My friend wants to know when she can use this to generate porn, are we close?
Currently waiting to scale-up for improve quality mainly for economic reasons, not quite sure I could recoup the training costs yet. Even more so if I go with cloud training.
NVidia will release the 4090 in september, and ethereum may do "the merge" that will make GPU useless for mining so GPU price could be affordable so I can update my home cluster with affordable 3090s. (But electricity prices are also up).
Also there are new algorithms every month like the stable diffusion, that would obsolete your previous training.
The video generation cost is probably still too expensive compared to just paying a cam girl in a low wage country. But it will probably go down soon.
This is also some sensitive data, as plagued with copyright issues, so it's quite troublesome to legally share training datasets to share costs.
It also has its own challenges with respect to custom dataset creation with text description, so it's probably a better idea to adapt the algorithm to the currently available data to keep the costs low.
Finally once someone releases a model, in the next month there will be at least 3 clones.
There is also the problem to find an adult friendly payment processor.
And the multitude of potential legal issues.
But it's probably inevitable.
So no, your "friend" can't use it for that.
A porn model should really be it's own thing.
Since we certainly don't want to show generated image of porn or violence to someone that didn't specifically ask for that, the easiest way to ensure that's not going to happen is to just not train on that kind of data in the first place. The worst that can happen with a model trained on "safe" images is that the image is irrelevant or makes no sense, meaning you could deploy systems with no human curator on the other end, and nothing bad is going to happen. You lose that ability as soon as you integrate porn.
Also with techniques like in-painting, the potential for misuse of a model trained on porn/violence would be pretty terrifying.
So the benefits of training on porn seems very small compared to the inconvenience. I don't think it's anything to do with puritanism, it's just that if I am the one putting dollars and time to train such a model I am certainly not going to be taking on the added complexity and implications of dealing with porn to just to make a few people realize their fetishes at the risk of my entire model being undeployable because it's outputting too much porn or violence.
uh have you seen American/European mainstream pornography? it's already pretty violent (ex. face slapping, choking, kicking, extreme bdsm).
I just don't see why this stuff is allowed and protected by the law (if its not recorded and published its illegal) and then we are suddenly concerned about what text can do.
Just one of the many double standards I see in Western society.
Someone like Trump would just shrug and say the pee tapes are deepfaked. I don't think its possible for AI to bypass forensics either. So again this narrative that "deepfake blackmail" would be dangerous makes no sense.
Of course, after a while it'll probably stop working, but there will be a period of time where it can be done profitably and a longer period where it will be obnoxious.
And, of course, you could probably always use the tool to scare children, who, even in the future, might not know that everyone would shrug off the generated pictures.
...and the more people will be aware of and stop believing in the "fake reality".
Ensuring this technology is only available to a tiny subset of the population is to essentially give all the power of distorting reality to that tiny group of people.
In fact, I suspect that is precisely the reason.
That's not at all what I am talking about. What I am saying is that such a model would give everyone the ability to create extremely realistic fake images of someone else within a sexual/violent context, in one click, thanks to inpainting. This can become a hate/blackmail machine very fast.
Even though Dalle-2 is not trained on violence/porn it still forbids inpainting pictures with realistic faces that have been uploaded by users to prevent abuse, so now imagine the potential with a model trained on porn/violence.
Someone is eventually going to do it, but back to your initial question about why it's still not done yet, I believe it's because most people would rather not be that someone.
Make no mistake this is an indirectly legal hurdle.
There are attempts to gather porn images and train or fine-tune existing networks on it, here's a recent attempt by an art student mentioned in the article above (NSFW!!): https://www.vice.com/en/article/m7ggqq/this-furry-porn-ai-ge...
However, if it is possible to re-start the training process from the weights of a non-sexually aware model, this finetuning might not take all that long..!
Making money at it should be easy, and places like PornHub wouldn't care about any outrage. The real challenge would be limiting criminal and civil liability, at least to my not-in-the-business thinking.