Just a heads up: This project can be followed on twitter @GenesVision
Hello anons. After getting tired of swiping and getting latinos and all sorts of mutty filth, I decided that technology today is good enough that we can solve this problem manually. Today, I put the finishing touches on a general adverserial neural network ("GAN"), a type of neural network ("AI"), that is particularly good at identifying traits in images. Basically what you do is have a bunch of tagged data of traits (e.g. blond hair, weight, height, lips, race, make-up, and 65 other markers) meaning it can detect 71 traits in a person with its current set of training data.
The network it uses was actually a learning experience to implement a GAN from scratch in C#, but then I decided to evolve it. It automagically swipes, and saves, any girls that match your preference. Tinder has a bunch of interesting data for each profile, including music, Instagram account, etc. that can be used for further investigation. It has a pretty good accuracy. I went through a sample size it had picked out from me of NOPE/YES, and it had an 96% accuracy (144 correctly classified out of 150 total). All were identified correctly as being white, there was just some makeup flaws that it misdetected 6 times. I do not like make-up, especially not excessive make-up.
I'm thinking of making this a SaaS, since it's too GPU-heavy to run in the browser, I cannot make it an open-sourced extension. I can make it an extension that contacts my GPU-farm and does the work for anons. I wonder if this is something /pol/ would be interested in.
I call the projects Tinder E-ugnics. My personal background is a masters in Computer Science working towards a PhD.
Ok I'm interested lol Do you plan on implementing this so you can run it on your phone? Also, why does it use GPU?
Oliver Thomas
It uses GPU to train the network continuously as it gets more data (images) from Tinder and other sources taking up hundreds of gigabytes of VRAM. Forward propagation (the act of classifying a picture) is less GPU intensive, but the model (the file it outputs, that is the classifier network) has a tremendous size.
Evan Reed
Excellent. You're gonna get busted by tinder though
Evan Carter
Nope. The AI swipes in a manner that is indistinguishable from humans. They'll have no way of knowing.
Daniel Allen
Ahh I see... Any plans for trying to move towards a more mobile usage?
Jordan Lewis
So no api to access tinder? Can you filter on text as well?
Isaac Martinez
There will be mobile usage - however only for Android, or jailbroken iPhones. Follow @GenesVision to get the software as it's released (alternatively the source code).
Kevin White
A solid idea and will def follow. Twitter?
Jacob Thomas
>I'm using a generative adversarial network >Isn't generating anything You don't need an adversarial network for trait recognition you goober.
Logan Powell
Yup. :) Some leftist cuck has followed me, make sure to retweet what you see if you have a based right-wing following.
Isaac Davis
Lol just noticed you said general adversarial network. That isn't what GAN stands for my friend.
Daniel Cooper
Oh my god. Will you please just Google before being retarded.
Lol. I've been following ai research for the past 5 years. I know what a fucking GAN is. Maybe you should Google it? lmgtfy.com/?q=What is a GAN
Landon Lopez
ML (machine learning) algorithms will not run on any phone, they run on very high-end PCs. a lot of people that use ML actually rent machines and send the data to that other computer over the internet and have that other high-end computer to the actual computation
this would have to work by you sending a list of profiles or a list of pics to OP's server, and then that figures out the 'type' that you want, and then scans tinder for that 'type' and returns to you a list of relevant profiles, probably sorted by distance to you
i don't see how any of this would be illegal or anything, as long as you don't charge money for it. if you do, then it might be a hairy situation, don't know much about the law. an interesting idea, though.
Nolan Diaz
It's literally the first hit on Google when you search for gan
Dylan Perry
It says generative not general. And you would most likely want to be using a convolutional neural networks. There is no need for an adversarial network in this case.
Brody Perry
If you'd read the post it's my own implementation from scratch. It determines whether the points on the face was placed correctly, and the discriminator removes them if it determines them to be incorrect. It has patterns of CNN in there.
Again for anyone reading: Follow @GenesVision on Twitter, and retweet the 2 tweets. The more we can get this out, the more data we have, the more accurate the service becomes.
You know the girl in the OP is wearing make-up don't you retard?
Grayson Perry
The neural network detected that with 91% certainty.
Landon Morales
This is interesting bump
Isaiah Hernandez
Based on your assessment, right? I also don't really like make up I think it's fake but I don't care too much
Hunter Walker
Cool idea but good luck putting this on a resume. Even though this is cool tech everyone is going to judge the shit out of you for creating this.
Thomas Murphy
Point is all girls wear make up. What's the issue is when they wear too much and get clown like. Almost all wear some eyeliner or cover a blemish or too.
Nathan Smith
i'm in the same boat. i make a lot of tampermonkey scripts/browser extensions, but 80% of them are for the porn sites i use.
so i mention on my resume and in my cover letters that I write browser extensions, but only show the few i have that are work safe and that i wrote a year ago. it looks kind of silly, but, ultimately, i have the skills of a mid-to-high tier front-end developer from all of the porn scripting, and that's what matters.
Jack James
My CPU is a neural net processor, a learning computer.