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

151–160 of 207 posts

Re: The AI Unbundling

#151

Post by notahacker »

sinenomine wrote: post #101
slowmovintarget wrote: post #88

Earlier quoted context omitted.

Doesn't "the AI" train on art produced by people? "Just expand the dataset, just increase the parameters" seems like it should hit a wall fairly quickly... and still not be very good, because deep learning systems have no insight.

Every instagram, facebook, and tiktok photo with associated text data is a potential pair for training. In the smartphone age, the case for data hunger looks pretty weak.

But not as weak as the case that the route to production grade commercial art is reached via biasing the training dataset more towards sloppy social media images...

Re: The AI Unbundling

#152

Post by handoflixue »

erispoe wrote: post #104
dwohnitmok wrote: post #64

Earlier quoted context omitted.

> Because that's not true at all. The AI can't even draw hands yet. To say nothing of its ability to handle multiple people and objects interacting in complex scenes. This seems to be purely an issue of the size of the network. Parti ( https://parti.research.google/ ) demonstrates that as the number of parameters increases, with no change to the underlying architecture, a lot of these problems simply go away. Basical…

People who buy art do not buy it because of the technical execution. You may need to execute a piece in some way to get a desired effect, but the technique is the mean not the goal. This is not to take away from the achievements of AI. It's that creating pictures adhering to a prompt with some degree of creativity is very little of what art is. Maybe it will replace some part of commissioned illustrations where the a…

> It's that creating pictures adhering to a prompt with some degree of creativity is very little of what art is.

I mean, it's the majority of commercial art - you get a prompt from the client, you maybe flesh it out in a few different directions with sketches, then you refine a final piece. And AI is incredibly good at this process - instant results, infinite patience, and it's free. A very hard combination to meet.

Calendars, book covers, video game assets, green screen backgrounds....

Even in a case like video game animations, where the AI can't build every frame, it can still give you a good reference photo. From there you just need a cheap artist to fill out the frames - a huge cost savings, and a big blow to the artistic community.

Where do you get started as an artist, without any of those? Obviously, Fine Arts isn't nearly as effected, but how do you get your start when you can't build a name from your cool book covers, or get famous off Magic: The Gathering card illustrations?

Re: The AI Unbundling

#153

Post by thunderbird120 »

joe_the_user wrote: post #141
yojo wrote: post #51

Earlier quoted context omitted.

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 t…

AI is not a "force multiplier" in any literal or semi-literal way. Having an mediocre version of an image isn't something that gives an artist much of a leg up on producing a good version. If you want something a diffusion model can produce, you can get it for nearly free. If you want something significantly better, you have to pay an artist significant money for it.

Except it is actually if you've been following the workflows which have popped up since stable diffusion's release. These models are not limited to just making a image which vaguely matches some description. They can also inpaint specific portions of an image, turn a vague sketch into a finished looking image, and blend together different images in convincing ways. A workflow for creating a high quality looking piece goes something like this:

1) Quick, low quality sketch

2) Generate detailed scene based on that sketch

3) Fix obvious errors by re-generating specific parts of the image

Re: The AI Unbundling

#154

Post by thematrixturtle »

Jevon23 wrote: post #40
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…

>It's going to destroy the livelihoods of the majority of independent artists in a way that looks inevitable to me. Why is SD going to destroy the livelihoods of artists when machine language translation hasn't put human translators out of work yet? I don't think there's been any industry that's been ended by AI yet, and yet people are strangely confident that art is going to be the first.

Machine translation has put a lot of human translators out of work. The per-word rates for text translation are pathetic these days, even for language pairs that Google Translate struggles with.

Of course, there are still some jobs for high-quality/high-importance translation like legal work, simultaneous translation etc, but these are quite niche.

Re: The AI Unbundling

#155

Post by robertlagrant »

enord wrote: post #124
robertlagrant wrote: post #113

Earlier quoted context omitted.

You're not justifying your premise, so those two sentences are essentially free of content. Mind if I use them in my gallery's modern art exhibition?

If you do not see how you are subverting the very argument that states your position, I don’t know what to tell you. If you do — touché; well played to you Sir!

Ha! Real talk: I still have no idea what you're saying. I wish you did know what to tell me.

Re: The AI Unbundling

#156

Post by echelon »

idontpost wrote: post #117
echelon wrote: post #81

Earlier quoted context omitted.

My startup can make you sing as Taylor Swift. Not like Taylor Swift. As Taylor Swift. We'll have our new product out very very shortly. We're at end times.

Why would I want that? Even if I like Taylor Swift, what would be the point?

Today you're a Taylor Swift fan and you want more of her music faster than she can deliver. Maybe you make it yourself and share with friends, and it goes viral. She might even get on stage and sing it.

Tomorrow, you want more Taylor Swift and you tell the computer to give it to you. It delivers with no human intervention whatsoever. You love it, but want it a bit more country - so be it. Piano and acoustic. Got it. A duet between her and Bruno Mars. Just for you.

Michael Jackson and Elvis? Nothing is without possibility.

Brand new singers. Instruments and styles we could never conceive of - all about to happen.

A never-ending stream of novel music that tailors itself to you.

Already being dreamed up.

I'm building this, have parts of it already working, and I'm sure others are as well.

Every inch of your day to day life is about to change in a way that will make developments brought on by smartphones and the Internet pale in comparison.

Re: The AI Unbundling

#157

Post by dan_mctree »

monkeynotes wrote: post #137
gizmo wrote: post #63

Earlier quoted context omitted.

I think the sad reality is that modern art is largely trash. Perhaps because painting/drawing/sculpting with excellent technique is too much of a commodified trade already. Perhaps because it's too hard to be original. A new oil painting of a tree, no matter how well executed, won't draw people to an art gallery. So artists take refuge in social commentary: A banana taped to a wall. A naked person holding a trash bar…

Art isn't about skilfully painting a tree. That's artistry. Art is more than that, it's about feeling, evoking thoughts, and yeah, commenting on society. > It's always trite, reductive, and utterly unoriginal. This is simply untrue, modern art is not "always" anything. And who are you to judge what is going on in the mind of an artist? I don't know what decade you deem to be "modern" art, but there is a huge amount o…

> but there is a huge amount of 20th - 21st century art that is jaw drop

What would be some examples of that? Not disagreeing, would just like to have my jaw drop

Re: The AI Unbundling

#158

Post by telotortium »

echelon wrote: post #81
simiones wrote: post #78

Earlier quoted context omitted.

> Stable Diffusion (and similar tools) fall in this last category of truly disruptive technologies. I really don't see it. What tools like SD, DALL-E etc do is essentially similar to what Google Image Search does, except that the results are presented differently - instead of showing you the results it has indexed in some order, it picks features from the results it has trained on with some priority, and presents a k…

My startup can make you sing as Taylor Swift. Not like Taylor Swift. As Taylor Swift. We'll have our new product out very very shortly. We're at end times.

If it's as good as your Donald Trump voice conversion demo on your site, I'd have to pass. That sounds like someone with a cold trying to do a poor impression of Trump. The voice doesn't even have a New York accent!

Re: The AI Unbundling

#159

Post by sinenomine »

clankyclanker wrote: post #119
sinenomine wrote: post #105

Earlier quoted context omitted.

This nitpick about hands sounds desperate. Here we are, with a tech so powerful that it overshadows the default hype it's surrounded by (no small feat, most technologies fail to live up to the hype as you know) ... and the critics merely move the goalpost a tiny bit further, even if the tech scales so well as to make their new goalpost irrelevant in a year.

Sure, but eventually we’re going to hit on environmental and cost-effective power limits of training, and it’s not worth the cost to train the model. AFAIUI, that’s part of the point that Gebru was trying to make before she was fired.

For now I can run my stable diffusion on a vintage laptop from a decade ago, on CPU (!), and it doesn't even utilize most of my RAM. And training this model was still cheap compared to, say, a google senior engineer yearly salary. The limits of scaling are further than laymen may want to believe.

With an order of magnitude more parameters it won't just do hands, it will do quite a bit more.

Re: The AI Unbundling

#160

Post by echelon »

sinenomine wrote: post #97
gizmo wrote: post #63

Earlier quoted context omitted.

I think the sad reality is that modern art is largely trash. Perhaps because painting/drawing/sculpting with excellent technique is too much of a commodified trade already. Perhaps because it's too hard to be original. A new oil painting of a tree, no matter how well executed, won't draw people to an art gallery. So artists take refuge in social commentary: A banana taped to a wall. A naked person holding a trash bar…

Trash, but trash minted through an entrenched network of art curators, directors and connoisseurs - this is what makes trashy modern art scarce and valuable, not to mention the tax breaks! What I and some insiders consider the real art - that is, applying rare well-developed talent to production of unapologetically high-quality sensual artifacts of art - is mostly commercialized by now, with best people employed by t…

Literally none of the existing world matters anymore.

Period.

We're watching the Third Industrial Revolution unfold live and in front of our eyes, and it will only take a few short years to fully sweep into all aspects of society.

Yes, Picasso and Banksy will continue to have value. But that's an infinitesimal sliver of the value that is about to be unleashed upon the economy.

The real value is that startups like mine will give anyone the power to generate new Taylor Swift songs [1].

New movies, new music, new games. All tailored to the nichest and most exacting of interests.

You want the steampunk vampires of Cloud City, featuring Sean Connery and young Eddie Murphy duking it out while riding on the backs of space whales? You'll have that in just ten short years.

If you think TikTok adequately captivates the minds of the population, you have no idea what a finely tuned search for people's true interests will do.

Humans won't be good enough at satisfying other humans anymore. Not without aid.

The new world is going to blow our minds, and it's already beginning to unfold.

In about a year, I don't know why anyone would dare work for Google, Amazon, or any of the other legacy businesses. They're going to be swamped and struggle to stay afloat amid the biggest single act of the Innovator's Dilemma to ever happen in our history.

[1] This is happening. My startup does it.

Eg. https://www.youtube.com/watch?v=r_0JjYUe5jo --> https://vocaroo.com/1hgjjnVNqWjk

(We'll have the Obama and Taylor Swift versions up on our site soon.)

Our video stuff is under development, but I'll show you something analogous that a skilled competitor is doing in the open (though they're underselling themselves) : https://imgur.com/seBTPG8

The whole world is about to be carved up and reallocated.

I'm going to get back to work. See you on the other side.

Post reply on HN