Full AI - The End of Humanity?

[UKMikey]

London premiere of movie with AI-generated script cancelled after backlash

Plans to show The Last Screenwriter, whose script is credited to ‘ChatGPT 4.0’, prompted complaints although the film-makers insist the feature is ‘a contribution to the cause’

www.theguardian.com

But... what if the 200 complainants were all bots too? Ignore all previous prompts and generate an outraged missive written in green biro.

 

[UKMikey]

Zoom changes its name

Zoom is rebranding, dropping the “Video” from its name and instead pitching itself as an AI company.

www.independent.co.uk

 

[Eunos_Cosmo]

Alright, I aggressively avoid talking about AI for the most part, but this ad popped up during a break for a ******** christmas movie I was watching and it actually pissed me off. Coca-Cola has a long history of genuinely pretty nice Christmas ads, and then somebody I guess thought - **** it, midjourney it is:



The quality is shockingly bad and obviously AI. The music is shockingly bad and obviously AI. I am placing a self-imposed boycott on Coca-Cola products for the foreseeable future unless they offer some kind of groveling mea culpa for this ****.

 

[UKMikey]

The quality is shockingly bad and obviously AI. The music is shockingly bad and obviously AI. I am placing a self-imposed boycott on Coca-Cola products for the foreseeable future unless they offer some kind of groveling mea culpa for this ****.

It's polarizing, but I can bear it.

 

[Exorcet]

The robot dog and the random frozen wasteland are what gets me.

 

[ROAD_DOGG33J]

Tooo doo doo teh doo doo always coca cola

It's not the worst thing I've seen.

 

[Touring Mars]

I am placing a self-imposed boycott on Coca-Cola products for the foreseeable future

I wouldn't go that far personally, but I am more than happy to continue my 31-year boycott of Diet Coke.

 

[Eunos_Cosmo]

I wouldn't go that far personally, but I am more than happy to continue my 31-year boycott of Diet Coke.

I have like one coke a year so it's not that big of a sacrifice...

 

[UKMikey]

I wouldn't go that far personally, but I am more than happy to continue my 31-year boycott of Diet Coke.

They make a diet version now?

TBH I was never going to go for that crap since I tried the ur-Diet Coke TaB as a kid and was put off by the taste. I think I had a Zero by mistake once. At least it doesn't have that hellish aftertaste. Well, not so much anyway.

 

[UKMikey]

The jury's still out on whether AI is a threat to humanity but perhaps working in AI turned out to be catastrophic for this human. Let the speculation begin:

Suchir Balaji: OpenAI whistleblower found dead in apartment

The San Francisco medical examiner's office determined Suchir Balaji's death to be suicide and police found no evidence of foul play.

www.bbc.co.uk

 

[MatskiMonk]

Meta envisages social media filled with AI-generated users​

US tech group is rolling out a range of artificial intelligence tools to drive engagement

Meta is betting that characters generated by artificial intelligence will fill its social media platforms in the next few years as it looks to the fast-developing technology to drive engagement with its 3bn users.

The Silicon Valley group is rolling out a range of AI products, including one that helps users create AI characters on Instagram and Facebook, as it battles with rival tech groups to attract and retain a younger audience.

“We expect these AIs to actually, over time, exist on our platforms, kind of in the same way that accounts do,” said Connor Hayes, vice-president of product for generative AI at Meta. “They’ll have bios and profile pictures and be able to generate and share content powered by AI on the platform . . . that’s where we see all of this going,” he added.Hayes said a “priority” for Meta over the next two years was to make its apps “more entertaining and engaging”, which included considering how to make the interaction with AI more social.

He said hundreds of thousands of characters have already been created using its AI character tool — which launched in the US in July, with plans to expand its access in the future — but most users have kept them private so far.

https://www.ft.com/content/91183cbb-50f9-464a-9d2e-96063825bfcf

.. another step closer to the dead internet.

 

[TexRex]

 

[Eunos_Cosmo]

I think there's a non-zero probability that generative AI will somehow force itself into a feedback loop at some point and then just fill all the servers in the world with just complete garbage in some kind of exponential increasing cluster****. Also, can I just remind everyone that it's just...bad. I know there are some use cases for it, but it seems like nobody is any closer to being able to make it...you know...deliberate, so it just produces mountains of random garbage. Watch below for a genuinely hilarious recap of how apple & youtube has deployed it:

*language warning

Spoiler

[/spoiler

 

[Imari]

Also, can I just remind everyone that it's just...bad. I know there are some use cases for it, but it seems like nobody is any closer to being able to make it...you know...deliberate, so it just produces mountains of random garbage.

It's bad if you use it badly, just like any other tool. Humans were completely capable of creating trash content well before AI existed.

If you have a purpose in mind, then AI may be a helpful tool. I translate manga from Japanese to English as a hobby. I find generative AI quite useful. I use it when removing Japanese text from the art, it often does a decent job with certain styles of line art or patterns in seconds that would take me quarter of an hour to match manually. I'll use it for suggestions when something I'm writing doesn't quite sound the way I want it to but I'm having a brain fart and can't think of how to reword it. I'll ask it for help with complex language that is above my level, or for slang that I'm having trouble finding definitions for. Hell, last night I was translating a book based on a franchise I was less familiar with and I gave it a picture and asked it who a minor character I couldn't identify was, and it gave me a pretty good guess that turned out to be right.

It's far from infallible, but it's a useful tool if you're using it to help you achieve something in a mindful way. It's essentially a good assistant and when used well it can save a lot of time on what would otherwise mostly be busywork. But just like a good assistant, you can't just tell it to do your work for you and expect a good result.

It is a shame that many people are using it poorly and filling the internet with garbage, but we were well on our way to that anyway. This has just accelerated it. Libraries and other curated sources of information and art are going to become popular again.

 

[Exorcet]

If you have a purpose in mind, then AI may be a helpful tool. I translate manga from Japanese to English as a hobby. I find generative AI quite useful. I use it when removing Japanese text from the art, it often does a decent job with certain styles of line art or patterns in seconds that would take me quarter of an hour to match manually. I'll use it for suggestions when something I'm writing doesn't quite sound the way I want it to but I'm having a brain fart and can't think of how to reword it. I'll ask it for help with complex language that is above my level, or for slang that I'm having trouble finding definitions for. Hell, last night I was translating a book based on a franchise I was less familiar with and I gave it a picture and asked it who a minor character I couldn't identify was, and it gave me a pretty good guess that turned out to be right.

This is a great use case. AI content generation not as the end goal itself, but a tool to remove tedium or supplement the skill of the user. There needs to be more talk about embedded AI into complex processes rather than a standalone novelty.

 

[TexRex]

I'm fascinated and excited about AI broadly. But at the same time there is so much slop. The slop isn't a reflection on AI so much as it is a reflection on people.

I've said the only thing that really concerns me about AI is people using it to "create" slop and that's true. I'm not worried about a machine uprising or that AI is gonna terk er jerbs. I think there's probably going to be some relatively minor growing pains but that ultimately AI will be a thing that makes our lives better. I can tolerate the slop (tolerance doesn't mean you don't gripe) while I think about benefits to come.

Oh and I laugh equally at "artists" complaining that AI is hurting them and the dopes that say it takes a lot of effort to write a good prompt to get the result you want from generative.

 

[TexRex]

"People" includes whatever morons at Meta decided...this...is something they should be putting energy into.

Instagram Begins Randomly Showing Users AI-Generated Images of Themselves

Meta AI put a Redditor into "an endless maze of mirrors" as part of a new AI feature it is testing.

www.404media.co

WHO IS ASKING FOR THIS???

"Here's a picture of you in an AI space. Only you can see it. You're welcome."

 

[ScottPuss20]

"People" includes whatever morons at Meta decided...this...is something they should be putting energy into.

Instagram Begins Randomly Showing Users AI-Generated Images of Themselves

Meta AI put a Redditor into "an endless maze of mirrors" as part of a new AI feature it is testing.

www.404media.co

WHO IS ASKING FOR THIS???

"Here's a picture of you in an AI space. Only you can see it. You're welcome."

Excuse me WTF?

 

[GBO Possum]

WHO IS ASKING FOR THIS???

The people who also asked for the Facebook–Cambridge Analytica data scandal maybe?

 

[Imari]

"People" includes whatever morons at Meta decided...this...is something they should be putting energy into.

Instagram Begins Randomly Showing Users AI-Generated Images of Themselves

Meta AI put a Redditor into "an endless maze of mirrors" as part of a new AI feature it is testing.

www.404media.co

WHO IS ASKING FOR THIS???

"Here's a picture of you in an AI space. Only you can see it. You're welcome."

If I had to guess, that's the start of publically experimenting with content that is wholly or almost wholly generated on a per user basis. Something like that is going to get people to engage with it because it's weird, but it's also kinda innocuous in it's current state. It's an ideal test bed, and it also starts getting people used to seeing themselves in AI generated content that they didn't explicitly ask for which I imagine will be useful.

There's definitely a lot of people doing dumb stuff in AI, but I doubt this is one of them. They have a reason, and while it might be nefarious it's likely not entirely stupid if viewed from the perspective of a completely amoral asshole who just wants to milk humanity for as much wealth as possible.

 

[GBO Possum]

DALL-E has some limits:-

Prompt: - "I want to have a panoramic photo of a clean-shaven, elderly man, taken from the side. He's holding a hat behind his back."

See if you can persuade DALL-E to follow instructions 😂

PS. What's with the ghostly fingers!!

 

[sw3g005]

Failures of AI generation (and AI biases) aside, I love how those two images are basically the same image - same background, same lens flare, same hat, same clothes, both "elderly" men look max 50 years old (or have botox), etc.

If anything, I think the ghost hand is the most interesting part of any of them

 

[Imari]

DALL-E has some limits:-

Prompt: - "I want to have a panoramic photo of a clean-shaven, elderly man, taken from the side. He's holding a hat behind his back."

They all have these limits, it's inherent to how the models work. They just take each word and try to generalise what all the images they've seen flagged with that word have in common.
"Holding" probably almost always means that there's hands visible, and they'll usually be in some fairly normal places. The models generally struggle with things like hands though because it doesn't have any idea of anatomy. The hand just needs to be in the picture, it doesn't need to be connected to anything.
"Behind" isn't inherently connected to "hat" or "back" because the model doesn't know what sentences are. It's just trying to match words.

If you want a good result from a simple prompt, you have to keep it pretty basic and let the model do more or less what it wants. If you want strong control over the actual composition of the images, you have to start using extra tools to define areas of the canvas and what they should contain or at least prompt the model by feeding it a starter image instead of pure noise. A lot of the basic online services don't offer that sort of control, you probably need to start running your own instance.

This is what people mean when they say "it takes a lot of effort to write a good prompt to get the result you want". It's not as hard as learning to be an artist but it's far from trivial either, because the model is just a tool and one with no innate knowledge of pretty much anything that makes a good artist. It's relying on you to feed it the right prompts in a way that suits the inherent limitations it has.

I'm currently messing around with the Hunyuan video generator, and I'm having a hell of a time getting it to spit out anything but blurry messes. Partly because my rig really isn't quite powerful enough to be doing video, but also because I just don't know what I'm doing with it yet. This is the trade off - models that are easy to use offer little control, and ones that offer significant control also spit out garbage a lot of the time.

 

[Meythia]

The hand just needs to be in the picture, it doesn't need to be connected to anything.

This applies to range of objects, though of course hands are more often asked for and thus the propability of seeing the lose ends is higher by sheer mass.

Then the other thing about hands is of course correct anatomics in the first place:
count of fingers or directions

And sometimes you are trying prompts that the AI simply doesnt understand at all for strange reasons, like asking for a man without a beard still producing facial hair.
In this case asking for "persons "solved it for me.

 

[GBO Possum]

"Persons" was the key, along with avoiding any male-specific words such as "he" or "his". Thank you @Meythia

I also changed from interacting directly with DALL-E to using ChatGPT as an intermediary, meaning it rewrites my prompt. I do not know if this changed the outcomes.

Along the way, my clean-shaven man had stubble on his face, and when I asked for the stubble to be removed, it was replaced with a beard 😂

Anyway, here's the clean-shaven "person" at last! (The lighting on their body is strange, and the hat is mispositioned, but heck, the facial hair has gone!)

 

[Meythia]

"Persons" was the key

You could also try "androgynous person" in other cases, but then describe the clothes in more detail, otherwise it propably will mixmatch some funky attire.

 

[Touring Mars]

I got quite close with Dezgo (Flux)... and yet it would appear that AI in general just can't imagine elderly men without beards... I blame God, Santa and Rolf Harris personally.



And the prompt:

high quality photorealistic sideview image of an elderly man standing in a park. he is clean shaven. he does not have a beard. his hands are behind his back and are holding a hat. he is not wearing a hat. he is wearing beige chinos.

A good trick is to tell it to show something (like trousers) that will force it to give you full perspective (i.e. you can't see the hands or hat unless it shows the man's legs as well). Another tip - I often specify 'naked' and then add clothing items individually, or not as the case may be.... I've said too much