Just stumbled over the video below. It is mostly about ChatGLM6B, a model that is somewhat similar to ChatGPT, used 6 billion parameters and runs on RTX 2080 (Ti?). That’s certainly only for inference, when it is used. For training, more beefy hardware is needed.
It still shows how much progress is being made and that running very good models on hardware that many Blender user may have access may be an achievable goal.
Yes, I am actually surprised by how fast it’s progressing. Just a few days ago I’ve played with 2 models: vicuna 13b and alpaca 30b. Both are LLaMA based, were quantized to 4bits and fine tuned on gpt data. They are both really good, and the first one runs on CPU (i.e., almost any hardware, with more than 10gb of system RAM) and the 2nd on 3090 (there are 13b and 7b version that can run on less VRAM).
Here is how fast Vicuna 13b 4bit works on a CPU:
And here’s how fast Alpaca 30b works on a GPU:
(they are not as good as chatgpt, and for sure way behind gpt4, especially when it comes to code stuff. But it’s already very, very impressive what they can do)
This video was generated from inside the Blender Video Sequence Editor with the Generative AI add-on for the VSE: Pallaidium - genAI in the Video Sequence Editor
I matured some opinions.
I think we are getting scammed. This framing of the issue is fraudolent. basically while we focus on “AI vs Artist” we are not paying attention to another more important question. “why only my social mnagers should have AI and I cant?”
we all know that amazon, google, fecesbook and all the other social media, use AI and enjoy the unregulated situation, there are few lawyers in total with specialization on digital rights and the tech world so these megacorps have lots of freedom to what they want
the fact we dont know what the various AIs are doing, is also another major concern because we have no means to even decide if this is a threat or not.
as for now, the only good thing to do in my opinion, is getting our hands dirty scavengng info on AI and trying to learn what we can about it. All we discover, in any case, is only what we are allowed to discover.
the topic should be framed in this way: We want AI for the masses, nobody has a right to have exclusive usage of this very dangerous technology, we have to do everything we can to give AI to the masses or at the very least to make it widely available or easy to have access to.
Another big distraction is about “AI and licenses”… which is completely irrelevant. like any other tool that does artsy stuff, AI will eventually be regulated, it will be a market process, like in the history of cinema happened years ago… Nobody can really expect law to be fast… law in general is slow and requires slow time to adapt… no art medium was born with a set of laws that defined it.
Could you explain a little bit more in detail what you mean by that and more importantly how it could be achieved? And what kind of AI solution exactly are you talking about?
Unfortunately i had to remain vague. because basically I lack basic law competence so I rely on comparisons. The comparison I did was the cinema industry. Because in the history of cinema there was this issue of licenses. The professor I was listening to, told me that basicaly “greed” and “crass vile willingness to make money” created the premise for a great art => cinema… and laws emerged, licenses were created out of market necessities… its a phenomenon that we can observe now and we know that it took lots of time. Heres then how I solve “the issue of licenses”… I just say that eventually it will be figured out.
As for the “nature of art and AI”… I think this is a mega distraction, because first we purposefully avoid to discuss “art”… And I suppose we avoid this discussion because its convenient for us… Basically the word “art” was separated years ago from “decoration” and its also separated from technicalities and artisan profession… what it means right now is sort of a pretentious declaration that you are “intellectual” and you have to be “philosopher”… its a circular definition. Art means “what me and my buddy intellectuals agree in this context to define as such”.
So for me both these topics are completely irrelevant. Therefore the “AI vs Artist” discussion is also a topic that is irrelevant.
Instead what is relevant to me, is the idea of owning an AI, and use it. I am lowstatus (it means I dont have the mony to develop the AIs), but I would like the opensource AIs to remain as such so even a guy like me could have access to it. Reason: I think given the premise that our social managers use AI against us, it would be a good idea to learn the tool best we can, also to increase our decision making skills…
I think this is more important: should only a selected few have AI and we dont have it? Becuase if you stay here arguing on humanitarian premises and constructing wacko ideas that focus on a sense of drama and you talk about “oh poor artist he sees hmself replaced”, maybe you satisfy your heroism you get a sense of drama where “the artist” is “good” and AI is evil, but you also avoid discussing the more important issue, which is the fact that AI exists even if you dont like it… and facebook, amazon, apple, theyre all going to use it even if you dont like it.
In my opinion at this point is best for us to all get into AI, and use it. And also learn how it works.
Sure, open source AI is great. But even if they are open, they usually require some quite beefy and expensive hardware. Even if you have all the necessary code and trained neural networks to run it, it is not guaranteed you could use it.
Could you be as specific as you can with this, preferably with an actual example? Because I have no idea what you are talking about.
2 obvious examples would be tiktok and fecesbook
there would be also google, who uses every sort of “AI” since lots of time. Amazon uses it.
These are just the most obvious examples, that you have under your nose and you can see with your own eyes.
So, yes, all I wanted to say is that if you keep focusing on the construction of this sense of drama, you just fulfill an emotional need… I know we all have this need for drama, myself included, we want to dramatize everything… “Oh I am a good guy, I am an artist, I am so profound and human but AI is evil and is stealing my lunch money”… I get it, we want drama, we want escapism and we want to see ourselves as the hero of the story, we like to have this storytelling narrative… fine, but its leading us all to the wrong direction and we miss the mark.
I am very confused. Am I understanding you correctly that you want those to be open source too?
I try to be reasonable.
I do want them but I obviously recognize their AI tools cost money so I can’t have them.
I want to go to the moon, I have a right to go to the moon. But nobody owes me the means to go there. I hope it’s clear.
How do you figure? Space travel isn’t part of the Universal Declaration of Human Rights as officially recognized by the United Nations 
That is to say, there’s no organized government or regulating body existing in our current society that would formally recognize your desire to go to the moon
Its a general thing. I hope you don’t take it literally.
If you say “I have a right to X”, I assume you mean “I have a right to X”. That doesn’t really feel like much of a stretch in understanding 
Am I understanding you correctly that you are referring to some sort of democratization of AI?
Let’s put it this way, it would be easier for me to enter argument to defend my right to go to the moon, rather than defining the whole premise to create this narrative of drama of “artist vs AI”.
Because this sense of drama about poor artist and his lunch money, has no legitimacy, and for as absurd as it may seem the claim that I have a right to go to the moon does have legitimacy. I could talk about it, obviously I prefer not to, because that would derail the topic, but maybe another time we can talk about it.
It doesn’t have legitimacy. You have no right to go to the moon. A right means- in legal terms- that anyone denying you that privilege or opportunity can be held legally responsible. No one can be held legally responsible for preventing you from going to the moon.
If you’re going to enter a discussion using legal arguments, you need to actually know what those legal arguments mean if you want people to take your contributions seriously.
i disagree but I leave you the final say because afterall it was me who brought up this example. perhaps I should have used another example. Lets just forget about it, or we will enter into argument for an example I did.
yes sort of . Could be good idea.
obviously we realize that AI is expensive, its costing loads of money, we can never expect to have access to it fully…
Fine, i think we still have to get our hands on it one way or another… I accept it if you start with your idea of “democratization”, I think it coudl be one good way to look at this issue. Why not? I certainly have nothing against.
I am not trying to throw my idea into the pot, but I am trying to understand your view. I have been asking clarifying questions to understand what you mean, to understand what you think is the issue.
Is there some specific issue in your view? Is there something specific going on which prevents AI solutions of being more publicly available? Are the necessary tools not available to the public? Should there be public funds to push for more open AI research?
i dont have the solution. I like some ideas of richard stallman, but I understand theyre too “extreme”, if we talk about “solutions”. I do like the idea of stallman of using universities to “reverse engineer” and rob knowledge, I like his idea of promoting hacking to steal knowledge.
Those are all sparse ideas, many people more prepared than me already try solutions and many of them fail. If I knew how, I would already be a guru of some sort, but I am afraid I dont know.
it is my personal opinion that we could achieve a lot if we did vounteer work, maybe 4 hours a week or even less, but I also know thats never going to happen. I dont have “a solution”, just sparse ideas.
I hope youre not going to press me to “find solutions”, because I found out about AI quite recently, and I am still in the process of intaking info.