Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

The AI Unbundling

stratechery.com

Post reply on HN
Print view View original

51–60 of 207 posts

Re: The AI Unbundling

#51

Post by yojo »

megaman821 wrote: post #6

I am not sure I see the unbundling here. AI generated art may be a threat to generic stock photography and small jobs for custom blog art, but artists higher up on the food chain now have a new tool for them to use. They can quickly generate a bunch of concepts to tryout ideas, and polish those into a higher quality piece of art than if left to their own devices.

If it’s a 10x force multiplier, then one person could do the work of 10. There are two possible outcomes:

1) the lower rate per piece enabled by higher productivity unlocks a much larger market for commissioned work. No one has to lose their job, people make up the price difference in volume.

2) there is not a much larger market waiting for cheaper art. The existing pool of artists is too big, and some of them need to find new work.

My guess is lowering the price does grow the market some, but not enough to avoid displacing at least some of the artists. Whether that’s 10% or 99% depends on how good these tools ultimately end up being.

Re: The AI Unbundling

#52

Post by OkayPhysicist »

bambax wrote: post #24

Good article, although "unbundling" is an odd choice of word. Before AI, the execution step was necessary, but not necessarily attached to the "coming up with an idea" step. Conceptual artists have for years (centuries?) produced ideas that were executed by others. AI makes the execution step near-instantaneous and almost free. It kind of erases it. It's more disintermediation than unbundling. Execution artists are b…

  does that mean that all possible images are somehow contained in the model?
That would depend largely on what kinds of non-linearities the model is using, particularly towards the later steps. It's entirely possible for there to exist spaces that a given trained model cannot output. It's unlikely those spaces are particularly interesting: perhaps a pixel-by-pixel checkerboard of pure black and white, for example.

A perfectly linear model with linearly-independent matrix columns could generate any possible value, but would be exactly equivalent to a single vector-matrix multiplication, unable to do any interesting multi-step reasoning.

Re: The AI Unbundling

#53

Post by enord »

gizmo wrote: post #33

Although we talk a lot about disruption, only very few technologies are truly disruptive. You can tell by the panic and awe in the air whether you're dealing with real disruption or incremental change. Dropbox made filesharing easier. It's a good product, but not disruptive. Nobody panicked that Dropbox would make their job redundant. Uber was hard on the taxi industry, but fundamentally you still have drivers taking…

You quite studiously compared every ostensible disruption to the preceding state of the art to (quite convincingly IMO) reveal the material consequeces to be… less than world shattering, shall we say. Your example of Amazon and walmart does the converse.

And yet, with «stable diffusion» it is just so. Disruption is declared wholesale, without even passing mention of the preceding state of the art of computer generated art.

Not to put too fine a point on it, but do you not see the own goal here?

Re: The AI Unbundling

#54

Post by yunwal »

egypturnash wrote: post #27

Welcome to the world where every blog has a vaguely painterly smear of a header image that falls to pieces when you actually look at it. Kinda tacky, really.

Yeah, I really hope people stop just letting AI generate images with no human direction beyond the text prompt. It looks terrible 99% of the time, and Midjourney especially is so same-y.

There are good uses for human guided AI, but I fear most people are down for accepting whatever the algorithm happens to crap out at you

Re: The AI Unbundling

#55

Post by lacker »

time_to_smile wrote: post #22
andreilys wrote: post #19

Earlier quoted context omitted.

I’m curious why do you think that AI generation will plateau at its current quality level?

Same reason that self driving cars have plateaued. The things that are really important for that last mile are also really hard and we're not really sure how to get there. The output between all the major AI image synthesis tools, while impressive, is quite clearly very high dimensional interpolation, and it shares the same limitations in that space as it does the 2d space. We've already thrown insane amounts of data…

Same reason that self driving cars have plateaued.

But they haven't plateaued. I can go take a trip in a self-driving Cruise in San Francisco right now. I couldn't do that last year.

There's certainly a lot of room for them to improve, but the self-driving car industry is making progress.

Re: The AI Unbundling

#56

Post by lacker »

hertzrat wrote: post #5

Isn’t it likely that patent law is going to lock ai down? One company will finally crack something general-enough in a way that gets it way ahead, will patent it, and will use that monopoly to gain control of the whole space, and the monopoly will never go away because all further improvements will also get patented by the ai legal team? The brief openness of ai right now is a glitch in the system that will get irone…

It's pretty rare that patent law makes a big difference in slowing down a field of software innovation. The main examples I can think of are eFax suing fax machine software companies, and Blackboard suing education companies.

In particular the large tech companies simply ignore all existing patents, develop what they want, and defend themselves in court if need be. Google is never going to say "well someone else got a key patent, let's give up development in the AI space".

Re: The AI Unbundling

#57

Post by OkayPhysicist »

hertzrat wrote: post #5

Isn’t it likely that patent law is going to lock ai down? One company will finally crack something general-enough in a way that gets it way ahead, will patent it, and will use that monopoly to gain control of the whole space, and the monopoly will never go away because all further improvements will also get patented by the ai legal team? The brief openness of ai right now is a glitch in the system that will get irone…

It is heavily unlikely that there is exactly one route to intelligence: to take a card from biology, look at octopuses versus humans. Since only relatively specific architectures can be patented, I expect that one would have no issue circumventing any patent that could claimed.

Re: The AI Unbundling

#58

Post by ryoshu »

pyaamb wrote: post #43
cleandreams wrote: post #41

I think Stable Diffusion is a both a threat and a tool for illustrators but will mostly be a tool for artists. The reason is that AI intelligence is probability generated mimicry. By extension it mimics art but is not art. Think of the character arcs in Shakespeare, the narratives, the layers of meanings. This sort of thing cannot even be modeled as a problem in AI. The ability to define what is necessary to produce…

Can Dall-E be extended to generate 3d models?

https://www.kaedim3d.com/ is working on 2D -> 3D.

Re: The AI Unbundling

#59

Post by mikepurvis »

bioemerl wrote: post #16
mikepurvis wrote: post #10

Earlier quoted context omitted.

I agree. I see AI-generated art as a force multiplier for skilled individuals, same as better editors, tools, languages, and frameworks have been for developers.

I think this will be many things. A force multiplier, allowing existing professionals to generate more art. The creation of a new profession. Digital artists are already doing a very different job with very different tools. These tools will be another step and there will be demand for human and generated art in different applications. The empowerment of non artists. Same as how digital art democratized art. This will…

Devalue, yes. Put people out of work? That's less obvious to me. Imagine:

- Instead of video advertising with a handful of regional variations, hundreds or thousands of variants tweaked for specific preferences, target profiles, as part of A/B testing, etc. Changes to accent, emphasis, copy, maybe even which parts of the product being advertised are focused on.

- Increasing variation in open world game locations and interactions, with AI able to insert much more meaningful distinction between instances of a template (as compared with recent Assassins Creed games, where identical towns and forts are copy-pasted dozens of times all over the map).

- Much greater accessibility of visual art as an accent for other kinds of creators— illustrations to accompany poetry, blog posts, fanfic, etc.

- Increasing ability for AI to do first-cut assembly of video content, particularly review-type YouTubers (think: SkillUp, Critical Drinker) where the video is kind of secondary and the majority of the audience may even be listening to it as a podcast anyway. The AI being able to understand from the script what is under discussion and select matching clips would likely match or approach what

This is far from a sure thing— sometimes people really are the buggy-whip manufacturers in a situation and technology completely displaces them. But other times (as with software development), the market and use-cases have expanded considerably faster than the efficiency gains made by better tooling.

Re: The AI Unbundling

#60

Post by rco8786 »

mejutoco wrote: post #39
gizmo wrote: post #33

Although we talk a lot about disruption, only very few technologies are truly disruptive. You can tell by the panic and awe in the air whether you're dealing with real disruption or incremental change. Dropbox made filesharing easier. It's a good product, but not disruptive. Nobody panicked that Dropbox would make their job redundant. Uber was hard on the taxi industry, but fundamentally you still have drivers taking…

> People will insist that SD art isn't real art. Artists will fight back, and lose. When talking about Stable Diffusion and art there are usually two different aspects of art. I am not going to try to define art, but sometimes we refer to art as illustration, or stock images (what SD puts in danger) and some as broad modern art. I am no trying to say that one is more valuable than the other, but want to qualify these…

Worth noting that this is also currently on the HN front page: https://waxy.org/2022/09/online-art-communities-begin-bannin...
Post reply on HN