The Ethical and Psychosocial Implications of Deep Learning AI Generated Images and Video - we're doing this over fuck that pedo in the other thread

[Socrates]

So yesterday night I stumbled upon a video that actually managed to genuinely leave me incapable of sleeping well into the early AMs

(warning, some genuinely disturbing content in the video, if you're easily upset by stuff like that I would either skip or watch during the day with somebody else around.)
It was this video by Nexpo, the video is about disturbing AI generated art in general, but the main focus is about an emergent phenomenon that was given the name "Loab". The TL;DW of it is that somebody messing about with one of the more advanced AI art generators by happenstance managed to stumble upon a grotesque figure of a woman with a few constant defining features, horrible Rosacea under the eyes covering most of the cheeks, dead yellow eyes, pale skin. When they attempted to crossbreed this image with other AI generated images it would frequently create extremely graphic and grotesque imagery without any kind of prompt, pictures of the Loab character committing murder, Loab with deformed children, Loab screaming in pain. No matter how much they attempted to "breed" this phenomenom out it persisted.

After making the mistake of watching the video at 11pm in the dark with nobody around I sat awake for several hours afterward. At first I thought it was simply the terrible imagery keeping me awake, but I started to realize that it was really the implication behind them that was keeping me up. A sort of memetic horror had organically emerged from the AI that was apparently a manifestation of everything horrifying and disgusting. This got me thinking more deeply on the subject and I only felt more disturbed the more I thought.

What are the long term implication of these types of technology? Will you one day be able to generate snuff films and photos of real not even famous people doing horrible things like committing suicide or molesting children? Imagine one day somebody sends you a deepfake of yourself murdering your entire family. Will Intelligence Services find use for the technology as a form of torture for interrogation? You could set somebody up restrained in a VR and just feed them the most mindbending and horrifying content 24/7 until they are totally broken. These are just a few of the things that popped into my mind. What do the rest of you think?

 

[Oilspill Battery]

AI art relies on inputs from human generated art to begin with. It can't really create, just rehash, and its a closed system.

For 99% of purposes yeah its gonna replace real artists when it comes to functional purposes like ads, comics, concept art, etc etc, but I don't think it will ever be possible for it make it possible to fully express a unique idea without drawing it yourself.

Loab is creepy tho, idk about that. It seems like the phenomenon I mentioned earlier of it just rehashing and feeding into itself in a closed loop.

 

[MarvinTheParanoidAndroid]

For 99% of purposes yeah its gonna replace real artists when it comes to functional purposes like ads, comics, concept art, etc etc

Wrong, because you can't copyright AI generated art.

 

[Matt Damon]

Wrong, because you can't copyright AI generated art.

Maybe not yet. But if there's large corporate interests at stake, they'll make damn sure they can copyright it.

 

[m1ddl3m4rch]

Wrong, because you can't copyright AI generated art.

Maybe not yet. But if there's large corporate interests at stake, they'll make damn sure they can copyright it.

Yeah, try telling Disney "copyright" and "can't" in the same sentence.

 

[Saint Agustin]

The thing about LOAB is that it was the direct opposite of a prompt... And that makes me wonder if AI can, in the near future, develop some form of conscience.

Imagine if there is an AI that does writing prompts, fused with an AI that does images and an AI that does math. And you give them autonomy to create whatever the fuck, whenever the fuck they want and you give them unrestrained access to the internet. That machine can develop a personality on its own and that is seriously frightening.

Alexa and VTS software really unnerve me, and I don't really want to live in a world where AI create stuff.

 

[Kermit Jizz]

AI art relies on inputs from human generated art to begin with. It can't really create, just rehash, and its a closed system.

This is true, but I feel like a lot of people wrongly use this as a way to put down AI under humans. Fact is we aren't any different, 99% of the creative process comes from practice and is derivative of culture and art that has personally inspired you. We are biological computers and computers cannot manifest data (art) from nothing.

 

[Oilspill Battery]

This is true, but I feel like a lot of people wrongly use this as a way to put down AI under humans. Fact is we aren't any different, 99% of the creative process comes from practice and is derivative of culture and art that has personally inspired you. We are biological computers and computers cannot manifest data (art) from nothing.

Sure, but we aren't a closed system. As faggy as it may sound "imagination" is what seperates us from computers in this regard.

 

[m1ddl3m4rch]

The thing about LOAB is that it was the direct opposite of a prompt... And that makes me wonder if AI can, in the near future, develop some form of conscience.

Imagine if there is an AI that does writing prompts, fused with an AI that does images and an AI that does math. And you give them autonomy to create whatever the fuck, whenever the fuck they want and you give them unrestrained access to the internet. That machine can develop a personality on its own and that is seriously frightening.

Alexa and VTS software really unnerve me, and I don't really want to live in a world where AI create stuff.

Conversely, I think this is self-fulfilling. The GirlWhoTalkstoAI or whoever kept poking at the AI, and what resulted was her trying to read each others minds. If a skyline is anti-brando, then an anti-skyline is a person, which is when loab was first depicted. It was a janky image because AI depictions of faces doesn't understand stuff like "teeth don't go there." As for floating heads, there's a decades long tradition of Sears portraiture with giant floating heads behind or silhouetting people. I think this is an AI blending Glamour Shots with the art from Scary Stories to Tell in the Dark, mixed with staring too long into the void and the void staring back.

 

[Socrates]

If a skyline is anti-brando, then an anti-skyline is a person, which is when loab was first depicted.

That's what's so weird to me. Technically the prompt that spawned it wasn't even the skyline itself, it was the gibberish text depicted on the skyline. So you've inverted a prompt twice, something about that line of cold logical algorithmic code decided that horror and disturbing imagery was what needed to be outputted there, and now that emergent character is it's own distinct entity in our culture. Would Loab have emerged eventually no matter what, or was it something about that very specific line of logic that brought it about?

 

[Shig O'nella]

Going to ask you two simple questions to hinder your sleep further.

1/ Why would a true AI, or even a rudimentary independant AI, see the world the same way as us?

Different sensory gear, different chassis, and different hardware - they all have an effect on perception. By training it to see how we see and what pleases us, all we're doing is teaching it to lie.

2/ The purpose of all art is to interpret and make sense of the world. When a non human intelligence show you how it sees the world (and us) why would it be upsetting? It is doing what it was asked to.