Hi, I've developed a prototype neural network which detects the banana cancer image with 90% accuracy. As we all know, the banana meme is cancer and by using the power of neural networks (powered by Tensorflow) we can use it to autoban banana posters
Also, big thanks to the guy who collected those thousands of banana images, you did a great job of putting together the dataset for the neural network and I didn't even have to lift a finger.
I'll provide source code and dataset once it reaches 98% accuracy. Maybe the admin might be interested in integrating this into 4chan?
It could also be adapted to counter all kinds of spammers and image flooders; spiderman, desu, sans, etc. The spammers and flooders are the biggest cancer of 4chan, and with neural networks we can put a stop to it.
>>514658
The servers struggle to maintain a fucking word filter
and your solution is a god damn AI which will take over the world through 4chan.
Thread theme:
https://www.youtube.com/watch?v=XU7VtN6txmQ
>>514661
moot works with google though and they got neural networks.
Let's ask him to help implement
:^)
>>514664
It would also be pretty important for the network to track your preffrences too.
4chan would be great if there was a large
[RECOMENDED THREADS]
Below the banner.
>>514703
You banned the trump wall picture too. lol
>>514658
kek I thought about this being a thing in the future one time, and now someone actually did it.
It's pretty cool OP but still a very bad idea, even if it could be practically used.
This isn't the direction 4chan should be headed in. The more I think about this I'm starting to realize that tightening the reigns on anonymous users who can easily evade bans just makes everything worse and worse.
Unless you can get rid of people who have the intent to abuse things (which even includes high quality posters occasionally, everyone gets in weird moods sometimes) then by treating the users more and more like children you're going to get more and more childish behavior.
We really need to be putting our effort into giving the people who want to do good things, creative things, fun things, the features they need to do so, instead of putting all our effort into figuring out how to stop the people doing things we don't like.
I don't think 4chan has done that in many years now. Even Hiro is getting caught up in it I think, he made /trash/ to house the trash. He made /qst/ to get quests off /tg/.
We're all always looking at the negatives. We need to focus more on the good aspects of 4chan.
Let's please stop babysitting and trying to childproof 4chan. Treat them like adults and maybe they'll act like it. Just let them post what they want.
>>514722
This guy gets it
>>514722
The other option I was thinking of was a locally ran neural network which either would be a firefox/chrome plugin or background program, which would simply hide the banana images automatically, and gives the user control over what they want to see and not see.
However, im not sure how other computers would handle it. I'm using Nvidia cuDNN which greatly speeds up neural etworks, but I havent tested it on anything without an Nvidia chip. Though Tensorflow is rather scalable, so it can be customized to whatever platform its running on.
>>514760
>and gives the user control over what they want to see and not see.
People shouldn't be given this right. People are retarded and try to turn everything into their personal hugboxes instead of maning the fuck up and learning how to be tolerant and ignore shit.
>>514770
Kil yourself bananafag.
>>514658
Isn't using a NN a bit overkill for this task? How many false positives does it generate?
Have you thought about using a perceptual hashing method like pHash instead? I feel like this would be much more lightweight.
>>514661
Maybe users could run an instance of the program locally and automatically hide any post with the image attached.
