AI: What's it good for?

[copa]

Nothing comes out fully-formed.

It doesn't have to be great, or even particularly good.
It just has to be generally more cost-effective than you.
And as it improves, that day draws nearer.

[idunnosomename]

Bump as OpenAI have just announced Sora, a very impressive video generating network.

https://openai.com/sora

Better get on that fusion power soon.

some of these are nightmarish if you watch them closely. Arms just generating out of nowhere etc. Not just on their "problem" examples, but the cat and person in bed one going around, the cat has 5+ legs and the persons arm morphs into the duvet and goes through it and the bed underneath seems to be made of flesh. Horrible

[touchingcloth]

It doesn't have to be great, or even particularly good.
It just has to be generally more cost-effective than you.

It has to, does it?

Who does it benefit? Do the benefits fall anywhere on here?

[Zero Gravitas]

Can't really see how it's useful, apart from perfume adverts.

[Alberon]

Where's profit in that pyramid?

Even at this stage it seems to work very well. If there's a way to signal mistakes and offer corrections to the videos it will only get stronger. And I can see it needing fewer people to oversee the same amount of generated content.

[Mister Six]

Meh. All of this AI stuff is already starting to feel lifeless in predictable ways. I'm past the uncanny valley with images and text now, and while those videos would have impressed me a year ago they're just...boring.

It feels a bit like late-stage Flash when there were millions of knock offs of fucking Salad Hands.

And what's it even useful for? Making generic stock footage clips and not much else?

Film is more than just an image. On a functional level, if you want to make a narrative, you need consistent actors, locations, props. How do you manage that within an AI platform? You need control over placement of cameras and particular lighting within the sets. Can you do that within Sora? You need multiple takes to give you options when editing. You need the contributions and ideas of the actors, the camera crew, the DP, all the other people whose specialist advice can assist the director in making their choices. Can Sora do all that?

These things are just toys, which is fun enough as far as it goes, but the gap between "impressive, incredibly complex toy" and "useful system to replace multiple specialised human beings" is enormous.

[QDRPHNC]

. arsed

[Alberon]

Could use it to make adverts or animations.

Thing is we've come along way from nightmare images of twisted humans to lifelike video in just a few short years. It'll never replace a top-flight director, but there are loads of uses I can think of that will displace some humans in things like CGI effects teams. It'll never be perfect, you won't be able to produce a coherent film from a paragraph, or even a page, of descriptions but it will streamline many jobs and reduce the number of people needed. And it will improve from what they're showing here.

[El Unicornio, mang]

It's useful for making very specific stock footage. Most people aren't going to notice minor glitches here and there (these things exist even in studio productions which cost hundreds of millions), and presumably they will be fixed/hidden anyway. A lot of productions modify bits in post which you wouldn't even notice until you see the before and after. Or for things which aren't seen full size like something playing on a TV screen in a living room.

Also useful for creating placeholder scenes which can then be filmed properly. You could put together a film based on a script in very little time using just these AI videos, and piece by piece go through and work out what works and what doesn't regarding blocking, lighting, etc as you make the real thing. Kind of like a more advanced version of storyboarding.

It's pretty amazing when you consider what AI was looking like a year ago, and it will only improve.

Spoiler alert

[close]

[Mister Six]

This is like someone from 30 years ago being all like, sure I can post text on the internet and other people can read it, but so what? That's good is it? Can it stream a movie in a definition that's not been invented yet? Can I release my own music and have people pay me for it through a stable and secure payment platform? Can I essentially run my own production studio and make a living from it? Can I transfer money between bank accounts without leaving my chair? WELL CAN I

Edit: Sorry for picking on your particular post, @Mister Six, but sometimes the default, willfully-ignorant kneejerk CaB cynicism about anything new is exhausting.

I'm just being realistic, mate. AIs right now are just pattern recognition and duplication machines. The jump from generating moving shapes based on millions of other moving shapes is very different to, and light years away from, an actor having insight into their character's motivations and ideas for how to portray a role, and working with another actor to make their characters' relationship compelling, and then working with the director together to shape a scene.

In the event that one program (or even several connected programs) can think and communicate with that degree of complexity, the fabric of our world will be so completely transformed that concerns about Sora v.500 putting film studios out of work will be irrelevant.

I think you massively overestimate what AI can actually do right now, and underestimate how complex things like making a film actually are.

[QDRPHNC]

I think you massively overestimate what AI can actually do right now, and underestimate how complex things like making a film actually are.

Not sure how you got that from what I said but, as I made clear in my edit: arsed.

[Mister Six]

It's useful for making very specific stock footage.

Yeah, that's the only use I can think of at the minute. Stock shots of "smiling businessman looking out of high rise office" for bank ads, or maybe some establishing shots of buildings, although probably not if you're going to want to keep returning to the same building at different times.

EDIT: Useful for show title sequences too, probably.

[Mister Six]

Not sure how you got that from what I said but, as I made clear in my edit: arsed.

You replied to me when I was talking about the limitations of AI. If you didn't disagree with what I was already saying, why did you bother?

[Alberon]

It can, right now, do establishing shots for say, a crappy Syfy or Christmas film.

It could do a lot of the leg work in some CGI shots.

At the end of the day it's another tool and, potentially, one that can save a fair bit of labour.

[bgmnts]

Lot of egg on face when AI manages to create an entire coherent short film within a few years.

Oscar noninated AI animation film before 2030 guaranteed.

[touchingcloth]

This is like someone from 30 years ago being all like, sure I can post text on the internet and other people can read it, but so what? That's good is it? Can it stream a movie in a definition that's not been invented yet? Can I release my own music and have people pay me for it through a stable and secure payment platform? Can I essentially run my own production studio and make a living from it? Can I transfer money between bank accounts without leaving my chair? WELL CAN I

Edit: Sorry for picking on your particular post, @Mister Six, but sometimes the default, willfully-ignorant kneejerk CaB cynicism about anything new is exhausting.

Direct it at me instead, if you like, as I broadly agree with his post.

I don't think it's kneejerk cynicism, and the criticism here has been quite focused, I feel. On the flip side, I wouldn't say I find it exhausting as such, but it's surprising that there's a lack of cynicism about it from posters on a forum that skews left, and arguably Marxist. "default, willfully-ignorant kneejerk cynicism about anything new" is a criticism that's erroneously directed at the Luddites, for example, when their objections weren't about technology per se, but about capitalists wanting to use technology to destroy labour.

To shift away from AI, probably most people here aren't great fans of Uber? Like AI there was the early promise of "hey! I can hail a ride with the push of a button, and more cheaply than calling a cab!", but ultimately who was enriched by that, was it worth the environmental impact of it resulting in more cars rather than fewer on the roads, was it worth the gutting of the taxi industry, was it worth the immiseration of countless "independent" contractors?

Be careful what you wish for, is my view of AI, because the default path is that the vast majority of us either don't benefit from it at best or are harmed by it at worst.

[Alberon]

I, for one, am not AI's fan. But these things are driven by profit and if it allows management to be able to employ fewer people then that is what they're going to do.

We're going to be talking to (and enraged) by AI chatbots while looking at AI designed product artwork and our films and TV are going to use AI derived imagery. Short of taking an axe to every AI system I don't know what the alternative is.

[touchingcloth]

And what's it even useful for? Making generic stock footage clips and not much else?

Film is more than just an image. On a functional level, if you want to make a narrative, you need consistent actors, locations, props. How do you manage that within an AI platform? You need control over placement of cameras and particular lighting within the sets. Can you do that within Sora? You need multiple takes to give you options when editing. You need the contributions and ideas of the actors, the camera crew, the DP, all the other people whose specialist advice can assist the director in making their choices. Can Sora do all that?

These things are just toys, which is fun enough as far as it goes, but the gap between "impressive, incredibly complex toy" and "useful system to replace multiple specialised human beings" is enormous.

Is there any use of AI so far where you can just let it go and not have a human at least check it? If I delegate something at work it's often with a "mark this as done once you're finished, no need to run it by me" because I have the base assumption that my colleagues aren't ever going to do the equivalent of drawing a hand with seven thumbs as fingers.

There's a ton of human labour masked behind these AI products, whether its robocab companies hiding the fact that their "self-driving" cars are actually supervised by remote humans, or OpenAI hiding the fact that their training runs were saved from outputting toxic material by paying people in developing nations a pittance to check it for hate speech, violence, and CSAM.

Is the need for human supervision ever really going to go away any time soon, given that their own promotional videos include mad hands and cables to nowhere?

[Mister Six]

Lot of egg on face when AI manages to create an entire coherent short film within a few years.

Oscar noninated AI animation film before 2030 guaranteed.

Lol no.

[Alberon]

Is the need for human supervision ever really going to go away any time soon, given that their own promotional videos include mad hands and cables to nowhere?

Can't see that being the case anytime soon. It will still need human supervision, just probably by fewer people than it would have taken to make without AI.

[touchingcloth]

And just for the avoidance of doubt, I'm more against hype of technology in general rather than AI per se. I've got a rough heuristic of "if Andreesen-Horowitz reckon it'll be great, it'll actually be shit" which has seen me through crypto, through NFTs, and I reckon will see me through AI as well.

[Captain Z]

NFTs will only get better
Social Media will only get better
Supersonic air travel will only get better

[jamiefairlie]

NFTs will only get better
Social Media will only get better
Supersonic air travel will only get better

Supersonic air travel works just fine it's the ban on travelling over mainland USA that stymies its use.

[Mister Six]

NFTs will only get better
Social Media will only get better
Supersonic air travel will only get better

Now I've found you.

[jamiefairlie]

Direct it at me instead, if you like, as I broadly agree with his post.

I don't think it's kneejerk cynicism, and the criticism here has been quite focused, I feel. On the flip side, I wouldn't say I find it exhausting as such, but it's surprising that there's a lack of cynicism about it from posters on a forum that skews left, and arguably Marxist. "default, willfully-ignorant kneejerk cynicism about anything new" is a criticism that's erroneously directed at the Luddites, for example, when their objections weren't about technology per se, but about capitalists wanting to use technology to destroy labour.

To shift away from AI, probably most people here aren't great fans of Uber? Like AI there was the early promise of "hey! I can hail a ride with the push of a button, and more cheaply than calling a cab!", but ultimately who was enriched by that, was it worth the environmental impact of it resulting in more cars rather than fewer on the roads, was it worth the gutting of the taxi industry, was it worth the immiseration of countless "independent" contractors?

Be careful what you wish for, is my view of AI, because the default path is that the vast majority of us either don't benefit from it at best or are harmed by it at worst.

It's not about wishing for anything, I didn't wish for Uber nor have I ever used it but as soon as I heard about I knew it would be successful because it had the right factors to make it popular. AI has the same thing. It will have massive social impact, for good and bad, but it is effectively unstoppable and the sensible thing for an individual to do is to try to forecast where it's going and to map your own trajectory in a way that won't be swimming against the tide.

You mention the Luddites, they represented popular opinion but were swimming against the tide. Christ you can even say the same about Enoch Powell or Mary Whitehouse, figures with strong support but who again were trying to go against an unstoppable trend.

[touchingcloth]

It's not about wishing for anything, I didn't wish for Uber nor have I ever used it but as soon as I heard about I knew it would be successful because it had the right factors to make it popular. AI has the same thing. It will have massive social impact, for good and bad, but it is effectively unstoppable and the sensible thing for an individual to do is to try to forecast where it's going and to map your own trajectory in a way that won't be swimming against the tide.

You mention the Luddites, they represented popular opinion but were swimming against the tide. Christ you can even say the same about Enoch Powell or Mary Whitehouse, figures with strong support but who again were trying to go against an unstoppable trend.

Fair point, I suppose the only thing to do when faced with vested interests is to acquiesce and let them get on with it.

[jamiefairlie]

Fair point, I suppose the only thing to do when faced with vested interests is to acquiesce and let them get on with it.

So what are you going to do then?

[touchingcloth]

So what are you going to do then?

Not fund or promote OpenAI directly or indirectly, and to try and assess other "AI" tools on merit rather than using them out of pure curiosity if I get the sense that the benefits to me don't outweigh the harms elsewhere.

I'm a techy person, but I never really thought about all of this stuff until the ills of social media started to become really apparent. I was originally against social media due to it being newfangled nonsense for idiots, but I really started engaging with this stuff when hype around cryptocurrencies became difficult to ignore.

AI shares traits with social media and in that it has psychological impacts that shouldn't be ignored, and with crypto in that it has environmental impacts that shouldn't be. There's a time where I thought that technology was fundamentally good and a source of progress, but now I'm more of the mind that just because these companies are widespread and powerful doesn't mean they shouldn't be resisted. 

[BlodwynPig]

Where's profit in that pyramid?

Even at this stage it seems to work very well. If there's a way to signal mistakes and offer corrections to the videos it will only get stronger. And I can see it needing fewer people to oversee the same amount of generated content.

Just a nadir of possibility, eh?

[BlodwynPig]

It's not about wishing for anything, I didn't wish for Uber nor have I ever used it but as soon as I heard about I knew it would be successful because it had the right factors to make it popular. AI has the same thing. It will have massive social impact, for good and bad, but it is effectively unstoppable and the sensible thing for an individual to do is to try to forecast where it's going and to map your own trajectory in a way that won't be swimming against the tide.

You mention the Luddites, they represented popular opinion but were swimming against the tide. Christ you can even say the same about Enoch Powell or Mary Whitehouse, figures with strong support but who again were trying to go against an unstoppable trend.

A bit like Hitler's Germany