this post was submitted on 13 Sep 2023
351 points (96.8% liked)
Technology
58303 readers
7 users here now
This is a most excellent place for technology news and articles.
Our Rules
- Follow the lemmy.world rules.
- Only tech related content.
- Be excellent to each another!
- Mod approved content bots can post up to 10 articles per day.
- Threads asking for personal tech support may be deleted.
- Politics threads may be removed.
- No memes allowed as posts, OK to post as comments.
- Only approved bots from the list below, to ask if your bot can be added please contact us.
- Check for duplicates before posting, duplicates may be removed
Approved Bots
founded 1 year ago
MODERATORS
you are viewing a single comment's thread
view the rest of the comments
view the rest of the comments
My point is that this description literally applies just as much to humans. Humans are also trained on vast quantities of things they've seen before and meanings associated with them.
This is genuinely a misunderstanding of how these programs work.
Because the only art anyone has ever done is when someone else paid them for it? There are a lot of art forms that generally aren't commercially viable, and it's very odd to insist that commercial viability is what advances an art form.
I do actually get regularly paid for a kind of work that is threatened by these things (although in my case it's LLMs, not images). For the time being I can out-perform ChatGPT and the like, but I don't expect that that will last forever. Either I'll end up incorporating it or I'll need to find something else to do. But I'm not going to stop doing my hobby versions of it.
Technology kills jobs all the time. We don't have many human calculators these days. If the work has value beyond the financial, people will keep doing it.
Human brains don’t have perfect recollection. Every time we retell a story or remember a memory or picture an image in our head it is distorted with our own imperfections.
When I prompt an AI to create an image it samples the images it learned from with perfect recollection.
AI does not learn the same way humans do.
This is incorrect actually. The models these AIs run from by definition have imperfect recall otherwise they would be ENORMOUS. No, that's actually exactly the opposite of how these work.
They train a statistically weighted model to predict outputs based on inputs. It has no actual image data stored internally, it can't.
This is incorrect actually. The models these AIs run from by definition have perfect recall and that is why they require ENORMOUS resources to run and why ChatGPT became less effective when the resources it was allocated were reduced.
-ChatGPT
No, they take exponentially increasing resources as a consequence of having imperfect recall. Smaller models have "worse" recall. They've been trained with smaller datasets (or pruned more).
As you increase the size of the model (number of "neurons" that can be weighted) you increase the ability of that model to retain and use information. But that information isn't retained in the same form as it was input. A model trained on the English language (an LLM, like ChatGPT) does not know every possible word, nor does it actually know ANY words.
All ChatGPT knows is what characters are statistically likely to go after another in a long sequence. With enough neurons and layers combined with large amounts of processing power and time for training, this results in a weighted model which is many orders of magnitude smaller than the dataset it was trained on.
Since the model weighting itself is smaller than the input dataset, it is literally impossible for the model to have perfect recall of the input dataset. So by definition, these models have imperfect recall.
In other words they require exponentially more input because the AI doesn’t know what it is looking at.
It uses its perfect recollection of that input to create a ‘model’ of what a face should look like and stores that model like a collage of all the samples and then uses that to reproduce a face.
It’s perfect recollection with an extra step.
Well, what you described is simply not a perfect recollection. It is many small tidbits of information that combined together can make a larger output.
That's exactly how our brains work too
If our brains worked exactly the same as AI programming then AI wouldn’t be needed because it would be no different than how we are doing things without AI.
I feel like you keep misrepresenting what I'm saying. Nowhere did I say that our brains work completely and exactly the same as AI. However, we do learn in much the same way. By amortizing small amounts of information and drawing connections between them
I'm pretty sure that the way they constantly fuck up hands is a solid demonstration that these AI tools do not have a perfect recollection
The reason they fuck up hands is because hands are usually moving during pictures and have many different configurations compared to any other body part.
So when these image AIs refer back to all the pictures of hands they’ve been fed and use them to create an ‘average approximation’ of what a hand looks like they include the motion blur from some of their samples, a middle finger sticking up from another sample or extra fingers from the sample pictures of people holding hands etc and mismatch them together even when it doesn’t fit in the picture being created.
The AI doesn’t know what a hand is. It is just mixing together samples from its perfect recollection.
What? No
How many pictures do you see online where the hands are in motion, or even blurred?
Hands are usually behind objects when they hold something and can indeed have tons of variations and configurations. Even human artists fuck up all the time or just not draw them at all.
AI don't combine samples. If they did they wouldn't be able to generate new pictures of whatever subject you want in a specific style you want and then have multiple variations of that picture.
It isn't a copy and paste, it is interpreting the drawing and modifying it based upon the prompt.
In which case the machine would get the copyright (which legally they can't now), not the prompter.
Copyright just isn't compatible with AI, we need to abolish it.
If a picture gets generated, who is the owner? The one writing the prompt? The AI that generated it? The researchers that created the AI? The artist on which the picture is based?
How about none of them? It is a picture, a piece of information. It doesn't need an owner.
Can we get UBI before we start abolishing people's income though?
I agree. Well, that is assuming there's no human editing of the results of the AI tool afterwards. There was heaps of it in the piece referenced in the article, and there usually is if you want to get something actually good. The piece referenced was entered in to a photomanipulation and editing category too, which seems like it's very much in keeping with the spirit of the competition. But the reason I said that was because the comment I was replying to wasn't about who has the copyright of the tool's output, it was about the value of the output and tools in general
The tools are valuable for sure.
Where the law is on copyright it looks like we're figuring out. For now I'm glad to see rulings like this as it will, hopefully, take some of the wind out of Hollywood studios and aide union negotiations.
If there was, then the artist should have discussed those heaps of human editing that went into the creation of this piece of art, and he would have been granted a copyright.
The fact that he refused to disclose what - if anything - was done after the AI spit out the result is what resulted in him not being granted copyright.
He did? This article mentions it only briefly, but he talked about it more when it was first getting attention for winning the competition. Is this something he did in the court case that you've read elsewhere?
But also, if you used Midjourney at the time that the image was made, you'll know that you did not get an image like that straight out of it
This wasn't a court case.
This was a copyright application.
The Copyright Office asked him to provide them with an unedited version of the image generated by Midjourney in order to determine how much (human) work went into the final version.
Allen refused to provide them with an unedited version, so the Copyright Office had no way to verify how much or how little work was actually done by the artist compared to work that was done by the AI, so they had to assume that the vast majority of the work was done without any human artistic contribution.
They were essentially forced to reject his copyright application because he refused to provide evidence that he actually did any kind of creative artistic work.
What? Humans don't learn to paint by looking at paintings, most people learn by just painting. Humans can also draw art without having ever seen any. AI on the other hand can only draw from other people's works, it has no creativity of its own.