|
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 ... 25 26 27 28 29 [30] 31 32 33 34 35 ... 170
726
« on: January 01, 2023, 21:02 »
...
All this is true, however one must be aware about plagiarism when something "new" is created.
There is a level beyond which plagiarism fades away, but I can imagine that when someone is trying to create some niche images, with few training samples, elements from the images used for training may appear with little or no modification in the "newly" created content.
This is no different than copying entire paragraphs from a book and then claiming that you are the author of the "newly" created content.
This will most likely continue to be something for the lawyers to argue about.
bold added
that's a strawman argument - no training set contains only a 'few' images, and no 'elements' of any image is copied. that's just not how ML works
I know how ML works (using several variants in my daily job).
So how can you tell the number of elements used to train a specific request, without being involved in the algorithm development? What you say may be true only if there a minimum threshold for the training set, a threshold beyond which individual image characteristics are fading away.
You have to know it, before making such statements.
If such threshold doesn't exist, some requests may simply plagiarize the few images used to respond to that query.
If there is only one image describing, let's say a clown, in the training set, then it's very likely that all queries requesting clowns will plagiarize that unique clown image, because that's the only thing the algorithm has learned about clowns.
still setting up strawmen for your arguments - how about a real-life example of a request that won't have hundreds if not many thousands of images it's been trained on? simple way would be to find any search on SS that has < 100 images & then ask dall-e et al for that. SS has 200,000,000 images for their training set - web scrapers can have many more (one recent example claimed 2 billion) so the chances of your scenario are pretty small and no one has been able to demonstrate this sort of result.
727
« on: January 01, 2023, 18:56 »
My 2022 year end review in a single chart....
what is it showing? number of bugs in SS contributor page? time spent watching youtube?
728
« on: January 01, 2023, 16:05 »
I traveled a lot in the last year but most effort has gone into building my pixify site & increasing SocialM presence, including a recent start on mastodon. For anyone using FAA & who's also on twitter, Mastodon et al, or with a personal website, check out 2023 A Year For Art And A Successful Marketing Promotion More details there, but basically, it's a group of artists who cross-promote each other's work. there are currently about 40 active members with twitter and/or mastodon accounts. also lots of great tips and suggestions. Lately a lot of help in setting up a mastodon account. I first learned about this discussion/group from another MSG member. There's a lot of potential symbiosis here between MSG and SM, but so far it's underused. and for bloggers, there's also Drop Your New Blog Post Links Here! which cross-promotes blogs thru comments my social media: blog: https://cascoly-images.comhttps://www.facebook.com/steve.estvanikhttps://mastodon.art/@cascoly https://twitter.com/cascolypixify/shopify https://cascoly.photographyPost your sites here - i'll followback anyone who follows me
729
« on: January 01, 2023, 14:59 »
My only comment is that from where I sit (mostly photo portfolio) SS revenues are declining and AS revenues are increasing.
the more important question is which agency produces the most income? for me SS consistently earns about twice what AS does.
730
« on: January 01, 2023, 14:51 »
2022 mtm was sporadic but only 6% less than 2021 and down 14% from 2020
I traveled a lot in the last year but most effort has gone into building my pixify site & increased SocialM presence, including a recent start on mastodon. many fewer submissions haven't affected monthly sales by much (classical ruins & vintage art haven't changed much over the last year)
731
« on: January 01, 2023, 14:30 »
...
All this is true, however one must be aware about plagiarism when something "new" is created.
There is a level beyond which plagiarism fades away, but I can imagine that when someone is trying to create some niche images, with few training samples, elements from the images used for training may appear with little or no modification in the "newly" created content.
This is no different than copying entire paragraphs from a book and then claiming that you are the author of the "newly" created content.
This will most likely continue to be something for the lawyers to argue about.
bold addedthat's a strawman argument - no training set contains only a 'few' images, and no 'elements' of any image is copied. that's just not how ML works
732
« on: December 30, 2022, 20:45 »
I just got an email from SS titled: Contributor earnings now reflect generative AI Here's the first paragraph of the email:
News for artist recognition in advanced technology worldwide, and news for Shutterstock contributors!
You will now accrue royalties when your IP is used in the training of AI-generative models or used for licensing of generative assets created using Shutterstocks software. Contributing artists can view their earnings in their account in the Earnings Summary 'By month' tab as 'Contributor fund.'
...
, but otherwise, I will not allow that, when given the option.
problem is, you've already given them permission according to the TOS you agreed to. SS may choose to allow people to opt out but haven't done so yet
733
« on: December 30, 2022, 14:10 »
the TOS seem to let SS use images for training (and doesnt even require any payment)
By submitting any Content to Shutterstock, you grant to Shutterstock a worldwide, sublicensable, non-exclusive right and license to index, analyze, categorize, archive reproduce, prepare derivative works incorporating, publicly display, sell, advertise and market, any Content uploaded by you and accepted by Shutterstock for any reasonable business purpose, including but not limited to the distribution of your Content to Shutterstock customers, to optimize the performance and operation of Shuttestocks platform and services, and to develop new features and products.
734
« on: December 30, 2022, 14:05 »
I just got an email from SS titled: Contributor earnings now reflect generative AI Here's the first paragraph of the email:
News for artist recognition in advanced technology worldwide, and news for Shutterstock contributors!
You will now accrue royalties when your IP is used in the training of AI-generative models or used for licensing of generative assets created using Shutterstocks software. Contributing artists can view their earnings in their account in the Earnings Summary 'By month' tab as 'Contributor fund.'
This is gonna be a few cents so they are able to reduce the likelihood of a successful lawsuit. And it is gonna be only for a few years until we are completely phased out.
this is what we've been saying all along - each artist is a tiny drop among 200 million images they can use for training and the overhyped fears are just silly - why would they delete those 200 million images which are already highly placed on google? shift happens assuming they used all 200 million, based on my numbers, the payout seems to be about .3 cents ($.003) /image. so someone w 10K images would get about $30 - does that correlate with what others have reported?
735
« on: December 30, 2022, 13:47 »
would upscaling work for old 2mpix (1600x 1200) photos? because I have a lot of good photos taken in the early 2000s but they were taken on my old Nikon Coolpix & Olympus P&S cameras.
it's worth a try - topaz has a free trial
736
« on: December 29, 2022, 15:16 »
...
New images are not created based on something previously copyrighted. The new images are based on Machine Learning and the images are new creations. AI doesn't use our images, it uses what the machines has learned about something.
So if the machine has been fed 2,000 images of a Banana, it learns that a banana is is basically, black, or brown, or green or yellow. Then the machine learns what a banana shape is. The machine also learns characteristics of the middle and ends and lines and other parts, maybe the insides as well as other parts or slices or variations of a banana. When someone says they want an image, created by AI that includes a banana, the computer uses everything it knows and creates a new banana. Our images are not directly used.
yep but no matter how many times this is explained, some folk continue to spew stories about images being used directly by the AI. the issue is about the one-time use to create the dataset which contains NONE of those images. Who owns the results? That depends on the service someone uses. Open AI / DALL-E2 we own the rights and the images can be used commercially. But Open AI also retains the right to use the image, if they want.
Others you can only use commercially if you paid for the service. And I'm surer there are other versions of the contract.
This will have to go into the courts and be decided, and this is new territory, so there isn't much case law to use for an answer. TBD. One case the courts decided that an AI creation, can't be copyrighted.
...
which becomes really silly since there's no way to tell an image is AI or just 'normal' use of PS unless it's tagged that as AS requires. even more critical for text created from chatGPT where it's even harder to detect "Did a Fourth Grader Write This? Or the New Chatbot?
Dont be surprised if you cant always tell."https://www.nytimes.com/interactive/2022/12/26/upshot/chatgpt-child-essays.htmli correctly identified only about 70% of the essays I've been using chatGPT for blogging & generic descriptions for my pixify site. I treat the results as a first draft, making small to substantial changes. but it is useful to overcome writer's block by getting something on the page
737
« on: December 28, 2022, 17:46 »
...and will make an effort to learn how to use the new tool to my advantage...
I wouldn't bother. The interface will get more user friendly and the results better so fast even the minimal effort to learn how to use it now is time wasted. Just wait a couple of months and it will be a lot more intuitive. No point learning typesetting cast metal sorts when the wordprocessor will be invented before you can master it. AI will be able to do the post processing much better and faster than us soon enough.
thus becoming a self-fulfilling prophecy of doom. and those who adapt will at least have a chance rather than none at all. it's dancing among elephants - few will be Stieglitz or Ansel Adams, but niches are there to be exploited. Not every buyer will even have a specific need when they're looking to illustrate a concept the wordprocessor didn't put all writers out of work, it streamlined the process. oldtime programmers will remember the idea of desk checking - spending time checking & rechecking code for typos, logic errors,etc before submitting to be keypunched, then getting time on THE computer, only to have it spit out a vague "you've got errors" msg. Along came interactive compilers that displayed your errors, then suites that did large parts of design. downloadable drivers eliminated coding for each device; internet browsers replaced programming modem input/output; databases replaced the need to for coding the most efficient sorting algorithms ('bubble sort' anyone? 'hashing' algorithms? linked lists?) we had workarounds that did multi-tasking before windows, and modems & semi-hard disks were useful, but they were cumbersome & limited. each step eliminated some work/ers but those who swam with the flow had more interesting careers (and higher salaries). many computer consultants would not be able to find work if they had to constantly re-create the wheel. now chat &others are starting to write code, but people will still be needed to design the app. (and current state of design needs more intelligence in general - with too many poorly designed apps being thrown untested into the wild with horrible design flaws)
738
« on: December 24, 2022, 14:11 »
yes, likely next steps - but what difference are you making by hiding your head in the sand while some of us see the future and are learning to deal with it. it's not a fast buck, it's facing reality.
if you've actually used an AI you'd realize that sending a phrase doesn't automatically create usable images without a good deal of post-processing. & buying from agencies will still be cheaper than paying salary & benefits to an in-house AI user.
Why do you think you are "learning to deal with it" in regards to the problem in my post, meaning, how are you learning to deal with the fact that in the future customers will enter their search terms in AI image generators instead of microstock sites and you won't be needed to do that for them?..,.
because it's not a 'fact' - just your improbable prediction - there's no reason to believe most customers will switch for reasons discussed here many times. they can buy images for pennies rather than pay the cost of an employee's time to create images 'dealing w it' = accepting it's coming (even faster by your estimates!) but your solution is to surrender w/o adjusting - just as many film photographers lost by refusing to adapt to digital. luddites are always on the wrong side of innovation & history
739
« on: December 24, 2022, 14:05 »
I am curious if anyone is up scaling images for stock?
I have some smaller images and I was thinking about this.
i use topaz giga with good results - i upscale x4 then reduce to 6mp and accepted by all. even scans of slides if grain is minimal for PS HDR i almost always have to post-process because it usually has a lot of sky noise, so i use deNoise (if i can't use a blue sky-replace) then giga. again almost always accepted
740
« on: December 23, 2022, 13:45 »
...
And the whole problem goes much further than this. The next step after AI generated paintings and photos will be AI generated videos, AI generated stories, AI generated music. In a 100 years there won't be any artists left, because who ever will persuit art, if no one can live from it anymore, because an AI can do it faster and cheaper than you anyways. And that's incredibly sad, because humanity is basically killing off part of what makes it human. I am glad I will not live long enough to see that art-deprived mankind where human creativity has been snuffed out.
yes, likely next steps - but what difference are you making by hiding your head in the sand while some of us see the future and are learning to deal with it. it's not a fast buck, it's facing reality. if you've actually used an AI you'd realize that sending a phrase doesn't automatically create usable images without a good deal of post-processing. & buying from agencies will still be cheaper than paying salary & benefits to an in-house AI user.
741
« on: December 22, 2022, 15:00 »
...,
They could make two step verification standard and optional. Some people can't do that, but hardly anyone is without a cell phone anymore. This assures that the account holder who is registered, is the one logging in. Also prevents multiple accounts and returning thieves to some extent.
...
any fairly easy addition to 2-step is a security question - MailChimp does this. we use it for a board where i'm webmaster & get the verification notice but others have occ'l need for access
742
« on: December 22, 2022, 14:45 »
I liked very much your post and think you are right as you say that AI is a tool - but it's a disquietingly clever one.
Your words started a disturbing chain of thought that I made into a post ....
Maybe it's just me, or maybe I've read too much Asimov... 
check out https://www.cold-takes.com/ai-could-defeat-all-of-us-combined/
#1 unplug the computers?
His premise starts to get strange around this part "
AIs could recruit human allies, tele-operate robots and other military equipment, make money via research and quantitative trading, etc.
At a high level, I think we should be worried if a huge (competitive with world population) and rapidly growing set of highly skilled humans on another planet was trying to take down civilization just by using the Internet. So we should be worried about a large set of disembodied AIs as well.
"
They could tele-operate robots and military equipment, for what gain? What need is there for money if the AI controls everything? highly skilled humans on another planet Realy, now we are back to War of the Worlds kind of scenarios? Disembodied AI will be like alien invaders?
Could some AI network manipulate the entire human population? Whoever controls the network, controls the world.
Lets say all the "good" people in the world say, we won't create any AI that could take over. We won't create AI... How would they have the knowledge of how that AI works, when someone evil tries to build the system to take over the world? In other words, the best defense, is to understand the weapon that will be used to attack us. Maybe create something that can defeat it?
But I still say, if some AI network starts to self create, flip the switch, turn it off.
that post has been followed by more details & addresses the points you raise - esp'ly about whether 'good' companies should develop dangerous AI to counter those who don't care https://www.cold-takes.com/ai-safety-seems-hard-to-measure/overall summary & references https://www.cold-takes.com/most-important-century/
743
« on: December 20, 2022, 17:17 »
I liked very much your post and think you are right as you say that AI is a tool - but it's a disquietingly clever one.
Your words started a disturbing chain of thought that I made into a post ....
Maybe it's just me, or maybe I've read too much Asimov...
check out https://www.cold-takes.com/ai-could-defeat-all-of-us-combined/
744
« on: December 19, 2022, 13:42 »
How would you like stock photography platforms to handle the uploading of AI-generated content?
Should there be a difference between completely-AI-generated content, and AI-assisted content in which there is substantial human work and editing involved?
If AI-generated content is allowed, should royalties and earnings of the AI-generated content be shared with human creators? If so, what share would be appropriate?
AI generation is a tool - just like the embedded AI in PS neural processing and others. So it should be treated exactly the same as any other submitted content and how would you determine it's AI generated if the artist doesnt label it as such? I like AS approach which, for now, accepts it but requires it to be tagged appropriately
745
« on: December 18, 2022, 20:16 »
...
Man- is that difficult to understand.... I had it translated into german but that makes it even worse to understand 
What the heck is a "Quellensteueragent"
...
maybe they shpould be using chatGPT! looks like that's a 'with-holding agent' - eg agencies like canva which hold part of your earnings as estimated taxes. when you list them on your tax form they reduce your taxes owed. canva is the only agency i know who does this automatically
746
« on: December 18, 2022, 20:10 »
i've been using chatGPT for several weeks now & it's amazing in some areas- i've found a few factual errors but overall it's incredible - i've used it get descriptions for my blog. sometimes i can use it directly, others it breaks writer's block w a great first draft. I'm also working with it to write stories.
it also does poetry & limericks (sort of)- few laureates will be displaced, but greeting card writers should be looking for other jobs!
Ngorongoro Crater, a place of wonder
A true masterpiece, created by thunder
A natural wonder, for all to see
A treasure of the earth, for you and me
748
« on: December 17, 2022, 14:09 »
had to scroll way over - found $112 there today
that's more than AS earns in a month - except for the annual(?) addition to free collection
these 'extras' from canva,AS & now SS really make a difference in the bottom line
750
« on: December 17, 2022, 14:04 »
Interesting, rather than a long quote, here's a link to the Creative Common CC-BY 4.0 license.
https://creativecommons.org/licenses/by/4.0/
completely false - CC license specifically says credit must be given to the creator among SEVERAL other restrictions
i've contributed several images when editing wiki, but it's a bit of a hassle
I think you have a point there? I don't use CC or offer them, so this is the first I've actually read anything about how this is allowed or not.
Pages: 1 ... 25 26 27 28 29 [30] 31 32 33 34 35 ... 170
|
Sponsors
Microstock Poll Results
Sponsors
|
Show Posts
