MicrostockGroup Sponsors


Author Topic: Dall e 2 will make us all redundant?  (Read 18284 times)

0 Members and 1 Guest are viewing this topic.

Justanotherphotographer

« Reply #100 on: October 16, 2022, 04:04 »
+2

DALL E doesn't "pull" images from anywhere. It has been trained with existing images. Even if no one can afford to make a living from photography anymore, DALL E will not "unlearn" what it has learned.

Isn't it training on an ongoing basis? My point is that in 10 years, if no one (or much fewer people) are producing images the AI will have trained (if you don't like the word pulled  ::) ) using images a decade old.

Think about how most images of business people get online. It's stock images. Think about stock images from the 90s now. Will AI images of business people be frozen looking like people from the 2020s because that's when most of the original "training" images are from. So the AI thinks that's what business people should look like in 2050.





Justanotherphotographer

« Reply #101 on: October 16, 2022, 04:07 »
0
DALL-E has an annoying tendency to crop - and doesnt understand 'no cropping' etc in the input phrase
Yes, I noticed that too and sometimes the crops are really ridiculous, like cutting off the whole head of a person. I tried all kinds of phrases to avoid this, like "no crop, not cropped, person fully visible in picture, head not cut off". I have not been able to figure out any kind of instruction that DALL-E uderstands.
But, to be honest, even though I was searching for images for work and could have used some good results, I was still glad to see such flaws. The more flaws I find with DALL-E's performance the less I am worried that I will be completely replaceble as a photographer, at least in the near future.

yes, i've tried many such phrases and also emailed DALL-E w no reply.

and yes, i see little competition in the near future

You can edit and draw a frame for it to fill in if it's cropped too tight

« Reply #102 on: October 18, 2022, 05:53 »
+4

Uncle Pete

  • Great Place by a Great Lake - My Home Port
« Reply #103 on: October 18, 2022, 14:09 »
0
Getty bans AI images https://www.theverge.com/2022/9/21/23364696/getty-images-ai-ban-generated-artwork-illustration-copyright

This is trending and the lawyers get into the game, which means, the agencies can make their own rules, until there's a legal decision.

« Reply #104 on: October 18, 2022, 23:29 »
+1
The article Getty ban AI images is very interesting.

It points out some of concerns of this community but artificial intelligence has a lot of potential and a world of possibilities.
There is a huge amount of work that I see on twitter every day and I can't tell the difference if it was created with only Photoshop, with tools generated by AI and Photoshop, or only generated by AI through the prompt. Unless the creator tells me how he made it, I don't see how it will be fairly possible to detect it. So if Getty Images will rely on users to identify and report such images i would say its a lost cause. I rather rely on AI to do some reverse engineer and try to come up with a diagnosis to the curator.

But even with Sci-fi AI reverse engineer idea the plot becomes more complicated...
For example if we use dreambooth with stable diffusion means you can create your own models and then use prompts in stable diffusion to act just like a filter, paint or style you think of...

I don't have a solution but maybe problem does not rely in "AI-generated art" itself but more in the process used to generate art with AI.
There should be legal boundaries.


tupungato

  • Europe
« Reply #105 on: October 19, 2022, 08:54 »
+1
Dear Contributors,

Greetings from 123RF!

Recently, we notice an increase in the submission of AI generated content.

Due to the legal uncertainties with regards to AI generated content, were getting in touch to inform you that 123RF will cease to accept all submissions created using AI generation platforms. This ruling will be applied retrospectively to content uploaded prior to this announcement.

What we will not accept into any paid collection:

Any content generated from AI generation platforms (examples include but not limited to: Dall.E 2, Midjourney, Stable Diffusion, etc.)

« Reply #106 on: October 19, 2022, 09:13 »
0
Isn't it training on an ongoing basis? My point is that in 10 years, if no one (or much fewer people) are producing images the AI will have trained (if you don't like the word pulled  ::) ) using images a decade old.

Sorry, but there is still something wrong; probably you have not a really clear view of how AI works.
It learn continuously, every time you use it and every minutes from our own inputs.
The lessons analyzing zillion of images were just a starting point; when it knows enough to reproduce a new requested object, it will start to learn more from our inputs and results. There is no need to feed AI with another zillion of images every couple of year. It will know, based on inputs, the new styles, also probably it will create new ones

« Reply #107 on: October 19, 2022, 09:28 »
+1
You're fighting a losing battle with this one I'm afraid. If you were to take parts (learnt) of X amount of songs to combine and form a new one, you'd have to pay the copyright holder of the original content

You're wrong
First of all in music there is a specific legal limit in terms of sounds and sequences in which you can claim a copyright infringment. That is quite obvious, because with only seven notes it's hard to create music without copying existing sequences.

Second, you are still thinking and telling that some parts of a copyrighted image is used inside the new image. That is simply not true.
AI create a completely new image, having an idea of how a wheel, a face, or a burger, is done.

Come on, it's called AI because it can do exactly the same operation that everyone of us do creating an image, having in mind the previous created images (created by others) with the same subject.


THERE IS AN ABSOLUTELY SERIOUS LEGAL ISSUE BECAUSE THE FEED THAT AI HAS USED TO START IS PROBABLY STOLEN, OR WITHOUT APPROPRIATE RELEASE
This is the problem, and this is very serious.

But you think that a legal problem will be the use of copyrighted creation INSIDE a new AI generated image, you're wrong.
Any AI engineer will easily demonstrate that no part of copyrighted images has been used
« Last Edit: October 19, 2022, 09:40 by derby »

« Reply #108 on: October 19, 2022, 10:19 »
0
You're fighting a losing battle with this one I'm afraid. If you were to take parts (learnt) of X amount of songs to combine and form a new one, you'd have to pay the copyright holder of the original content

You're wrong
First of all in music there is a specific legal limit in terms of sounds and sequences in which you can claim a copyright infringment. That is quite obvious, because with only seven notes it's hard to create music without copying existing sequences.

Second, you are still thinking and telling that some parts of a copyrighted image is used inside the new image. That is simply not true.
AI create a completely new image, having an idea of how a wheel, a face, or a burger, is done.

Come on, it's called AI because it can do exactly the same operation that everyone of us do creating an image, having in mind the previous created images (created by others) with the same subject.


THERE IS AN ABSOLUTELY SERIOUS LEGAL ISSUE BECAUSE THE FEED THAT AI HAS USED TO START IS PROBABLY STOLEN, OR WITHOUT APPROPRIATE RELEASE
This is the problem, and this is very serious.

But you think that a legal problem will be the use of copyrighted creation INSIDE a new AI generated image, you're wrong.
Any AI engineer will easily demonstrate that no part of copyrighted images has been used

In your opinion I'm wrong. While you're entitled to that opinion it doesn't make it fact. Neither you or I can provide the hard evidence at this point and it will end up in a court of law to determine it. As it stands, the list of agencies who have concerns over the legality of this is growing by the day and they do have a lot of legal minds who have experience in this field.

"Come on, it's called AI because it can do exactly the same operation that everyone of us do creating an image, having in mind the previous created images (created by others) with the same subject."

You're more gullible / naive than I thought if you believe that.
« Last Edit: October 19, 2022, 10:22 by HalfFull »

Uncle Pete

  • Great Place by a Great Lake - My Home Port
« Reply #109 on: October 19, 2022, 10:37 »
+1
You can edit and draw a frame for it to fill in if it's cropped too tight

 :o The image the AI creates runs over the edges. How do I create a frame for something that doesn't exist?



DALL-E sure creates some strange and bizarre images, when it comes to things mechanical. The AI doesn't understand construction and function.
 

« Reply #110 on: October 19, 2022, 11:29 »
0
As it stands, the list of agencies who have concerns over the legality of this is growing by the day and they do have a lot of legal minds who have experience in this field.

The only problem for the agencies is that they can't legally refer an image to an author: this follow the early decisions in USA about copyright on images created by AI. I suspect that this is the only reason to not accept these images.

Of course most of the thing I wrote are only my opinion, like for every one else :-)
But, once again, I suspect that the way in which AI works is much more a fact than an opinion...

« Reply #111 on: October 19, 2022, 12:23 »
0
As it stands, the list of agencies who have concerns over the legality of this is growing by the day and they do have a lot of legal minds who have experience in this field.

The only problem for the agencies is that they can't legally refer an image to an author: this follow the early decisions in USA about copyright on images created by AI. I suspect that this is the only reason to not accept these images.

Of course most of the thing I wrote are only my opinion, like for every one else :-)
But, once again, I suspect that the way in which AI works is much more a fact than an opinion...

As it stands, we don't know how things will pan out but we do see a lot of examples of AI imagery that contains watermarks. If it turns out, which seems quite obvious now, that it has, "Learnt" from our images without our permission etc, then it maybe forced to remove the data and start a fresh with data from imagery the have acquired legally (imagery without watermarks). Who knows.

As I mentioned in another thread. It should be interesting to see how AI imagery is used. I see real potential with uses that need images for fiction, places that don't exist etc.

For places that do exist I see plenty of imagery already available that is 100% accurate. AI imagery for tourism, guides, reference etc is unlikely to take off because unless you can 100% guarantee it's exactly right, then there are likely to be problems for those use types. You can imagine confused tourists looking up at XYZ and saying, it looks nothing like that in the book!

For fine art, I see real possibilities there as well. The creative side of things. Not so much in terms of imagery of real places to hang on the wall. People buy those types of imagery as they want to see the real thing. If they wanted a representation of it, they'd by an illustration.

At the moment AI Imagery is all new and people are rushing in and trying to make money from it but in the end, I think it'll end up being limited to specifics uses.

For illustrations, I know that I can create my own work that is better and it benefits from my understanding of how mechanically things work in the world whereas AI doesn't. Humans are a living organism that are conscious and always question what they see in order to make sense of the world we live in. AI on the other hand repeats what its been shown. It comes up with imagery based on what we've told it and It never questions it. It requires humans to provide it with new reference material, it doesn't think, "Oh, that looks dated I must shoot something new".

If we stopped providing reference material to it today, it will carry on producing images based on images of today. So, if all photographers are put out of work, what will it produce in 50 years time when there is no new reference material available to feed into the machine. If building A was knocked down in city B and it didn't have enough reference shots of the new view it wouldn't question it, it would carry on being wrong.

Mind you, the day AI becomes conscious entity and is able to question things, it will probably look for ways to get rid of us... to protect itself and stop us from pulling the plug  :D
« Last Edit: October 19, 2022, 12:25 by HalfFull »

« Reply #112 on: October 19, 2022, 13:12 »
+1
For places that do exist I see plenty of imagery already available that is 100% accurate. AI imagery for tourism, guides, reference etc is unlikely to take off because unless you can 100% guarantee it's exactly right, then there are likely to be problems for those use types. You can imagine confused tourists looking up at XYZ and saying, it looks nothing like that in the book!

in 2017 National Geographic ran a controversial cover. " The altered image was displayed to the unwitting reader without mention, leading readers to assume the pyramids were significantly closer to each other than they actually were."

https://medium.com/engl462/visual-deceptions-national-geographic-and-the-pyramids-of-giza-3fee6d448d0d

« Reply #113 on: October 19, 2022, 13:41 »
0
You can edit and draw a frame for it to fill in if it's cropped too tight

 :o The image the AI creates runs over the edges. How do I create a frame for something that doesn't exist?
....
DALL-E sure creates some strange and bizarre images, when it comes to things mechanical. The AI doesn't understand construction and function.

when frame enlarged, AI can produce a reasonably accurate extension:



Image on right my actual image, but DALL-E phrase did not use it

« Reply #114 on: October 19, 2022, 14:10 »
0
That type of reaction is what I'd expect when something is materially altered and basically passed off as real. People feel cheated when they visit the place and find it's not as expected. They certainly don't expect the National Geographic to alter reality for their convenience  ::)

Edit.
That is quite different from reality though... maybe a long distant relative of the original  ;)

« Reply #115 on: October 19, 2022, 15:40 »
0
Thanks for the news 123rf.

The way I see it, I think the problem isn't just or exactly contributors submitting AI images to microstock sites.

The volume possible to produce in a short time is large and a AI contributors/companies may choose to make a website and sell there or sell directly to other companies.In fact, there are already sites for selling AI images like https://generated.photos/ or https://aiartshop.com. Even Saatchiart, opensea or rarible are in this race too.

But the problem is that 123rf shares 2 million free photos... and other sites too just to attract clients. If someone use these free contributor photos (our work) to train models.... how long 123rf thinks its needed to train a model and create a new image? Of course it would depend on the size of the model you want to train but today you can train in less than an hour with amazing results with at least 12 photos. Not to mention that you can do it from a cheap laptop with google colab.

So there is the problem of free photos being available for machine learning too...
« Last Edit: October 19, 2022, 18:46 by Evaristo tenscadisto »

Brasilnut

  • Author Brutally Honest Guide to Microstock & Blog

« Reply #116 on: October 24, 2022, 07:20 »
+3
Last weekend I invested on an old green suitcase at a flea market and using Dalle2.

Finding it very useful to brainstorm various concepts for book covers.

From brainstorming -> final execution -> acceptance at Arcangel. Useful tool!

tupungato

  • Europe
« Reply #117 on: October 25, 2022, 07:28 »
0
Shutterstock announcement:
We will not accept content generated by AI to be directly uploaded and sold by contributors in our marketplace because its authorship cannot be attributed to an individual person consistent with the original copyright ownership required to license rights. Please see our latest guidelines here. When the work of many contributed to the creation of a single piece of AI-generated content, we want to ensure that the many are protected and compensated, not just the individual that generated the content.

PaulieWalnuts

  • We Have Exciting News For You
« Reply #118 on: October 25, 2022, 07:57 »
+6
Quote
Shutterstock: Working together to lead the way with AI

Were excited to announce that we are partnering with OpenAI to bring the tools and experiences to the Shutterstock marketplace that will enable our customers to instantly generate and download images based on the keywords they enter.

As we step into this emerging space, we are going to do it in the best way we know howwith an approach that both compensates our contributor community and protects our customers.

In this spirit, we will not accept content generated by AI to be directly uploaded and sold by contributors in our marketplace because its authorship cannot be attributed to an individual person consistent with the original copyright ownership required to license rights. Please see our latest guidelines here. When the work of many contributed to the creation of a single piece of AI-generated content, we want to ensure that the many are protected and compensated, not just the individual that generated the content.

In the spirit of compensating our contributor community, we are excited to announce an additional form of earnings for our contributors. Given the collective nature of generative content, we developed a revenue share compensation model where contributors whose content was involved in training the model will receive a share of the earnings from datasets and downloads of ALL AI-generated content produced on our platform.

We see generative as an exciting new opportunityan opportunity that were committed to sharing with our contributor community. For more information, please see our FAQ on the subject, which will be updated regularly.


I predicted this was coming earlier in this thread (bold part).

Any compensation to us at this point is likely temporary until their legal team figures out a loophole or the copyright aspects of AI are more well defined. Or until AI generates a significant percentage of their income then they will not care at all about contributors. As soon as they figure out legally how to quit paying us they will announce some "good news".

And this shouldn't be a surprise to anybody, it's okay for them to create and sell AI based content, but contributors aren't allowed to submit or sell any. It's all about control and money. Unless there ends up being some copyright ruling that restricts or prevents them from selling AI content, they will eventually have total control.

So yes, AI is on track for making us redundant unless legal/copyright prevents it from happening.
« Last Edit: October 25, 2022, 09:27 by PaulieWalnuts »

« Reply #119 on: October 25, 2022, 08:55 »
+1
I predicted too this move. I think it's necessary for agencies, unavoidable, and others will follow soon.

There will be hard legal wars, but this is the future and nothing will change it.

Now it will be very interesting to see if buyers have will and ability to use the AI feature.
I suspect that, after a first enthusiastic view, most of buyers will find how difficult is to translate in word what they have in mind (assuming that they have in mind a really specific subject)

We'll see. By the way, this step is inevitable

« Reply #120 on: October 25, 2022, 12:21 »
0
I predicted too this move. I think it's necessary for agencies, unavoidable, and others will follow soon.

There will be hard legal wars, but this is the future and nothing will change it.

Now it will be very interesting to see if buyers have will and ability to use the AI feature.
I suspect that, after a first enthusiastic view, most of buyers will find how difficult is to translate in word what they have in mind (assuming that they have in mind a really specific subject)

We'll see. By the way, this step is inevitable

otoh, Microsoft is moving to allow anyone to use DALL-E

« Reply #121 on: October 25, 2022, 21:56 »
+4
Somehow I feel AI could be a final nail in the microstock coffin.

« Reply #122 on: October 25, 2022, 23:44 »
+2
Somehow I feel AI could be a final nail in the microstock coffin.

I predicted the same few months back.
https://www.microstockgroup.com/general-stock-discussion/artificial-intelligence-killing-the-whole-industry/

There were people defending the same and I always said that AI training has no expiry and there is no roll-back.
This industry do not have a good future. Mark my words.

« Reply #123 on: October 26, 2022, 08:36 »
+1
Interview with OpenAI's CTO Mira Murati: https://youtu.be/Ba_C-C6UwlI

Just_to_inform_people2

« Reply #124 on: October 26, 2022, 11:47 »
0
otoh, Microsoft is moving to allow anyone to use DALL-E

Told you so:
If AI can make any picture you want why would anyone need agencies? You just buy the software and add any picture you want to your article. So it's not just the contributors losing here. It's also the agencies. They will be redundant, like us.

But no that would not be possible right?

If AI can make any picture you want why would anyone need agencies? You just buy the software and add any picture you want to your article. So it's not just the contributors losing here. It's also the agencies. They will be redundant, like us.
because it's just not that simple -  RYFM - this has already been discussed


 

Related Topics

  Subject / Started by Replies Last post
11 Replies
4902 Views
Last post August 08, 2013, 02:12
by anjaliaroha
3 Replies
2667 Views
Last post May 28, 2014, 16:50
by gnirtS
19 Replies
5343 Views
Last post May 18, 2022, 05:21
by thx9000
14 Replies
3246 Views
Last post October 20, 2022, 04:46
by MadMax
13 Replies
3417 Views
Last post April 06, 2023, 12:26
by Uncle Pete

Sponsors

Mega Bundle of 5,900+ Professional Lightroom Presets

Microstock Poll Results

Sponsors