pouët.net

If you devotedly follow and foster the AI hype and display loyalty to entities behind it your faithfulness will be pontifically rewarded on the day of singularity.

(some rumors even mention 99 virgin droids)

As much as I am not a fun of using commercial engines for demos, at least I recognize that there has to be some prior knowledge for someone to do anything with them. Tell someone who is illiterate with programming to download Unity/Unreal and do something, even a basic thing, even a rotating cube, it would take time even to me to figure it out. But with AI art, it seems like it's made for "idea guys". Even if the first results after putting text prompt are obvious crap, I know as a non artist it would take me to learn how to paint even that in photoshop or something.

As much as it doesn't bother me too much, there is something that feels off with AI art. The equivalent for coders would be a text field where the "idea guy" could describe a complete demo/game and get something that works, without ever being able to code it on his own. Atm even doing demos using in Unreal/Unity, an "idea guy" without any skills wouldn't survive.

loady: and your written exam on 'machine learning & image synthesis A' is in 7 weeks!

Optimus: the thing that is overlooked IMO is that ideas tend to take on shape, and new ideas emerge during the process of making something. You have an idea that is more or less fuzzy, and you start exploring it and making it concrete, and during the process it evolves, takes shape, until you’re left with a well-defined concept. With AI, you get a result instantly, and lose the part where your ideas evolve and change. It’s shallow and short-sighted to outsource your thinking to an algorithm if you ask me… but that’s exactly what loads of people are doing, and proudly showing off.

[query]With AI, you get a result instantly[/query]

Only if you are lazy. But we're not, are we?
I would never call "a result" the first thing an IA proposes.

Damn typos.

REPEAT



Only if you are lazy. But we're not, are we?
I would never call "a result" the first thing an IA proposes.

Seems lazy to me that you don’t follow up and tell me what you would call it ;)

@farfar: It depends. I can call it a sketch or first step or inspiration for a specific aspect of an idea. But I would never call it a result because, like you, I think that the result comes after a process of dialogue with oneself. And within that dialectical process AI can enter but as a tool, not as an "artist" since the decisions are made by me.

It is amazing to see how strong are the opinions on AI art of people who clearly have no clue about the technique/process involved.
It is almost as if sometimes correlation does imply causation.

@introspec: It is perfectly possible to have an opinion on the social implications of a technology without being an expert of its exact inner workings. I also believe that you were the first to use the word 'idiot' in this thread on page 3 in a sorry not sorry fashion, why I choose to take your views on the matter lightly, researcher or not.

Havoc: since it's ai art, that logo needs two or three more "fingers" attached :)

I'm having fun replacing "artist" with "digger" and "AI" with "excavator" in these paragraphs - and the discussion in general. :)

Ham: right, then I think we largely agree :)

@El Topo:

I absolutely agree with you that anyone is entitled to have an opinion about social implications of any new technology, AI or not AI. However, I do find it very annoying when people try to support their opinions by making statements that are demonstrably false.

For example, I can understand some of the motivation of anti-vaxers in general, but I have very little sympathy for those of them who propagate misinformation that grossly exaggerates side effects of vaccines. The reason for this is very simple, once our opinions become disconnected from the objective reality, there remains nothing that can be discussed constructively, as we end up just sharing the differences in our systems of beliefs.

Very similarly, I have used the word "idiot" on the page 3 of this thread in a very specific context and applied it to a particular person from twitter, who repeatedly made statements that are manifestly incorrect and absolutely misleading, and whose misinformation campaign pretty much created the first wave of hype against generative AI on twitter. My "sorry not sorry" attitude, although you presented it as a sign of arrogance, was actually to do with the fact that I do believe that person to be grossly unclever, from my observations of his interactions on twitter. Hence, my "sorry" did not refer to him, it referred to everyone else in this thread and showed that I do, in fact, recognize a certain level of escalation that results from using such language.

I also explained what was misleading about this person statements earlier in this very thread. However, people keep repeating very similar claims (e.g. the claim that AI produce collages of works used for training). This does make me want to make snarky remarks, true. I am sorry about that. Will try my best to reduce the number of such remarks even further.

I have used the word ANANASAKÄÄMÄ on page 13 of this thread.

Like it or not that language is now in the class action lawsuit.

Isn't basic AI kinda used to generate stuff even in software like Photoshop? Imho the line between what is AI or not AI and how much of it, is (or should be) acceptable is a thin one...

Yes, I've seen it, and not entirely surprised given the state of this discussion even on a technical site like ours.

Replace AI with the machine learning and the line gets much clearer

So it all comes down to not if AI was used or not, but rather how much of it was used to automate the creation process, no...?

introspec: I feel that I have a pretty good understanding of how an AI image generator and AI in general works - but I'd be interested to read your definitive explanation and any concrete points you feel like you want to make - I'm sure I/we could learn something.

"How much"? The line between machine learning and a lack of it is rather qualitative than quantitative.

But a well trained AI can produce a rather high quality image these days.. I think the real question here is, 'is it ethical to use such tech in a demo compo or not?' Anyway, this discussion reminds me of 'should 64ks use directx', which was later evolved to 'should demos be made with using demotools', which was succeeded with 'should commercial game engines be used for making demos'.

Just a thought... so like modern sound engineering (autotune and alike) destroyed the singing profession, AI has a big chance to destroy gfx artist profession. It's not gonna be fully destroyed. It will probably just cut the "long tail" of mediocre artists that cannot deliver better quality than AI-based art. Lawsuits or not, the tech, VCs etc.. are unstoppable here.

But contrary to the hype, I really doubt we will have AI that will be superior to top human performers any time soon in any profession, including art, coding etc... I can give you a simple "proof": all current AI solutions are fed on the data derived from real human work and the trained model cannot really exceed median-case easily. And, in the training data, there always gonna be much more mediocre input than outstanding (sort of by definition of outstanding).

There are of course some attempts to improve the situation (think unsupervised learning), but IMHO they won't really take off any time soon (think exponential vs linear compute-power increase is needed) and are not even proven so far to work (despite some audacious claims).

Unfortunately, if anything, market shows that people are ok with mediocre quality as long as it's cheaper and there is more of it, so... civilization is doomed and will start eating its own tail very soon. AI generated content will be used for training new AI generated content while nothing fresh will be entering the closed loop that is being created.