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

Implementation of Imagen, Google's text-to-image neural network, in PyTorch

github.com

Post reply on HN
Print view View original

81–90 of 123 posts

Re: Implementation of Imagen, Google's text-to-image neural network, in PyTorch

#81

Post by turtledove »

tomatowurst wrote: post #24
dotnet00 wrote: post #18

Earlier quoted context omitted.

In my experience, scraping the data is the easy part. Once you've scraped it you've got to get rid of all the garbage, which is where the issues arise, especially if you're just blindly scraping everything you can find. For example, in a generative model I'm working on, I have a dataset consisting of ~5M images just blindly scraped from a website. After filtering, this drops down to ~500k images, yet a model trained…

Which is why porn is such a great dataset for crowdsource: - lots of people are stimulated by it - lots of people want DALL-E-2 for porn - and lots of people are willing to work towards that common goal The beauty of this is that people are just going to keep coming and coming to it. Like I'm trying to be mature and serious about this. What's it going to take? - Community responsible for scraping dataset, generating…

You are going to need to figure out how to filter out copyrighted works and images for which you don't have permission (eg, someone uploading a picture of an ex partner).

Re: Implementation of Imagen, Google's text-to-image neural network, in PyTorch

#82

Post by abrichr »

aero-glide2 wrote: post #74
jfoster wrote: post #60

Earlier quoted context omitted.

In fact, this kind of reverses things, doesn't it? Open source is built on the assumption that you can do more with source code than with binaries. In the case of AI models, the computed weights of models are what's valuable, and the source code used to achieve them is less useful.

How much would it cost in training to match dall-e 2?

https://twitter.com/alexjc/status/1347458546636619778?s=21&t...

> The blog post says 256 GPUs for 2 weeks, so:

> DALL·E would cost $131,604 to train on AWS, assuming a p3.16x-large at market rates. Could be as low as $40k if you already paid for reserved instances.

Re: Implementation of Imagen, Google's text-to-image neural network, in PyTorch

#83

Post by pmoriarty »

nautilus12 wrote: post #66
adamhowell wrote: post #57

Earlier quoted context omitted.

I'm currently working fulltime on AI-powered design suite Accomplice ( https://accomplice.ai ) and if you ask me on a good day I would tell you I do think there's already huge commercial potential. On a bad day, though ;) My current approach is a "model marketplace" ( https://accomplice.ai/models ) where the most popular open source text-to-image models (VQGAN+CLIP, Disco Diffusion, DALL-E Mega coming soon…), sit alo…

Why would people do this when aspiring artists are practically giving their real photos and paintings for free on places like deviant art? Why further commoditize something that's already been commoditize to practically free?

I'm an artist, and I'd absolutely love to use something like this to inspire me or to give me something to continue working with on my own.

In one sense it's kind of like a much "smarter" photoshop filter, where it can make your own art/photos look more like what you want (ex: Van Gogh, Dali, Picasso, or combinations of those, or something completely weird/new/different).

You could also train the models on your own work and have it generate art in your own style that could inspire you or could be useful to you either as a base to work from or that you could take interesting elements from to create new art.

Similar things can be done in music, by the way, and that would be really useful to musicians too.

Poets could use something like this to create poetry, novel writers to write novels, etc..

This is really an improvement on the collaboration potential between humans and computers -- which is probably why it's called "Accomplice".

Re: Implementation of Imagen, Google's text-to-image neural network, in PyTorch

#84

Post by pmoriarty »

tomatowurst wrote: post #28
DeathArrow wrote: post #23

Earlier quoted context omitted.

And what is going to take to generate videos, not pictures?

To run we must learn to walk first. To walk we must learn to be erect. To be erect we must master crawling. I don't think the leap is too crazy if we are talking short moving pictures without sound. However, when sound gets involved, this is where it would become very tricky.

"when sound gets involved, this is where it would become very tricky"

I don't think a neural net would have much trouble generating moans in sync to the motion.

Why pay humans for all that fake moaning when an AI could do it?

Re: Implementation of Imagen, Google's text-to-image neural network, in PyTorch

#85

Post by pmoriarty »

jstummbillig wrote: post #55
londons_explore wrote: post #50

Earlier quoted context omitted.

This tech is actually pretty bad at human faces - it all appears unnatural and distorted. You get things like the occasional extra earlobe...

While the "it will never do as well as humans at this one thing I feel strongly about" bias is still a common one, even on hn, at this point I am fairly certain that we will soon all have to live with the fact that we are no longer all that special in regards to, well, absolutely everything. One of my more interesting realisations over this development is that being human is apparently a religion to a lot of otherwis…

"we are no longer all that special in regards to, well, absolutely everything"

If/when general AI comes about, maybe so.

Until then all we've got are a bunch of highly specialized tools/helpers/slaves that may be good (in some sense) at one thing and awful at pretty much everything else.

You could argue that humans themselves are an ensemble of such highly specialized parts that are more than the sum of their parts in some ineffable way that's more of a "we know it when we see it" than something that's formalizable. Machines/computers lack that... so far.

Re: Implementation of Imagen, Google's text-to-image neural network, in PyTorch

#86

Post by pmoriarty »

nl wrote: post #41
TekMol wrote: post #16

What is the reason Google published their research details about Imagen? Why don't they just keep their findings to themselfes and build products on top of them? Public companies can't do stuff just for the fun of it, right? So there must be some commercial reasoning behind it?

> Public companies can't do stuff just for the fun of it, right? So there must be some commercial reasoning behind it? Yes they can actually. If you are a shareholder you can either sue (unlikely to succeed) or vote against the board. That's pretty much the only recourse.

"If you are a shareholder you can either sue (unlikely to succeed) or vote against the board. That's pretty much the only recourse."

Or you could sell or just threaten to sell your shares.

Buying more of the company's shares to take it over is another option.

Good luck doing that with Google.

Re: Implementation of Imagen, Google's text-to-image neural network, in PyTorch

#87

Post by pmoriarty »

29athrowaway wrote: post #79
lajamerr wrote: post #69

Earlier quoted context omitted.

Which part is illegal? Are you talking about porn? Or using potential copyright material in an infringing manner? Or perhaps the possibility of "deepfake" porn?

In some jurisdictions, deepfakes of any kind are treated as a form of defamation. In others, adult deepfakes specifically are illegal.

Which jurisdictions are those?

Re: Implementation of Imagen, Google's text-to-image neural network, in PyTorch

#88

Post by nmfisher »

adamhowell wrote: post #57
nmfisher wrote: post #42

Earlier quoted context omitted.

Two reasons: 1) Even though it's all technically very impressive, so far there's not a huge amount of commercialization potential here. OpenAI is charging for its GPT-3 model but its revenue is probably negligible next to the hardware costs (sunk + ongoing) to train it in the first place, let alone the researcher salaries they're paying 2) Most of the stunning examples are cherry-picked. These things fail much more o…

I'm currently working fulltime on AI-powered design suite Accomplice ( https://accomplice.ai ) and if you ask me on a good day I would tell you I do think there's already huge commercial potential. On a bad day, though ;) My current approach is a "model marketplace" ( https://accomplice.ai/models ) where the most popular open source text-to-image models (VQGAN+CLIP, Disco Diffusion, DALL-E Mega coming soon…), sit alo…

I meant commercialization potential for companies like Google, where anything less than a hundred million is probably a failure :) Hopefully for the non-Googles of the world (i.e. you), there's a good pathway forward!

Re: Implementation of Imagen, Google's text-to-image neural network, in PyTorch

#89

Post by Karawebnetwork »

tomatowurst wrote: post #5
eminence32 wrote: post #3

I wonder if this implementation was trained on the same dataset that trained DALL-E2, would it produce results of comparable quality.

I have to clarify I am visioning DALL-E-2-PORN. Couldn't you just scrape porn to get copious amount of dataset? Scope would be narrower and thus require less classification and in general it has common themes. I'm more concerned how expensive it will be to train it on GPU instances. We are looking at A6000s right? That's like $15/hr.

Sure, you could use images you do not have the legal rights to if you do not release anything at the end / just leak the final result. But this is a lot of effort and resources invested for something that would effectively never be shared on the clear web.

Re: Implementation of Imagen, Google's text-to-image neural network, in PyTorch

#90

Post by tbalsam »

Good to see LucidRains get the love he (rightly) deserves. He's a beast!

As a thank you to him -- he also does for work for commission/etc, check hus GitHub page for more info. I'm not fiscally or currently otherwise directly linked to him too closely, I've just hung around a while and think he deserves far more credit than he gets. This is literally the smallest piece of the pie of what the man does across several subdisciplines, send him a thank-you please, if possible!

Post reply on HN