AI vs Creators

[Mr Amore]

Something I liked about SL was the landscaping, it could take weeks to complete a scene with a spare hour thrown in at the end of day. Virtual landscaping is an art form as any other, it's relaxing and it's enlightening as you begin to view the world differently. And you find an appreciation for the builds of others.

This is all set to change. The future will not be an inventory filled with foliage, rocks and a canvas to build upon. It's going to be a sequence of words, instructing an AI to construct something for you and scarily, the results will be superior to anything you could build yourself - At least that's true in my case.

These past months youtube has been spammed with AI generated art and even to an AI skeptic as myself, the results are remarkable. To see more, search for DALL-E 2 or Midjourney.

 

 

 

Within the decade this level of AI generation will extend to 3D environments too. Feed the AI your terms and within seconds it will generate a tree, a building, or an entire scene. Whether we'll find it as satisfying remains to be seen.

 

[Seicher Rae]

Maybe in SL version 2122 something approximating Midjourney will be available. It is fun to think about. Then again, it is also fun to just use the land tools, make really high hills and then flatten them, over and over. I do think it would be nice to see more fantastical sim designs, which do exist here and there, in SL.

[Nalates Urriah]

In different places we have tools to build the various components used in region building. Blender has parametric tree maker. The AI you describe only sounds like the separate tools being brought together in a single app.

SL is a parametric world. Other than mesh everything is made from a parametric primitive.

The AI you describe, as described, is a parametric tool. If you follow any of the Mandelbrot fractal art you have seen the amazing stuff parametric equations can generate. The art is detailed beyond anything humans create by hand (may be more accurately said, without computer assistance).

None of these tools have stopped people from being creative or working in their favored media as the choose. Don't worry about AI and art.

[Coffee Pancake]

The tells that something is AI generated are only going to get smaller, the bulk is already good enough to get past human inspection.

It's not "just procedural" math.

[Prokofy Neva]

16 hours ago, Mr Amore said:

Something I liked about SL was the landscaping, it could take weeks to complete a scene with a spare hour thrown in at the end of day. Virtual landscaping is an art form as any other, it's relaxing and it's enlightening as you begin to view the world differently. And you find an appreciation for the builds of others.

This is all set to change. The future will not be an inventory filled with foliage, rocks and a canvas to build upon. It's going to be a sequence of words, instructing an AI to construct something for you and scarily, the results will be superior to anything you could build yourself - At least that's true in my case.

These past months youtube has been spammed with AI generated art and even to an AI skeptic as myself, the results are remarkable. To see more, search for DALL-E 2 or Midjourney.

 

 

 

Within the decade this level of AI generation will extend to 3D environments too. Feed the AI your terms and within seconds it will generate a tree, a building, or an entire scene. Whether we'll find it as satisfying remains to be seen.

 

I've been watching MidJourney and got an invitation to be in it for awhile but you have to pay a subscription fee to stay. There's another one called Craiion which has a similar premise.

And what I find that when I plug in any number of wild search terms with increasing granularity, it looks like kasha.

But when a real artist works it, it looks great.

I'm not sure whether the AI brain is violating copyright by mashing up a whole bunch of people's IP and serving it up this way. 

But it seems that it won't "replace" artists any time soon, just like Google Translate didn't replace translators.

 

[Mr Amore]

On 8/16/2022 at 4:54 PM, Nalates Urriah said:

None of these tools have stopped people from being creative or working in their favored media as the choose. Don't worry about AI and art.

In this challenge you can see the tension in the artist, his future is literally on the line. Eventually businesses will be deciding whether to employ a graphics designer with a full salary and overheads, or an annual subscription to an AI artist.

 

On 8/17/2022 at 6:23 AM, Prokofy Neva said:

I'm not sure whether the AI brain is violating copyright by mashing up a whole bunch of people's IP and serving it up this way.

Infringement is brushed over in this video, also an interesting guide on influencing the AI.

 

Excuse the youtube spam, but a video's worth a million words and all that ;)

[Bree Giffen]

On 8/16/2022 at 10:23 PM, Prokofy Neva said:

I've been watching MidJourney and got an invitation to be in it for awhile but you have to pay a subscription fee to stay. There's another one called Craiion which has a similar premise.

And what I find that when I plug in any number of wild search terms with increasing granularity, it looks like kasha.

But when a real artist works it, it looks great.

I'm not sure whether the AI brain is violating copyright by mashing up a whole bunch of people's IP and serving it up this way. 

But it seems that it won't "replace" artists any time soon, just like Google Translate didn't replace translators.

 

That's a good point. A  real artist would know what looks good. They would reject bad art and keep pushing the AI to create new artwork until it looks good to them. A non-artist layman can't just ask the AI to create art and decide which of the generated artwork is the best to use. The AI can't select among it's artwork which is the best.

The artwork I see generated by the AI seems like the art I see in NFTs. It's like the uncanny valley of artwork.  

[Fluffy Sharkfin]

I'm a big fan of AI experiments and AI art has produced some excellent results but, as usual, it seems that those in charge of developing the technology decided to breeze right past the question of "should we" and focus solely on the "can we".  In this instance it's a case of "should we take the intellectual property of others and use it as a dataset for our experiments without their permission?".

That being said, I think these new AI driven image generation systems could be quite useful as a tool for visualization purposes, like AI generated concept art for example.

It would be interesting to see what results an AI could produce if limited to a large dataset of PBR surface materials.  Having an AI that can generate unlimited variations of seamless PBR textures based on a few keywords like "Ivy covered garden wall" or "paint splattered rough wooden planks" would certainly cut down on production time considerably when creating 3D assets.

Anyway I doubt AI art is going to be taking many jobs in the design field (at least not from anyone that's actually competent at their job), as someone on Twitter put it ...

 

[Quistess Alpha]

On 8/16/2022 at 1:23 PM, Coffee Pancake said:

The tells that something is AI generated are only going to get smaller, the bulk is already good enough to get past human inspection.

It might be good enough to get past a cursory glance, or to someone who's not spent an hour or two looking at AI images, but I think 9 times / 10 they still produce tells that you can't un-see once you notice them.

(sorry for the week later quote)

Like pretty much any technology, it isn't going to "replace" people, so much as it's going to become a tool in the toolbelt of people who already do related work. (Although you can make the argument that increased efficiency without increased demand leads to consolidation: people who don't utilize the technology or don't use it as effectively as others will eventually be out-competed by those who do)

TL;DR "AI-assisted" images seem more likely to become mainstream in the near future than completely AI-generated ones.

[Coffee Pancake]

27 minutes ago, Fluffy Sharkfin said:

Anyway I doubt AI art is going to be taking many jobs in the design field (at least not from anyone that's actually competent at their job), as someone on Twitter put it ...

 

To get ONE good design from a human the client needs to accurately describe what they want. Broad range of skills is required by both the artist and the client.

Picking one good design from INFINATE variations. Simple skill on the part of the client, there is no artist.

Picking the one perfect design that maximally appeals to inspire a precise response in a chosen demographic of monkey brains takes no one at all.

3 minutes ago, Quistess Alpha said:

Like pretty much any technology, it isn't going to "replace" people, so much as it's going to become a tool in the toolbelt of people who already do related work. (Although you can make the argument that increased efficiency without increased demand leads to consolidation: people who don't utilize the technology or don't use it as effectively as others will eventually be out-competed by those who do)

TL;DR "AI-assisted" images seem more likely to become mainstream in the near future than completely AI-generated ones.

Right now there is some skill is poking the AI, because the tech is output focused. This will change.

We still need a human touch up artists to fix flaws and correct for "dreaming". Betting we could get an AI to do that too .. and another one to do context aware composition .. 

 

This represents the dawn of a new art period and bombs human art back to a niche populated by few with exceptional skill and personality. Future controversies wont be about the content of the work, but whether the worker had help .. or whether the worker even had agency.

[Fluffy Sharkfin]

50 minutes ago, Coffee Pancake said:

To get ONE good design from a human the client needs to accurately describe what they want. Broad range of skills is required by both the artist and the client.

Picking one good design from INFINATE variations. Simple skill on the part of the client, there is no artist.

Picking the one perfect design that maximally appeals to inspire a precise response in a chosen demographic of monkey brains takes no one at all.

I'm not suggesting AI won't have a transformative effect on the way in which people approach the design process but as with any new technology initial fears that it will replace humans entirely in a particular field are, I feel, a little alarmist.  It seems more likely that the end result will be more effective tools which can be used to streamline the process of creativity.

 

I think if any group of people should be in fear of AI replacing human artists it should be art critics who rather than pondering the deeper meaning of what an artist is really trying to express with a particular piece will be relegated to standing around trying to guess which keywords they used to have the AI generate said "masterpiece".

[Coffee Pancake]

50 minutes ago, Fluffy Sharkfin said:

I'm not suggesting AI won't have a transformative effect on the way in which people approach the design process but as with any new technology initial fears that it will replace humans entirely in a particular field are, I feel, a little alarmist.  It seems more likely that the end result will be more effective tools which can be used to streamline the process of creativity.

The fear isn't that this bumps one specific profession, it's that this is the tip of a very large iceberg and we're seeing similar inroads into other areas once considered protected human spaces. In the short term we can expect dramatic increases in capability, the size and scope of projects can grow well beyond what current teams can accomplish in commercial timeframes, in the medium term we're relegated to an oversight or editorial role and in the end we simply can't compete.

The only realm with some protected superiority is the physical one, and only because simulation can't match real world complexity and variability. It's not practical to build a complex enough simulation that can trial and error train a model in a reasonable timeframe. Yet.

 

Current AI is limited by hardware, that will change. Asianometry have done some very good videos on the subject from an industry perspective.

 

The Coming AI Chip Boom - https://www.youtube.com/watch?v=L0948yq2Hqk

Running Neural Networks on Meshes of Light - https://www.youtube.com/watch?v=t0yj4hBDUsc

 

[Mr Amore]

7 hours ago, Fluffy Sharkfin said:

It would be interesting to see what results an AI could produce if limited to a large dataset of PBR surface materials.  Having an AI that can generate unlimited variations of seamless PBR textures based on a few keywords like "Ivy covered garden wall" or "paint splattered rough wooden planks" would certainly cut down on production time considerably when creating 3D assets.

I'm certain it's branching into this direction, eventually AI will be creating the textures, the models, and the entire scene too based upon a few parameters.

In Star Citizen, each ship takes months, or even years to reach their gold standard from concept to completion. Some of these are showcased at every stage along the way.

https://starcitizen.tools/images/4/4d/Ship_Pipeline_-_Full.png

Accounting for the 'human hours' and overheads, each vehicle easily exceeds a million dollars and several months to create. When AI is sufficiently trained, it could breeze through the entire process in a matter of seconds. There may be some lingering discrepancies, balanced against the sheer savings in time and investment.

It's not so much a case of artists losing their jobs, but the smaller studios with limited budgets will start employing AI over artists. With procedural generation being an existing case with limited impact on the industry.

[Fluffy Sharkfin]

16 hours ago, Coffee Pancake said:

The fear isn't that this bumps one specific profession, it's that this is the tip of a very large iceberg and we're seeing similar inroads into other areas once considered protected human spaces. In the short term we can expect dramatic increases in capability, the size and scope of projects can grow well beyond what current teams can accomplish in commercial timeframes, in the medium term we're relegated to an oversight or editorial role and in the end we simply can't compete.

Without wandering too far off-topic (since the thread is called AI vs Creators, not AI vs Humanity) if your predictions are correct and the evolution of AI results in a large proportion of humanity becoming essentially redundant then perhaps that's just part of our own evolution.  Using productivity as a metric by which to judge the value of an individual whilst simultaneously striving to automate every possible task is obviously a recipe for disaster,  if AI expedites the process of people coming to that realization and becomes a catalyst for societal changes then perhaps that's not such a bad thing.

12 hours ago, Mr Amore said:

I'm certain it's branching into this direction, eventually AI will be creating the textures, the models, and the entire scene too based upon a few parameters.

In Star Citizen, each ship takes months, or even years to reach their gold standard from concept to completion. Some of these are showcased at every stage along the way.

https://starcitizen.tools/images/4/4d/Ship_Pipeline_-_Full.png

Accounting for the 'human hours' and overheads, each vehicle easily exceeds a million dollars and several months to create. When AI is sufficiently trained, it could breeze through the entire process in a matter of seconds. There may be some lingering discrepancies, balanced against the sheer savings in time and investment.

It's not so much a case of artists losing their jobs, but the smaller studios with limited budgets will start employing AI over artists. With procedural generation being an existing case with limited impact on the industry.

While I'm sure there are plenty of studios that will happily employ AI to generate a lot of generic, derivative assets, one of the key components of asset creation is storytelling.  A good 3D asset (be it a character, environment or prop) will often have a myriad of small details all of which serve as visual cues which provide the context and history of the character, place or item.  So essentially, even if you were to use an AI to create such an asset the core skills of imagination, observation and descriptiveness which are essential as a creator are still required in order to provide a sufficiently detailed prompt that would allow the AI to create a result that even came close to what the creator had in mind.

When it comes to creativity AI, much like procedural generation, will most likely end up as a toy for those who aren't inclined to spend countless hours learning how to use 2D/3D apps and as a very powerful tool in the hands of those with the expertise to use it properly.  No matter how "good" AI art may get in the future, it's not going to effect real creators because at its core art is about expression rather than commerce and I doubt AI will rob humanity of its desire to express itself.

[Bree Giffen]

I've seen technology start to replace or enhance everything that nearly everyone does. As humans we have strength, intelligence and creativity. The US used automation to remove the need to have large numbers of people using their strength in reducing farmers and factory workers.  AI will remove the  need for intelligence and creativity. Essentially, society will be filled with weak, dumb and uncreative people.  

[Quistess Alpha]

seems topical: