Another demonstration of chatGPT 4.0 capabilities

It's a bit difficult because a real artist can be influenced to the nth degree by another artist and yet still produce new art.  Is this not the same, just automated?

That's the argument used, but to me the reliance on these images is still there in the underlying maths.

For example I played around with a few of these [art generators] and it wasn't that hard to get them to produce somewhat altered collages of copyrighted work. I'm sure the AI art gen companies are getting better at hiding what art they took but we'll have to see what the IP lawyers are able to "trick" the AIs into making.

AI Spits Out Exact Copies of Training Images, Real People, Logos, Researchers Find

There are apparently people who guard their prompts because they don't want the AIs to make the image for someone else, and some even think these prompts are a sign of their own artistic merit...to me that's like a supposed lothario bragging about how much porn he watches...

On Deep Dream there is a button called "try it" which is designed so that you can use other people's prompts.  A lot of my own stuff uses other people's combined with my own images.  It's all a lot of fun but I am mostly with you; the mass production element is soulless.  I get far more satisfaction out of the process of drawing and painting than I get out of AI text prompts.  I do sometimes find it useful for inspiration as to how to finish my incomplete artwork.  I just load up the unfinished piece with some text saying what kind of thing I am looking for and see what it comes up with.

That's the argument used, but to me the reliance on these images is still there in the underlying maths.

For example I played around with a few of these [art generators] and it wasn't that hard to get them to produce somewhat altered collages of copyrighted work. I'm sure the AI art gen companies are getting better at hiding what art they took but we'll have to see what the IP lawyers are able to "trick" the AIs into making.

AI Spits Out Exact Copies of Training Images, Real People, Logos, Researchers Find

I have noticed that there are some rare moments when an almost exact Edward Hopper replica or something similar crops up, so you are right - it is an issue.

I don't doubt they'll give a robot a paintbrush to paint a generated image...this might be as early as next year as I can't think of any major programmatic barriers to this happening.

Of course reading about the life and art of, from personal experience, Tiepolo and then correctly identifying one of his paintings from afar in a museum...that delight has no place in the dull world of mass produced junk.

There are apparently people who guard their prompts because they don't want the AIs to make the image for someone else, and some even think these prompts are a sign of their own artistic merit...to me that's like a supposed lothario bragging about how much porn he watches...

What is at the core of Art --->  is experience!  AI's promptings and responses are based on algorithms - not actual connectivity to cultural feelings and personal emotions.  Data about experience is not the real thing.

AI cannot but corrupt common sense and in time, the coming generations ability to see deeply into social environments.  It's all gonna float on the surface.  What would Umberto Eco say?

I don't doubt they'll give a robot a paintbrush to paint a generated image...this might be as early as next year as I can't think of any major programmatic barriers to this happening.

Of course reading about the life and art of, from personal experience, Tiepolo and then correctly identifying one of his paintings from afar in a museum...that delight has no place in the dull world of mass produced junk.

There are apparently people who guard their prompts because they don't want the AIs to make the image for someone else, and some even think these prompts are a sign of their own artistic merit...to me that's like a supposed lothario bragging about how much porn he watches...

The gap for going physical is much bigger than you give it credit for. A training set for learning anything physical (like controlling a robotic arm as an example) doesn’t exist. It’s much easier to train on digital data.

The gap for going physical is much bigger than you give it credit for. A training set for learning anything physical (like controlling a robotic arm as an example) doesn’t exist. It’s much easier to train on digital data.

I agree with your assessment of real world robotics not having training sets (or at least not high quality ones)...but I don't think you need a trained robotic arm in the sense of machine learning here? I figure you can have the image first, then have the robot paint?




=-=-=

What is at the core of Art --->  is experience!  AI's promptings and responses are based on algorithms - not actual connectivity to cultural feelings and personal emotions.  Data about experience is not the real thing.

AI cannot but corrupt common sense and in time, the coming generations ability to see deeply into social environments.  It's all gonna float on the surface.  What would Umberto Eco say?

I don't think I've read enough Eco to know what he'd say heh, and it's been some years since.

I do think artists with a deeper story for each work will survive, as even now that carries weight in the art world. The whole "my kid could paint that" debate about works by Cezanne or Kadinsky that seem like blurs and scribbles.

But what about an artist like James Jean where it's the humanity that gives it weight, whereas I would find an AI version hollow:



Jean also "came up" from comic book work, building his career over years. I don't think we'll get artist career paths like that when the "lower rung" of work is not even available.

The "good" thing is I don't think it's very difficult to enter the AI art gen space, [so] you will get a major [decline] on profitability in tandem with mega-corporations using their power to restrict what can be used in the training set. That combo might ideally keep a career path for artists more open as it becomes less & less profitable to maintain art generators.

Signing up for Midjourney this week left me flabbergasted, seeing the images produced by the Dall-e I hadn't expected too much, but I was completely floored by just how good these AI generated images on Midjourney were, and how effective the prompt interface was. I needed a painted image of a small rowing boat on the ocean in a violent storm with dark clouds and waves, and I asked for it to be in the style of Turner, and in 16:9 ratio.

In 30 seconds it produced 4 images (below)... Until now, I don't think I quite comprehended the scale of this AI revolution... and the business types it's  going to depreciate... but on the other hand, it offers amazing creative abilities to everyone to create images so quickly. Needless to say I signed up for an annual subscription. It's also bloody addictive!

I do quite like some of those images.

As for the style of Turner, I can't really say - of course I'm familiar with his work but I've spent more time looking at French and continental European artists.

Recently I came across a set of about thirty individual images generated by Midjourney, all on the same theme and in the style of various well-known artists such as Van Gogh and Picasso. Though the results were impressive, they did tend to look a bit weak to me. Some of the artists I am most familiar with tended to use bold composition and interesting appearance at the level of individual brush-strokes. The AI images though interesting somehow missed both of those aspects, missing something which was characteristic of the intended artist.

I don't know where this is going. There are surely future developments to come.

On there I found people designing products for manufacture, designing business cards and business logo's. Marketing agencies producing images for brochures, architects producing architectural illustrations of projects for presentations, people designing wallpapers and fabrics, sports teams designing new kit, major publishers creating photographs to illustrate their publications, housing associations designing web sites and brochures, authors designing book covers, songwriters and music publishers designing CD covers... totally amazing.

Since you mention architectural illustrations, it had occurred to me that architectural design might be a useful practical application for AI. Certainly seeing some of the horrible designs that were deemed acceptable in the 1960s, it could be useful to have some more interesting buildings, hopefully able to be constructed for no extra cost as compared with a dull, drab one. This may not yet be practical but to me it seems it could be beneficial for the people who would have to live with the results for decades.

On Deep Dream there is a button called "try it" which is designed so that you can use other people's prompts.  A lot of my own stuff uses other people's combined with my own images.  It's all a lot of fun but I am mostly with you; the mass production element is soulless.  I get far more satisfaction out of the process of drawing and painting than I get out of AI text prompts.  I do sometimes find it useful for inspiration as to how to finish my incomplete artwork.  I just load up the unfinished piece with some text saying what kind of thing I am looking for and see what it comes up with.

I suspect if artists used AI generation, even in their own styles, people would demand discounts for the work because the "AI did most of it".

I actually think artists using AI to make work in their own styles has some merit, and this could actually make art as a career and hobby more viable. I hope that if AI art generators are here to say this is what happens otherwise I fear we will end up with less and less novelty.