Author Topic: So they do use AI to review then... (Read 29908 times)

Snow · « **Reply #25 on:** August 01, 2020, 04:42 »

Alright maybe I’ve missed this but I still don’t know if AI reviews how all those stolen images and spamfolio’s got online while some of us get tight and unjustified reviews?

Can anyone answer this?

I would perfectly believe AI was reviewing if it wasn’t for that. Are those portfolio’s part of The AI? Or part of keeping up apearance?

Wouldn't it be easy to code the AI reviewer to reject exact copies of other images to avoid stolen images?

Also is this happening only at SS?

To me it always felt like some do not get reviews at all while others have nothing but trouble with reviews and not because of low quality etc... so how do you explain that?

There must be flags on our ports that have influence on exposure (sales) and reviews.

There was (and probably still is) a time that people reported a halt in sales during a certain time period. As if your portfolio gets switched off while another gets turned on, to spread earnings and keep everybody somewhat motivated.

Im sure we have a few here who are familiar with creating search engines or even stock agencies so would love to read their input on this.

Tenebroso · « **Reply #26 on:** August 01, 2020, 04:59 »

The reviewer panel has a very large section for comments, orders, suggestions, for those who place the images. They know if a sunflower will appear as a sunflower, or in search of yellow, summer, virus, Christmas.
By this I mean that the reviewer panel is also full of annotations from the department head.

In addition, there are channels with specific Christmas themes, February 14, ......, those channels are always on Relevant. It is rare that their work is not constant if they sell a lot and well. It gives the feeling of channels on demand.

Similar recognition software works very poorly. It is very difficult. And besides, if it is cut and mirrored, it is almost impossible for the machine to determine if it is stolen.

In addition, images that everyone knows, famous images have been for sale.

I think answering your words are very difficult. They do not have the adequate image recognition software. But there is always the myth that it is a machine that examines, as the review is so inconsistent and ridiculous.

On the other hand, they accepted spam to the wild in COVID for example. All to cover the new files and send the client to Relevant.

If they change the name of the virus, it does not matter, the same images in Relevant.

If the World Health Organization calls it COVIDYA-19-20 today, the same relevant files will appear and the images that speak of the new name of the virus are covered with three pages of spam in new content.

Of course, they are all theories, conjectures, conspiracies and I speak of assumptions, and perhaps. I cannot guarantee anything.

Snow · « **Reply #27 on:** August 01, 2020, 05:51 »

Quote from: Tenebroso on August 01, 2020, 04:59

Of course, they are all theories, conjectures, conspiracies and I speak of assumptions, and perhaps. I cannot guarantee anything.

Thanks for your input.

Some like to put an assumption label on what we write but then if they can’t prove their counter-argument it’s nothing but assumptions as well and does not prove us Right nor Wrong.

Uncle Pete · « **Reply #28 on:** August 01, 2020, 10:17 »

Quote from: Snow on August 01, 2020, 05:51

Quote from: Tenebroso on August 01, 2020, 04:59
Of course, they are all theories, conjectures, conspiracies and I speak of assumptions, and perhaps. I cannot guarantee anything.

Thanks for your input.

Some like to put an assumption label on what we write but then if they can’t prove their counter-argument it’s nothing but assumptions as well and does not prove us Right nor Wrong.

You can't prove a negative, so terribly flawed reasoning.

Example: Prove that I don't have a Giant Invisible Frog living in my back yard? His name is GIF by the way.

How do you prove (counter-argument) that I don't?

The objective view is that someone making a claim, is responsible for the proof, not someone countering that there is no proof of the claim.

We don't know to what degree if any, that real IA is used. If the answer is, as reviewers, the human kind, make decisions, the program stores that data, it's not AI it's machine learning.

I suspect, as the OP posted and as many others might agree, that software is used for automatic intake accept/reject/pass on to a human.

Some easy rejections would be "lighting" reading a histogram. Size, color space, level horizon (there's a cans of works), contrast, focus? I'm sure there are many others, which could be reasonably accurate within the agency standards. What's that? 80% if it means not paying a human?

Grammar checker on the Title/Descriptions. That's easy and why the English and complete sentence rejections are so messed up. (oops, theory, I have no backing facts)

There's also software for the humans, that makes suggestions and reads qualities and parameters of the images.

I will agree that if the program has flaws, the results will also be consistently flawed. GIGO

Reminder I had accepted images, rejected a week or two later, because the whatever human or automated review made a mistake. That wasn't intake AI taking back the approval.

We don't know. We don't know if this started in 2012 or 2019. All we know is the SS report claims to be using software assistance and AI. They have listed their proprietary software for many years as a way they cut costs in reviewing.

The reviewers lawsuit, little details are available, was because they were paid as contractors not employees. The CA claim was based on that detail. https://www.classaction.org/news/class-action-claims-shutterstock-misclassifies-content-reviewers-owes-unpaid-wages

“As a result of [the defendants] misclassifying their Class Members as ‘independent contractors,’ [the defendants] have failed to itemize the total hours worked, overtime hours, and missed meal and rest periods on wage statements furnished to [the plaintiff] and similarly-situated Class Members.”

Of course filed in California the home of lawsuits because of the courts political leanings there.

cascoly · « **Reply #29 on:** August 01, 2020, 19:27 »

Quote from: Snow on August 01, 2020, 04:42

Alright maybe I’ve missed this but I still don’t know if AI reviews how all those stolen images and spamfolio’s got online while some of us get tight and unjustified reviews?

Can anyone answer this?

I would perfectly believe AI was reviewing if it wasn’t for that. Are those portfolio’s part of The AI? Or part of keeping up apearance?

Wouldn't it be easy to code the AI reviewer to reject exact copies of other images to avoid stolen images?

unfortunately, it would be costly - in development and server time plus SS isnt particularly motivated to fix the problem.

you'd have to search millions of images for each new image and do a cpu intensive comparison of each one. hacks would make it easier (eg skip if first 2K bytes are not similar) but there'd still be a hefty price.

additionally, we know what similars are, but the actual bits are probably very different - turning the object a few degrees makes a very different set of pixels

Uncle Pete · « **Reply #30 on:** August 02, 2020, 16:06 »

Quote from: Jo Ann Snover on July 31, 2020, 09:58

Quote from: DianeLambert on July 31, 2020, 08:26
... but there's no AI that can review a photo.

Sure there is - if you don't care about what ends up in the collection.

Have you seen the badly lit garbage they've been accepting over the last couple of months?

It should be an embarrassment to any agency to show that type of work, but Shutterstock is happily doing that. In the illustration department, there's misspelled garbage and endlessly repetitive flag combinations or simple patterns

I've been tweeting about this for weeks. Some examples:

https://twitter.com/joannsnover/status/1288187977311481856
https://twitter.com/joannsnover/status/1288153643045158915
https://twitter.com/joannsnover/status/1287887730462973952
https://twitter.com/joannsnover/status/1286739227300880384
https://twitter.com/joannsnover/status/1284252477575979010
https://twitter.com/joannsnover/status/1288683434852798464

There are many more examples (@joannsnover and #BoycottShutterstock) but you get the idea

Which is this one, in your opinion? System error, upload error or it's supposed to be this way, AI and it was accepted?

I was looking at new, since you started mentioning that.

If the image above disappears, someone found it. If not, there we are.

obj owl · « **Reply #31 on:** August 02, 2020, 17:09 »

Quote from: Uncle Pete on August 02, 2020, 16:06

Which is this one, in your opinion? System error, upload error or it's supposed to be this way, AI and it was accepted?

AI, Illustrations in my experience tend to pass review unhindered, unless there are similars. I would speculate that the lack of exif data has something to do with it.

Tenebroso · « **Reply #32 on:** August 02, 2020, 20:41 »

I recognize that this subject is beginning to be something personal to me.

If SS uses AI, the toads dance flamenco. It would have everything accepted if it were AI, in addition, it would make your AI dance flamenco too.

Any teenager from any corner of Bangladesh or Romania, Turkey, Israel, ...... we would have already put that AI to dance, to sleep, to feed it, to bathe it, on vacation, and we clothed it every night with the blanket and two kisses.

In addition, we would have prepared you to always place relevant images and not approve as many images from other collaborators, except from friends. We would take their AI for a walk, give it an ice cream and a coffee and study how they have achieved such a prodigy. We would make a source copy of your genes, to see your source code.

Then we would leave it in place, so that the parallel data does not detect the changes in the hosting, after six minutes, the CDNs from any corner of the planet would already give an updated version of the MariaDB database, an examination of the letter in 6 minutes.

Your CEO must give press conferences without sipping soup at the same time. Show off with ambiguous words to prove anything.

They are more computer scientists, software professionals at Adobe and more people, and more companies. Adobe takes less time to examine an illustration than an image, simply because they have more images to examine than illustrations, different equipment, however, it is evident that AS has new personnel in photographic examination, more numerous equipment in the last days.

So this is proof in your sentence of different teams of examiners. In your sentence, you yourself speak of different exams, a machine gives you the same illustration as it photographs, in short, it can neither see the file, nor understand, nor do anything with illustration or photography, in relation to microstock examination. Zero.

If your software reads the data from the .JPG to approve illustrations or photographs, we would have already reached the approval of the files in 5 hours. I don't hire the SS software team for web page design. They do not give me any confidence in their work. They must be trying to spend less money in this department too.

Two different teams, illustration and images. I highly doubt that an illustration has noise, blur. Her rejection is for the same nonsense as the rest, similar, title in English, keywords with advertising, brand and children's things, to comply with the order not to exceed 1,200,000 images a week.

Neither a microstock agency examines with IA, and less, SS, that they are not able to do anything in a coherent way. It has neither the financial nor the technical means.

Humanity cannot place machines today to decide whether an image can be sold or not.

A filtering software to detect a wifi icon in the batch of uploaded files and similar cancel, can be. Little more. If it were IA, we would have already made the IA, the CEO and the SS dance flamenco.

However, you can still think that SS has AI. Think what you want. They didn't approve files because they didn't want so many files. Only 1,200,000 a week. The rejections were absurd. An AI can't do that job today, right or wrong.

SS examines humans with orders not to pass more than 1,200,000 a week.

In short, they may approve everything, they are waiting for new orders, they cannot continue to throw good material in the trash. They are short of fresh material. I hope that those who upload material will confirm it when the HUMAN exam orders change.

The 0.10 Agency would like, in addition to sincerity without resorting to half-truths, to have modern software, for example, for the detection of similar and stolen images in the showcase of the database.

They would like to have an AI capable of examining. For starters, they don't need a SS if they have that tool, which would be a change in humanity. Something historical. At the moment, let the language translators improve. And then the text readers, OCR. AI to examine on microstock is science fiction today.

Tenebroso · « **Reply #33 on:** August 02, 2020, 21:29 »

IA can do the job of CEO in SS. Worse than him, it is impossible. AI cannot impersonate humans to examine microstock files. We are going to wait for them to learn in the vehicles to decide in case of an accident and collision, how to handle the AI situation in the vehicles. We will have security first in the technological devices of the vehicles, fight to avoid terrorist attacks in vehicles with technological devices.

Let AI learn to interpret driving. Let him learn how to remove noise from an image. Learn how to spot the red words and codes in random wiretapping for monitoring humans of potentially dangerous targets. And let the time pass before AI can see, understand, and decide on an image.

It will be the last place where an AI can act, related to art. It is the most human thing we have, the sentiments.

Tenebroso · « **Reply #34 on:** August 02, 2020, 21:41 »

The day that a machine sees the Eiffel Tower as something beautiful, instead of an iron that hinders the main focus of the landscape, then, we will have made the software begin to interpret the likes, common sense, trends and feelings of humans.

Tenebroso · « **Reply #35 on:** August 02, 2020, 21:58 »

The 0.10, believe themselves Heavenly Gods of the Galaxy, but you do not play their game, they are not Divine Gods in the Galactic Heavenly Universe. Let them do their job well, instead of showing off with half truths. If they had AI capable of deciding that a logo is good for sale, and that it's on the side to make the design easier for customers, that the image is not off-centered by mistake, then, I would say, SS are Gods of the Galaxy.

I'm going to do things, I'm too busy to mess around with SS.

gnirtS · « **Reply #36 on:** August 05, 2020, 12:51 »

This amused me. Not all reviews are unfair.

In other news, a "contributor" who literally doesnt understand composition. (or it seems anything else).

It does show how low the bar is now for contributing though.

Tenebroso · « **Reply #37 on:** August 05, 2020, 14:21 »

That image can be commercial, for forums dealing with photography and image composition.

cascoly · « **Reply #38 on:** August 05, 2020, 17:35 »

Quote from: Tenebroso on August 02, 2020, 21:58

The 0.10, believe themselves Heavenly Gods of the Galaxy, but you do not play their game, they are not Divine Gods in the Galactic Heavenly Universe. Let them do their job well, instead of showing off with half truths. If they had AI capable of deciding that a logo is good for sale, and that it's on the side to make the design easier for customers, that the image is not off-centered by mistake, then, I would say, SS are Gods of the Galaxy.

I'm going to do things, I'm too busy to mess around with SS.

you continue to deny AI exists - please give us YOUR definition of AI and what is required to be accepted as AI

Tenebroso · « **Reply #39 on:** August 05, 2020, 19:18 »

My interpretation does not exist. I do not have a particular interpretation, nor proper of the concept of AI. I don't have a different concept of AI in microstock agency.

A software with the appropriate parameters to accept or reject a file for sale in a Microstock Agency. A software that by itself is interpreting, improving and learning about the work done, always supervised by humans, learning its function.

There is no team on the planet today capable of coming out of the Alpha version of this supposed AI for microstock in 10 or 12 years. And without guarantee of success.

It is unfeasible today, even if you have money, technique and talent to achieve this purpose.

Simply an extraordinary job to make him understand the difference of the concept DEATH, by the title of a file, "Death to the Collaborators who criticize SS", "Death of potatoes by snails" or "Trapeze artist facing death in the Rio Canyon Little.

For a machine to see explicit sex, it is something that any multinational company of the platforms that we all know would gladly pay for, simply a software that determines a file that contains explicit sex would be a great advance.

Nowadays, explicit sex is usually seen where there is none, and sex from someone naked with a painted body sneaks in. The Hate Speech thing, another great advance, that Google and the others would like to have. Not only in English, but an insult in an image to Christmas. The machines only read completely flat text nowadays, and with many errors, not design text. It is impossible, there is no money, no technique to examine with AI in microstock.

As for SS, your team, when you modify something, don't even check the software in other languages, with longer words, the save button on the panel is missing. They are very disastrous, perhaps for not spending money.

To get an AI to examine, it would be chaos for SS. They must test privately in Alpha version, and for several decades they would be developing code. And with no guarantee of success. Of course, the money from this investment is not suitable for SS. The investment is more expensive than the entire company. Also, if they get it, it has more value than SS.

It is better to examine randomly, since there are no real possibilities of this technology in our days.

AI, software, that develops a preset function, and that learns by itself.

In this case, he cannot learn anything, nor develop his mission, he cannot understand that a swarm of bees are not strange artifacts in the movie. Teaching them to see birds, flies in the landscape was already an extraordinary advance, only each of the functions, if it could exist, would be a great advance. Now put all these advances to examine files, it is simply impossible.

Seeing an anonymous mask and detecting it as wrong for copyright, is easy, but it is not easy if the mask is on its side, or if the mask is at a protest rally in the streets of London. Any victory in each field would be a great achievement. Logically, it is not possible today to replace a human in microstock exams.

Teaching him that there are humans with a pale complexion and mustaches that is not an anonymous mask, it would be an achievement to achieve this. Multiply this example exponentially for each specific case that may occur to you. Without insisting, that there may be text in the illustrations, impossible to read. You must teach her the parameters for bikini on the beach. Define the bodies on a beach as non-gangbang and learn to identify these images. To learn about sex between animals and humans, since you can see affection in a zoophilia archive. It is simply unattainable today.

Interpreting child abuse in an inappropriate image or insinuation would be extremely helpful to security agencies on the planet.

Les · « **Reply #40 on:** August 05, 2020, 19:22 »

Calling the SS selection algorithms AI is a bit magnanimous. There are other terms which would be more fitting.

Tenebroso · « **Reply #41 on:** August 05, 2020, 19:25 »

Quote from: Les on August 05, 2020, 19:22

Calling the SS selection algorithms AI is a bit magnanimous. There are other terms which would be more fitting.

Exact.

Tenebroso · « **Reply #42 on:** August 05, 2020, 19:38 »

It is likely that they have some filter, mainly from similar files in the same shipping batch.

A filter before the supervisor. But with a lousy result. And that its CEO does not care. Not distinguishing an mp3 and mp4 illustration as very different things.

However, to avoid spending money, they may have a filter that interprets the approach. Also a noise filter.
However, it is humans who examine. And if you want it to be a software, randomly does it better than the examiners.

They simply reject some files to almost all the collaborators, because they do not want to expand the team of placing images, a team that has to pay money, with 3 million images a week, they accept 1,200,000 a week and that money stays in the box. With ridiculous rejections.

We will see how they examine now. In the coming months. With fewer files in your entry.

Cobra · « **Reply #43 on:** August 05, 2020, 19:40 »

Quote from: Tenebroso on August 05, 2020, 19:18

My interpretation does not exist. I do not have a particular interpretation, nor proper of the concept of AI. I don't have a different concept of AI in microstock agency.

A software with the appropriate parameters to accept or reject a file for sale in a Microstock Agency. A software that by itself is interpreting, improving and learning about the work done, always supervised by humans, learning its function.

There is no team on the planet today capable of coming out of the Alpha version of this supposed AI for microstock in 10 or 12 years. And without guarantee of success.

It is unfeasible today, even if you have money, technique and talent to achieve this purpose.

Simply an extraordinary job to make him understand the difference of the concept DEATH, by the title of a file, "Death to the Collaborators who criticize SS", "Death of potatoes by snails" or "Trapeze artist facing death in the Rio Canyon Little.

For a machine to see explicit sex, it is something that any multinational company of the platforms that we all know would gladly pay for, simply a software that determines a file that contains explicit sex would be a great advance.

Nowadays, explicit sex is usually seen where there is none, and sex from someone naked with a painted body sneaks in. The Hate Speech thing, another great advance, that Google and the others would like to have. Not only in English, but an insult in an image to Christmas. The machines only read completely flat text nowadays, and with many errors, not design text. It is impossible, there is no money, no technique to examine with AI in microstock.

As for SS, your team, when you modify something, don't even check the software in other languages, with longer words, the save button on the panel is missing. They are very disastrous, perhaps for not spending money.

To get an AI to examine, it would be chaos for SS. They must test privately in Alpha version, and for several decades they would be developing code. And with no guarantee of success. Of course, the money from this investment is not suitable for SS. The investment is more expensive than the entire company. Also, if they get it, it has more value than SS.

It is better to examine randomly, since there are no real possibilities of this technology in our days.

AI, software, that develops a preset function, and that learns by itself.

In this case, he cannot learn anything, nor develop his mission, he cannot understand that a swarm of bees are not strange artifacts in the movie. Teaching them to see birds, flies in the landscape was already an extraordinary advance, only each of the functions, if it could exist, would be a great advance. Now put all these advances to examine files, it is simply impossible.

Seeing an anonymous mask and detecting it as wrong for copyright, is easy, but it is not easy if the mask is on its side, or if the mask is at a protest rally in the streets of London. Any victory in each field would be a great achievement. Logically, it is not possible today to replace a human in microstock exams.

Teaching him that there are humans with a pale complexion and mustaches that is not an anonymous mask, it would be an achievement to achieve this. Multiply this example exponentially for each specific case that may occur to you. Without insisting, that there may be text in the illustrations, impossible to read. You must teach her the parameters for bikini on the beach. Define the bodies on a beach as non-gangbang and learn to identify these images. To learn about sex between animals and humans, since you can see affection in a zoophilia archive. It is simply unattainable today.

Interpreting child abuse in an inappropriate image or insinuation would be extremely helpful to security agencies on the planet.

You write like a lawyer. Long winded and really nothing is achieved. Just goes in circles forever...

Tenebroso · « **Reply #44 on:** August 05, 2020, 19:51 »

I know I get nothing. Thank you. Also, understanding myself is very difficult. I know.

I'm going to do things, it's still an extraordinary waste of time.

Tenebroso · « **Reply #45 on:** August 05, 2020, 20:05 »

I have decided to accept Octopus as a pet. I'm wrong, SS has AI in file monitoring.
A mistake on my part, sorry. I was confused, the night confuses me.
SS has AI in file exams.

cascoly · « **Reply #46 on:** August 05, 2020, 21:28 »

Quote from: Tenebroso on August 05, 2020, 19:18

My interpretation does not exist. I do not have a particular interpretation, nor proper of the concept of AI. I don't have a different concept of AI in microstock agency.

....

unbelievable!!!! i thought you were serious, but have absolutely nothing in support? i won't bother to read your screeds from here on in

Tenebroso · « **Reply #47 on:** August 05, 2020, 22:08 »

I am not used to having my own technical definitions of computer terminology and / or telecommunications, depending on the day or as a menu, to my liking, on objective concepts, established, known, supported and accepted terms. I work, study and live with them, I do not modify them according to my preferences or tastes or mood.
You make me laugh. Thank you.

leaf · « **Reply #48 on:** August 06, 2020, 02:24 »

Quote from: Tenebroso on August 05, 2020, 20:05

I have decided to accept Octopus as a pet. I'm wrong, SS has AI in file monitoring.
A mistake on my part, sorry. I was confused, the night confuses me.
SS has AI in file exams.

Perhaps this user is an AI.. or getting text from ai..

Very funny lol

Snow · « **Reply #49 on:** August 06, 2020, 03:44 »

Quote from: leaf on August 06, 2020, 02:24

Quote from: Tenebroso on August 05, 2020, 20:05
I have decided to accept Octopus as a pet. I'm wrong, SS has AI in file monitoring.
A mistake on my part, sorry. I was confused, the night confuses me.
SS has AI in file exams.

Perhaps this user is an AI.. or getting text from ai..
Very funny lol

It's just bad translation no? at least I hope it is!

MicrostockGroup Sponsors

Author Topic: So they do use AI to review then... (Read 29908 times)

Snow

Tenebroso

Snow

Uncle Pete

cascoly

Uncle Pete

obj owl

Tenebroso

Tenebroso

Tenebroso

Tenebroso

gnirtS

Tenebroso

cascoly

Tenebroso

Les

Tenebroso

Tenebroso

Cobra

Tenebroso

Tenebroso

cascoly

Tenebroso

leaf

Snow

Related Topics

Sponsors

Microstock Poll Results

Sponsors