in fact, this green text was made purely from asking chatgpt what ai will look like in 10 years
Unironically the best greentext I ever read was the bottomless pit one written by AI
That was like 3 years ago when generative AI was fun and whimsical
The first ai green texts made me laugh so much. They managed to perfectly capture the essence of a green text but because they were dumb they would create the most weird situations.
The last time I had fun with LLMs was back when GPT2 was cutting-edge, I fine-tuned GPT2-Medium on Twitch chat logs and it alternates between emote spam, complete incoherence, blatantly unhinged comments, and suspiciously normal ones. The bot is still in use as a toy, specifically because it’s deranged and unpredictable. It’s like a kaleidoscope for the slice of internet subculture it was trained on, much more fun than a plain flawless mirror.
much more fun than a plain flawless mirror.
yeah agreed! Back in the day I used to generate text for fun with n-grams and I never went higher than bigrams bc it was boring without those unexpected disfluencies. I thought of it being like an electric guitar, you want it to sound a little raw.
Link?
Reputation and PGP signatures could be used to verify real human made content. That is, of course, if people actually care, which I think will be rare.
There might be no-ai communities, that require this and are closed down to avoid being scraped for ai training.
Edit: Also AI is already enshittifiying itself, which might get worse if it becomes more widespread than it already is.
Why couldn’t AI use PGP signatures that suggest they’re human?
Trust is the most important part. You trust someone they made something themselves. They digitally sign their work with a public key that is known to be theirs. You can now verify they (the person you trust) made it.
Once the trusted creator’s key is leaked, they are no longer trusted for future works.
AI made content can be freely signed as well, but if you don’t trust the origin, the signature doesn’t matter anyway since it will just verify it is coming from the AI creator.
The key thing is trust, the signature is just there to verify.What is stopping a human from gaining a reputation with a signature, and then selling it (a la reddit accounts)?
nothing, however the signature could be untrusted after it is discovered that it was sold/stolen
Its not about “just having a signature”. Its about a web of trust. It only works if you verify if the key belongs to a creator that is actually a person.
Basically creators go to a convention and hand out their public key in person and have other creators sign their key. If you trust creator A is real and they signed the key of creator B, you can have some trust B is also real. And if your buddy went to the convention, met A and B, got their public keys and tells you they are real you can also trust they are real. The more steps/signatures you are away from a creator the less trustworthy they are and nothing really ensures a (human) creator doesn’t use AI secretly. If somebody is found to be a fraud everyone has to distrust their key.
Throw the technical bit away. Just think of it as a signature. Yes, an old school, written-with-hand signature.
Does the bank trust me giving you $100 by you having this cheque? Yes. Why? Because I told them what my name is and what my signature is like.
Will the bank give you $100 if you stole my cheque and sign your name on it?
I wonder if personal websites with links to each other, like in the olden days, will start growing in popularity again because of how trust is slowly eroded for anything not in your direct control, and search engines becoming more and more useless 🤔
Before the invention of video, humanity didn’t have video evidence either and still managed. We are approaching the end of a ~150 year time period in the history of humanity in which video evidence is persuasive.
Skynut is coming, in all its smutty glory and we can do nothing about it
Photorealistic porns? What’s your problem, man? You have realistic AI and this is all you’ll have? Just order a silicon doll and put an AI chip into it! Free
sex-slawife!Wouldn’t be against that
I love my computer, you make me feel alright! every waking hour, and every lonely night! I love my computer, for all you give to me! predictable errors, and no identity! and it’s never been quite so easy, I’ve never been quite so happy!!! all I need to do, is click on you, and we’ll be joined in the most soul-less way! and we’ll never ever ruin each other’s day, cuz when I’m through I just click, and you just go away! I love my computer, you’re always in the mood! I get turned on, when I turn on you! I love my computer, you never ask for more! you can be a princess, or you can be my whore! and it’s never been quite so easy. I’ve never been quite so happy!!! the world outside is so big, but it’s safe in my domain, because to you, I’m just a number and a clever screen name! all I need to do is click on you! and we’ll be together for eternity, and no one is ever gonna take my love from me, because I’ve got security, her password and a key!
Video evidence is relatively easy to fix, you just need camera ICs to cryptographically sign their outputs. If the image/video is tampered with (or even re-encoded) the signature won’t match. As the private key is (hopefully!) stored securely in the hardware IC taking the photo/video, any generated images or videos can’t be signed by such a private key.
So whatever way the camera output is being signed, what’s stopping you from signing an altered video with a similar private key and then saying “you can all trust that my video is real because I have the private key for it.”
The doubters will have to concede that the video did indeed come from you because it pairs with your key, but why would anyone trust that the key came from the camera step instead of coming from the editing step?
You, the end user, don’t have access to your camera’s private key. Only the camera IC does. When your phone / SD card first receives the image/video it’s already been signed by the hardware.
so you want the hardware to be significantly more opaque and almost impossible for new manufacturers to compete?
It’s pretty standard practise these days to have some form of secure enclave on an SoC - Arm’s TrustZone, Intel’s SGX, AMD’s SME/SEV. This wouldn’t be any different. Many camera ICs are already using an Arm CPU internally already.
You can enter the camera as evidence, and prove that it has been used for other footage. Each camera should have a unique key to be effective.
So if you create a new key, it won’t match the one on am existing camera. If you steal the key, then once that’s discovered, the camera should generate a new one.
But if you don’t actually check the physical camera and prove that key for yourself, then it can easily be faked by generating a key that is not coming from the camera and is used for the “proof” video and the fake video.
Mate, digital cinema uses this encryption /decryption method for KDMs.
The keys are tied into multiple physical hardware ids, many of which (such as player/.projector ) are also married cryptographically. Any deviation along a massive chain and you get no content.
Those playback keys are produced from DKDMs that are insanely tightly controlled. The DKDM production itself even more so.
And that’s just to play a movie. This is proven tech, decades old. You’re not gonna break it with premiere.
But how would one simple member of the audience easily determine if this whole chain of events is valid, when they don’t even get how it works or what to look out for?
You’d have to have a public key of trusted sources that people automatically check with their browser, but all the steps in between need to be trusted too. I can imagine it is too much of a hassle for most.
But then again, that has always been the case for most.
This is for restricting use, not proving authenticity of the videos recording. Anyone can spin up keys and sign videos, so in a legal battle it would be worthless.
The technology would be extremely easy to adapt, with the certs being tied to the original recording equipment hardware. Given i don’t see a $60 ip cam having a dolphin board it would probably be relegated to much higer end equipment, but any modification with a new key would break the chain of veracity
This is blatantly not true, it would be extremely simple to circumvent. How do you “tie” the cert to a specific hardware without trusting manufacturers? You just can’t, it’s like putting a padlock on a pizzabox.
I personally doubt that will happen, since the current models require a lot of data to get better, something we actually don’t have. The real danger is what happens once we figure out how to make models without an absurd amount of data.
Making models without a mountain of data is just engineering lol. That’s what we were doing before, are still doing, and will continue doing for the rest of our existence.
This guy bought so many rare monkey tokens. Ai is impressive in some aspects, but it’s not nearly as impressive as the marketing that drives the massive amounts of investment into it.
The US economy is doing anything it can to create growth, which is causing investors to create a bubble around AI that is “too big to fail”.
Read classics:
Pride and Prejudice" by Jane Austen, “Moby Dick” by Herman Melville, “The Great Gatsby” by F. Scott Fitzgerald, and “1984” by George Orwell.
Start here. There are thousands.
Why start there with British and US authors? Why not 100 years of solitude, Disgrace, and dream of red mansions?
Most readers would gladly read AI slop instead of real literature, and thousands likely already do. Just look at how much brainrot gente fiction is pushed on “booktok” and the now common practice of choosing books by tags only.
Also, for AI porn, it’s already all over /b/
