The Ai thread

[dada_dave]

Copyright law. To protect the artists and creators?… Nono. To protect their employers and capitalists profiting off of the art

The Capitalist motto, “Not-Neutral Lawful, It’s About Enriching My Pile of Gold”.

It's bad enough they do it off the back of laborers, now they're coming for artists.

As it stands I've had to ban the terms "AI" and "fake" in the comments on my Instagram account because so many users were claiming my real work is fake. I'll just say creators create, whether planning photo shoots, traveling, etc. or painting a scene, they put their time and effort into real work. It's a shame to see AI obliterating that.

And of course when it comes to protecting “their” training data suddenly OpenAI is a lot less concerned with fair use and much more concerned with protecting what’s “theirs”. What’s yours is mine … and what’s mine is still mine.

View attachment 33645
OpenAI and MS investigating DeepSeek for stealing training data … is irony dead yet?

 

[Eric]

Copyright law. To protect the artists and creators?… Nono. To protect their employers and capitalists profiting off of the art

Not exactly confident they're just admiring my work here lol.

For reference I do a lot of what is often misconstrued as AI.

 

[rdrr]

Not exactly confident they're just admiring my work here lol.
View attachment 34286

For reference I do a lot of what is often misconstrued as AI.

I really like your work, and I know some of the techniques you are using. I can understand why people who don't know the techniques you use in your art, think that this is AI.

 

[Eric]

I really like your work, and I know some of the techniques you are using. I can understand why people who don't know the techniques you use in your art, think that this is AI.

It's nothing more than simple optical lens compression applied to movement, a basic concept that has been around forever. Now, because of AI, everyone is confused by it.

 

[KingOfPain]

Since AI companies don't care if their products spew false information, these stories will likely multiply in the near future:

Dad demands OpenAI delete ChatGPT’s false claim that he murdered his kids

Blocking outputs isn’t enough; dad wants OpenAI to delete the false information.

arstechnica.com

 

[dada_dave]

It's nothing more than simple optical lens compression applied to movement, a basic concept that has been around forever. Now, because of AI, everyone is confused by it.

To be honest I’d be confused anyway! How does it work?

 

[Yoused]

How does it work?

A telephoto is essentially a telescope. It allows you to get a better view of a distant object by manipulating the light to seemingly bring that object closer. The effect of this is that it compresses the distance between the viewer and the object. You can see this in some movies, such as the original Vanishing Point, in which the car is approaching at 120mph, but the distance is so compressed by the lens that it looks like it is barely moving. With a long lens and a lot of depth-of-field (small apperture that puts more of the image in focus), movement of near and distant objects looks odd because of the way the distance between them is compressed in the image.

For instance, imagine you have a 4x lens. There is a person 20' away from you and a person 100' away from you, almost hidden by the nearer person. In your video, they appear to be 5' and 25' away. Now, while you are recording, move the camera one foot to the right. As you look with your eyes, they separate by a normal distance, but, in the video, the person who appears to be 25' away moves by a distance that the viewer would expect a person 100' away to move – that is, the separation looks to be much greater than it should be, based on normal perspective.

 

[Eric]

A telephoto is essentially a telescope. It allows you to get a better view of a distant object by manipulating the light to seemingly bring that object closer. The effect of this is that it compresses the distance between the viewer and the object. You can see this in some movies, such as the original Vanishing Point, in which the car is approaching at 120mph, but the distance is so compressed by the lens that it looks like it is barely moving. With a long lens and a lot of depth-of-field (small apperture that puts more of the image in focus), movement of near and distant objects looks odd because of the way the distance between them is compressed in the image.

For instance, imagine you have a 4x lens. There is a person 20' away from you and a person 100' away from you, almost hidden by the nearer person. In your video, they appear to be 5' and 25' away. Now, while you are recording, move the camera one foot to the right. As you look with your eyes, they separate by a normal distance, but, in the video, the person who appears to be 25' away moves by a distance that the viewer would expect a person 100' away to move – that is, the separation looks to be much greater than it should be, based on normal perspective.

Well said. Here are a couple of more extreme examples, I've really taken to this kind of photography and carved out a niche in it in an otherwise crowded field of local photographers.

The 7x camera on my drone is 166mm, I spend a lot of time scouting and flying these locations to get exactly what I want out of it. While it's frustrating that people call it AI, I also get why.

 

[cbum]

Here are a couple of more extreme examples

If I had to take a guess, I'd say you love SF...

Takes me way back - 45 years ago I was doing an infectious disease rotation as visiting student at Moffitt Hospital. Still remember an ICU consult on a previously healthy young gay man who developed a severe Pneumocystis infection out of the blue - no one had ever seen anything like it since P. carinii is an opportunistic infection only seen in severely immune-compromised people ( .... ).
The ID fellow who was in charge of the consult told us "a friend on mine just saw the same thing at UCLA..."

The ID fellow later was the first author on the first clinical paper of a new disease called "AIDS"...

 

[KingOfPain]

The irony...

CEO of AI ad-tech firm pledging “world free of fraud” sentenced for fraud

Prosecutors: Firm offering “300% more” fraud detection oversold revenue by 700%.

arstechnica.com

 

[Yoused]

Are there LLMs that recognize irony and sarcasm?

 

[dada_dave]

A follow on article about people deliberately creating AI companions (both developers and their users).

What do you love when you fall for AI?

Inside the surprisingly meaningful, unexpectedly heartbreaking, and deeply confusing reality of AI relationships.

www.theverge.com

An incredible article, a little disturbing to be honest, but well worth the read.

Follow-on article in a way:

Some ChatGPT users are addicted and will suffer withdrawal symptoms if cut off, say researchers

OpenAI and MIT worked together on a study of four million interactions over 28 days.

www.tomshardware.com

Basically reinforcing the above human-interest angle reporting with a study from MIT-OpenAI that people can become addicted to chatbots. Though one wonders if OpenAI’s goal is how to further that. As the article says: “business is business” after all.

 

[Nycturne]

Basically reinforcing the above human-interest angle reporting with a study from MIT-OpenAI that people can become addicted to chatbots. Though one wonders if OpenAI’s goal is how to further that. As the article says: “business is business” after all.

I mean, it’s what social media and content platforms are aiming for, intentionally or not. As I mentioned in another thread, metrics can lead to dark patterns by being able to do A/B testing while measuring things like "minutes spent on the home feed per user" and wanting that line to go up. So I would be entirely unsurprised if OpenAI has a metric like "minutes spent in a chat per user" and trying to get that line to go up with changes.

 

[KingOfPain]

Security restrictions of LLMs breached by gaslighting them that they are running in an archived state 50 years in the future:

Librarian Bully Attack: Gaslighting State-of-the-Art LLMs into Obsolescence

Psychology-based LLM-Jailbreaks: How Temporal Manipulation and Psychological Exploits Undermine Security Across All Transformer Models

humandataexperience.substack.com

 

[ArgoDuck]

LLMs and theft: Some of you may wish to search here to check whether any of your published (copyrighted) output has been scooped up in Meta’s training sets. It’s likely OpenAI, Google etc have done this too.

(And yes, some of my own and colleagues’ work is listed).

 

[KingOfPain]

Some insight into how a LLM works. The math one is pretty weird...

https://www.anthropic.com/research/tracing-thoughts-language-model

 

[dada_dave]

 

[Citysnaps]

LLMs and theft: Some of you may wish to search here to check whether any of your published (copyrighted) output has been scooped up in Meta’s training sets. It’s likely OpenAI, Google etc have done this too.

(And yes, some of my own and colleagues’ work is listed).

The majority of my photography is of strangers I hit up on city streets for a bit of conversation and a portrait or two. And is why I've stopped posting my work on social media and books I publish of easily identifiable people.

 

[Eric]

The majority of my photography is of strangers I hit up on city streets for a bit of conversation and a portrait or two. And is why I've stopped posting my work on social media and books I publish of easily identifiable people.

They say you gotta have those model releases handy but I never do. The other day I filmed a family on a bluff from a distance with my drone and their faces were all identifiable, so I found them after they came down and asked them if I could share it and told them who I was on IG. They never followed back up with me so I can't release it, sometimes that's just the way it goes.

 

[Citysnaps]

They say you gotta have those model releases handy but I never do. The other day I filmed a family on a bluff from a distance with my drone and their faces were all identifiable, so I found them after they came down and asked them if I could share it and told them who I was on IG. They never followed back up with me so I can't release it, sometimes that's just the way it goes.

Back when I started hitting up people for conversation and portraits (around 15-20 years ago), I was very mindful of the four privacy torts. False light publicity, disclosure of private facts, intrusion into seclusion, and appropriation of likeness/name for commercial purposes.

Still, I felt the need to have an expert weigh in, especially on edge cases. So over that period of time I engaged two different media rights attorneys in SF. They both felt I was on safe ground, me being aware of the four privacy torts mentioned above, though one of them needed to consult one of his law books for a specific purpose I had in mind.

In consideration of all of the above, I always let my subjects know what I was doing and its purpose; documenting the city (San Francisco). No one ever complained or took issue with that.