/cyber/ - Cyberpunk & Science Fiction

The vast majority of images are reposts, indeed the vast majority of content on the internet is reposts. If you're storing all of it, you only need to actually process/identify the OC, which is a tiny proportion of all posts. The image you just posted has probably already been identified and classified elsewhere, so the image recognition algorithm doesn't need to be able to know who the dude is or what his expression means, it just has to match it against the other instances of that image.

Also don't pretend that Google's commercially available OCR is the same as the government's classified OCR.

Why do you think they're developing A.I?

>>35800

duplicated thread

>>35715

>>35802

The purpose of image posting is not for the images to be totally unreadable but to tax the information gathering system.

It's one thing to hand them text files that can easily be processed and it's a completely another thing processing images and trying to figure out their contents.

For example, traditional web scraping becomes almost impossible for a site with no text.

>>35846

>but not too big that the NSA will increase their "TOR-related budjet"

I think its a little bit late for that.

>>35849

Yeah, absolutely. I2P is the way to go… Tor is flawed anyway because of it's inability to handle torrent traffic properly. IPFS is looking solid too.

>>35851

IPFS and Tor are in no way meant to cover the same niche. Tor is for communicating anonymously, IPFS is for distributing files via swarm. IPFS could easily be used to learn information about you since other people are able to figure out everything that you have browsed. Using it over I2P or gnunet won't protect you from this.

I2P is what you would use for torrenting.

>>35846

1) Image to text is never 100% correct (especially on a large scale). On the other hand, text reading is… 100% correct.

2) Image posting is LOW tech but in order to ocr it (especially on a large scale) you need HIGH tech. So, with little effort you can jam the agencies' spy efforts.

bumping.

posting images really is no solution.

it's pretty dumb.

if you don't want the nsa to read your shit then host a private irc on i2p or something like that. there are also people working on decentralized end-to-end encrypted social networking software on top of gnunet and maybe tor/i2p.

>>39225

what board?

>>39261

Damn, I thought it was Lainchan. Kill me.

>>35800

>That's why posting using images only is schway

>implying they don't do OCR over any text in images

>implying maymays with their big fonts aren't perfect for this shit

>>35800

Apparently OP is too much of a shazbot to know about Project Naptha

not sure if OP is trolling or really this retarded.

probably a christian who denies the existence of AI.

>>39415

If AI was that good, CAPTCHAS would be completely useless

>>39415

>probably a christian who denies the existence of AI

*tips cyberfedora*

>>39471

Many of them are worthless. The ones that aren't worthless are ones that are too expensive for spammers to compute.

If you have a bigger defense budget than the rest of the world put together then it's not much of an issue.

Why do you care if the NSA catalogs your data if all you do is post on imageboards?

I'm not saying it's ok for them to snoop on you I'm saying why go to all that bother when you're not even being monitored by them? Like if you were guilty of cyber crimes on a continuing basis then yeah you should take precautions but for your average poster why bother?

> their face when said data is mostly non-text and therefore unsearchable/useless

Doubt it. OCR exists. They can scan audio for voice IDs. Facebook is VERY good at identifying a person in a photo, so I'm sure the NSA could handle that.

just encrypt your shit.

>>39471

>If AI was that good, CAPTCHAS would be completely useless

it is.

remember when that troll company hired halfchanners to solve recaptcha for their botnet but in public they claimed they built an AI that solved 90% of captchas?

(at least that's how i imagine the story went)

google wanted to do that too so they built a real AI that could solve captchas at 99.8%.

www.zdnet.com/article/google-algorithm-busts-captcha-with-99-8-percent-accuracy/

that's why they kinda deprecated reCAPTCHA and instead use AI to guess who is a bot and who isn't just like the US uses AI to guess which mobile phone to send a drone at.

That's why twitter did it right. They made fashionable the use of metadata tags. And now everyone else is following, including facebook.

In a future, the odd image with no hashtags will be looked in derision and some subcontracted indian will tag it invisibly for you.

The solution is not to use neither twitter nor facebook nor google.

Isn't Google worse than all of that?