nature, idk
|
|
#
?
May 30, 2019 15:44
|
|
|
journalist: "can you explain what a 'matrisee' is?"
|
|
#
?
May 30, 2019 15:45
|
|
|
whats the most racist number
|
|
#
?
May 30, 2019 15:51
|
|
|
power botton posted:whats the most racist number one thousand four hundred eighty-eight
|
|
#
?
May 30, 2019 15:52
|
|
|
power botton posted:whats the most racist number 104.70.190.51
|
|
#
?
May 30, 2019 15:55
|
|
|
power botton posted:whats the most racist number #ffffff
|
|
#
?
May 30, 2019 16:14
|
|
|
let’s codify some law of robotics for ml algos a robot must be racist at all times and cannot, due to inaction, cease to be racist a robot must deepfake all possible combination of human heads and porn videos except when the deepfaked vid turns out non-racist
|
|
#
?
May 30, 2019 16:48
|
|
|
has anyone developed an ML algorithm that can tell you if the thing you're about to post is problematic?
|
|
#
?
May 30, 2019 18:06
|
|
|
can anyone rec a small ML project or guide for someone who doesn't want to do ML but wants to understand it a little more?
|
|
#
?
May 30, 2019 18:08
|
|
|
a classic is always a twitter bot https://dzone.com/articles/generating-tweets-using-a-recurrent-neural-net-tor
|
|
#
?
May 30, 2019 18:27
|
|
|
i miss markov polov
|
|
#
?
May 30, 2019 18:33
|
|
|
code up a shitposting bot, have it take over ur forums posting duties, write a wired/huff post article “I was replaced by an AI and no one noticed or cared: the pernicious implications of a fully automated luxury gay shitposting bot”
|
|
#
?
May 30, 2019 18:35
|
|
|
DONT THREAD ON ME posted:can anyone rec a small ML project or guide for someone who doesn't want to do ML but wants to understand it a little more? 3blue1browneye has a short series https://www.youtube.com/watch?v=aircAruvnKk&list=PLZHQObOWTQDNU6R1_67000Dx_ZCJB-3pi&index=1 this guy gives a understanding of image processing: https://adeshpande3.github.io/adeshpande3.github.io/A-Beginner%27s-Guide-To-Understanding-Convolutional-Neural-Networks/ and fastai (course 3) is a the best intro ive found. he does a top-down approach tho, so you wont learn the meat and potatos until halfway through. and youre probs looking at a good 80 hours for doing this course https://www.youtube.com/watch?v=XfoYk_Z5AkI
|
|
#
?
May 30, 2019 18:35
|
|
|
|
| # ? Apr 25, 2024 10:06 |
|
|
you could make a shaggar bot but im afraid the limited corpus might give you issues
|
|
#
?
May 30, 2019 18:38
|
|