|
MicrostockGroup Sponsors
This section allows you to view all posts made by this member. Note that you can only see posts made in areas you currently have access to.
Messages - cascoly
Pages: 1 ... 12 13 14 15 16 [17] 18 19 20 21 22 ... 170
401
« on: October 07, 2023, 11:56 »
So far it's difficult, because artists have a hard time proving that their specific image was used to train Midjourney's AI.
But there is talk about an EU law that will require developers of AI to make all material that was used to train an AI public. It's unclear if this will pass, so far it's just a draft. But if it happens, artists will have a better chance with lawsuits, because then at least they will be able to prove that their images/text/music/voice/etc. was even used.
not practical for anyone w 100+ imag es on multiple sites, having to do individual searches & recording - all for .00001c per image
Not sure what you are talking about. I am talking about potential lawsuits of copyright violations. Winning a lawsuit against a big company for copyright violation should give you way more than .00001c.
>>> EU law that will require developers of AI to make all material that was used to train an AI public.so a company makes hundreds of millions of images available - how do you find yours w/o searching for each of your images? for 100 image portffolio that's reasonable - for 10,000+? unlikely additionally who is responsible when an AI developer buys their trained dataset from someone else (ie, never actually touched any image)
402
« on: October 06, 2023, 14:04 »
Half of the images displayed on the first page for almost any search term are images "generated with AI". 5% of all images in the database are displayed as "generated with AI". ...
quick test - none of these are even 20% AI: india varanasi rainer mountaineering sherpa - a few of my Ai & others on 1st page, but still mostly non-AI which search terms yield half AI? if only 5% are in the AS collection, but half are displayed on 'almost any search term' that implies a seriously biased algorithm
403
« on: October 04, 2023, 13:21 »
as i mentioned elsewhere we can't really tell if rejection is for ai-gen as there are 3 reasons for rejection & ai-gen is one of them. rejections have always given little info about the reasons other than 'lack of quality. this is ridiculous
404
« on: October 04, 2023, 13:15 »
most of mine have been sitting for 2+ wks, many for a month & a few for 3 months. seems the only recent reviews are images that are rejected
more annoying is we have no way to detect false positive rejections as ai-gen since te are 2 other nebulous rejection reasons. so, as on SS, there's possibility for suspension that's the fault of reviewers.
405
« on: October 04, 2023, 13:06 »
So far it's difficult, because artists have a hard time proving that their specific image was used to train Midjourney's AI.
But there is talk about an EU law that will require developers of AI to make all material that was used to train an AI public. It's unclear if this will pass, so far it's just a draft. But if it happens, artists will have a better chance with lawsuits, because then at least they will be able to prove that their images/text/music/voice/etc. was even used.
not practical for anyone w 100+ imag es on multiple sites, having to do individual searches & recording - all for .00001c per image
406
« on: September 30, 2023, 12:52 »
likely they're training for a 3-legged race parcours style
407
« on: September 30, 2023, 12:43 »
how are these to be used for AI training if they have no metadata? are all images to be labeled with generic data? eg, food, rather than steak, hamburger, chicken etc?
408
« on: September 28, 2023, 12:59 »
...
Even if I should had a lot of images already done for their missions the time I have to spent to search images in my HD, re-export, collect and organizie and upload, is too much. ...
doesn't apply to this instance where no captioning is required, so easier to sh oot than gather individual older images --. i keep a spreadsheet with date, captions & meta data. So i can search my entire inventory with 1 search across the entire worksheet, getting a list of the images i need in a few seconds
409
« on: September 26, 2023, 13:35 »
Ok. Thank you.
Pardon my ignorance but is that just 16:9 upscaled or what size could that image be before it is upscaled?
I think it could have been any size. Most resizing tools will let you set the output dimension, so I could, for example take an image that has 1500x1000px and set the output dimension to 6592 x 2880px. But you would of course get an image where either part of it has been cropped off or that is distorted, so that doesn't make much sense. Probably was something like 1648x720px before resizing.
i scale up using 4x in Topaz giga AI, then reduce to what i need in LRC using just xxMP rather than an exact calculation
410
« on: September 25, 2023, 15:07 »
I think,but it is just my guess,that if you have opted out,your images suitable for the data catalog will be added,but they will not be used because you have chosen opt-out.
I say this because I see that someone who has chosen opt-out still sees their images added to the data catalog.
I'm in,I don't see why I shouldn't get some extra money for otherwise rejected images.
are these images submitted after you opted-out? you'll still get the message "ELIGIBLE for data licensing" after you opt out
411
« on: September 24, 2023, 14:31 »
....
The judges should try themselves to create an image like that.
They would quickly understand how much work and how much human thinking goes into this.
of course it's futile to hope for rational thinking by many US congressfolk & judges. their level of ignorance concerning science and technology is staggering consider evolution-, vaccine- & climate-deniers admittedly it was a long time ago - there's former Senator Ted Stevens, chair of the Senate committee responsible for regulating the internet: the Internet is not something that you just dump something on. It's not a big truck. It's a series of tubes. And if you don't understand, those tubes can be filled and if they are filled, when you put your message in, it gets in line and it's going to be delayed by anyone that puts into that tube enormous amounts of material, enormous amounts of material.
412
« on: September 22, 2023, 13:47 »
same here, sales normal
413
« on: September 19, 2023, 14:36 »
you should learn how ML works before making such inaccurate statements - #1 is just wrong - ai generators do NOT use 'parts' of your image - that's done once during training only
#2 you cant find it because they are NOT 'perpetually' using your image
I disagree. Your statements are just as opaque as the original posts about 'training' and a one time payment. At a basic level AI generated images use previous images to make new ones. So I was asking if 'training' is code for taking parts of our images to be used for making new ones. The original post said this is a one time payment. There has been no clarification on any of these terms that I've seen by Mat from Adobe Stock. So far all he has said in response to my questions is that if I remove my images from AS they won't be used for training. If I leave them there when does the training cease?
I also thinking you're an immoral person. Supporting this immoral company shows no care or concern for peoples hard work. They didn't consult us. They just think they can do it under the current agreements. They haven't explained carefully the terms they are using and what they are actually doing. You're dumping on me for asking for clarification.
a one time payment is all that is possible, since if you study ML, you'd realize there's no way to track back from the newly created dataset to the original images. you mistake my post - i criticized your ideas - not you personally. this is not twitter -- attacking a forum member as 'immoral' and continuing such will get you reported & banned. make your arguments, attack the ideas, but leave out the personal attacks.
414
« on: September 19, 2023, 13:59 »
I had some editorials go through pretty quickly....maybe they are finally getting a handle on this?
I am afraid not. Editorial photos were reviewed fast all the time, nothing has changed about that.
Waiting 28 days for real non-editorial photos now.... 
more delay than that for my AI batch; oddly, the other images i submitted at same time were reviewed a week or so ago
415
« on: September 18, 2023, 14:29 »
1. What exactly does "training purposes" mean? Is it code for perpetual use of the images for a one time payment? Using parts of the image in new AI images isn't training. That's baking it in as part of the AI recipe.
2. I've been through my emails from Adobe Stock. I found one email at Dec 2022 saying you're now accepting AI generated images. I can't find one that notifies us that you're using our work perpetually for AI. When did you send us a clear notification that this was to happen?...
you should learn how ML works before making such inaccurate statements - #1 is just wrong - ai generators do NOT use 'parts' of your image - that's done once during training only #2 you cant find it because they are NOT 'perpetually' using your image
416
« on: September 18, 2023, 13:59 »
my canva income was down 10-15% from the last 2 years thru June. it's now down 20-30%. still, canva's almost always my 2nd largest earner (after SS)
re the pool - many are ignoring the fact that the total # of images is not static - so, if you're not keeping pace w new uploads, your monthly payment will continue to decrease. combine that w the fact that their rejection rate has gotten much higher. we know canva's income has tripled, but we don't know how much their overall portfolio has increased. they had stopped accepting new contributors for awhile, but now they do accept newbies
417
« on: September 18, 2023, 13:42 »
Matt, when accepting the terms of use of the Firefly plug-in in Photoshop, a message appears that commercial use of the beta version is prohibited. Does this mean that the Photoshop plug-in is still in beta or is it safe to upload files to Adobe Stock?
Firefly is out of beta. You can use it for commercial purposes. Update your main Photoshop on your computer through the Creative Cloud and you should be good to go.
-Mat
the confusion is what commercial means - we can use the images almost anywhere but can't submit these 'commercial' images to Adobe while we can submit all our other images, including ai generated
418
« on: September 18, 2023, 13:33 »
once again, you neither understand how these generators work, nor the massive programming involved. have you worked on such huge projects?
if everyone can optout at any time the training would have to be continuous, and there's no indication original images used would still be available - where are those billions going to be fud and how would they be able to identify your work?
but again, you dont understand how these work -; once trained, there is NO way to trace back to original training set.
419
« on: September 16, 2023, 13:16 »
my method which stores meta data both in images & an excel worksheet & tracks submission to various agencies The result is a solution that allows: - Storage of images for easy retrieval
- Fast sorting by date, topic, etc
- Instant access to any submission in progress
- Ability to track and analyze images accepted
https://cascoly-images.com/building-a-microstock-tracking-system-part-1
https://cascoly-images.com/building-a-microstock-tracking-system-part-2
I use this method to track over 100K of slide scans & digital images. These are stored on an external HD and cloud using Back Blaze. The main reason I don't use other systems is I don't process images in chronological order, especially during frequent travels. Eg, while I rename ASAP and enter minimal info as described in part 1, after that files move thru a variety of paths. Some go immediately to captioning & submission, others need processing thru Topaz AI Phot, deNoise, Sharpen& Giga pixel (usually in batches of 50-100 of images with a variety of dates). Posting processing is done later so files are stored in folders whether they need color correction, cropping, isolation, rotation, sky replacement, HDR, etc). Some finished files go only to agencies, others upload to my Pixify store and others are set aside for use in blogs. Originals and some intermediate images are archived.
It sounds complicated, but the use of many folders means everything is organized and there are no steps that block the pipeline. At any time there are thousands of images along the way, with the biggest holdups in post and in captioning.
420
« on: September 15, 2023, 14:03 »
Otherwise, I can't see any reason for which Shutterstock is giving recurring payment (every six months they said) for data training
Maybe it does revisit the same images from time to time. Or perhaps they're feeding it new images with each payment.
given the mechanics of the latter is almost surely the case (it would require saving all the original images somewhere, and much more likely given the billions of images involved, these are processed in large batches, then original images deleted) it's what SS explicitly says in its description of their data licensing program. a much larger problem that many of the images used have no owner info, and many images are duplicates as few only distribute thru 1 agency or service.
421
« on: September 15, 2023, 13:57 »
As far as I understand once AI is "trained" on the images it can't be untrained. So if we remove our images there's no difference, it's already done.
that's a function of the basic workings of data training, which is constantly misunderstood or ignored in these discussions (despite being explained countless times here) - it's not commutative - there's no way to go back from the dataset to the original images.
422
« on: September 13, 2023, 13:34 »
you're off by several orders of magnitude and show you haven't really studied how ai training works - there millions++ of datapoints (not '100 ') and the identity of the images used in training is lost in creation of the dataset as there is no longer a simple correspondence between initial pixels (24MP/image at minimum) and the resulting dataset. so when an ai image is generated there's no way to trace back
dont know your programming bkgd, but the solution you present is enormously complicated
423
« on: September 13, 2023, 13:21 »
a few rejects still appear - for focus, NLC and 'altered editorial' (which can just be a slightly blurred bkgd, accepted on resubmit.)
424
« on: September 12, 2023, 13:30 »
confused - if the order matters, why does LRC automatically alphabetize them?
One of the top features of Lightroom is the organizational capabilities. Alphabetical keywords makes sense for the vast majority of Lightroom use cases. We are the only exception I can think of.
As a compromise, the Lightroom team made an update in which the original order of keywords is retained as you add them. In LR the keywords are still displayed alphabetically, however when you upload to Adobe Stock, the keywords are added in the order you originally listed them. It's not the cleanest UI, but it is effective.
Personally, I add my keywords in Adobe Bridge to embed in metadata. I only use LR keywords to keep track of what I've got and where.
-Mat Hayward
thank s - does that change apply to both LR & LRC? i just use LRC for captioning & bulk resizing, as the catalog features dont matvch my workflow & naming conventions.
425
« on: September 10, 2023, 13:02 »
confused - if the order matters, why does LRC automatically alphabetize them?
Pages: 1 ... 12 13 14 15 16 [17] 18 19 20 21 22 ... 170
|
Sponsors
Microstock Poll Results
Sponsors
|
Show Posts
