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

Imagen, a text-to-image diffusion model

gweb-research-imagen.appspot.com

Post reply on HN
Print view View original

531–540 of 661 posts

Re: Imagen, a text-to-image diffusion model

#531

Post by tomp »

howinteresting wrote: post #418
tomp wrote: post #224

Earlier quoted context omitted.

You're confused by the double meaning of the word "bias". Here we mean mathematical biases. For example, a good mathematical model will correctly tell you that people in Japan (geographical term) are more likely to be Japanese (ethnic / racial bias). That's not "objectively morally bad", but instead, it's "correct".

Well that's not the issue here, the problem is the examples like searches for images of "unprofessional hair" returning mostly Black people in the results. That is something we can judge as objectively morally bad.

Did you see the image in the linked article? Clearly the “unprofessional hair” are people with curly hair. Some are white! It’s not the algorithm’s fault that P(curly|black) > P(curly|white).

Re: Imagen, a text-to-image diffusion model

#532

Post by SirHound »

rg111 wrote: post #493
qz_kb wrote: post #437

I have to wonder how much releasing these models will "poison the well" and fill the internet with AI generated images that make training an improved model difficult. After all if every 9/10 "oil painted" image online starts being from these generative models it'll become increasingly difficult to scrape the web and to learn from real world data in a variety of domains. Essentially once these things are widely availa…

People training newer models just have to look for the "Imagen" tag or the Dall-E2 rainbow at the corner and heuristically exclude images having these. This is trivial. Unless you assume there are bad actors who will crop out the tags. Not many people now have access to Dall-E2 or will have access to Imagen. As someone working in Vision, I am also thinking about whether to include such images deliberately. Using imag…

Most images you see from these services will not have a watermark on them. Cropping is trivial.

Re: Imagen, a text-to-image diffusion model

#533

Post by rhacker »

animal_spirits wrote: post #452
rhacker wrote: post #448

Earlier quoted context omitted.

Look at carpentry blogs, recipe blogs. Nearly all of it is junk content. I bet if you combined GPT and imagen or dalle2 you could replace all of them. Just provide a betty crocker recipe and let it generate a blog that has weekly updates and even a bunch of images - "happy family enjoying pancakes together" I can see the future as being devoid of any humanity.

The future digital landscape might be void of humanity, but there will still be real humans living next door to you ;)

The only interaction with people now is installing bark detector automatic dog whistles for our neighbors dogs and ring doorbells.

Re: Imagen, a text-to-image diffusion model

#534

Post by sexy_panda »

UncleOxidant wrote: post #63
sexy_panda wrote: post #10

Would I have to implement this myself, or is there something ready to run?

I think implementing this yourself is likely not doable unless you have the computing resources of a Google, Amazon or Facebook.

It seems like lucidrains is currently working on an implementation [1] of it.

I would love it.

[1] https://github.com/lucidrains/imagen-pytorch

Re: Imagen, a text-to-image diffusion model

#535

Post by rhacker »

slimsag wrote: post #465
rhacker wrote: post #448

Earlier quoted context omitted.

Look at carpentry blogs, recipe blogs. Nearly all of it is junk content. I bet if you combined GPT and imagen or dalle2 you could replace all of them. Just provide a betty crocker recipe and let it generate a blog that has weekly updates and even a bunch of images - "happy family enjoying pancakes together" I can see the future as being devoid of any humanity.

I wrote a comedic "Best Apache Chef recipe" article[1] mocking these sites. I guess the concern would be: If one of these recipe websites _was_ generated by an AI, the ingredients _look_ correct to an AI but are otherwise wrong - then what do you do? Baking soda swapped with baking powder. Tablespoons instead of teaspoons. Add 2tbsp of flower to the caramel macchiato. Whoops! Meant sugar. [0] http://slimsag.com/best-…

I don't know. If we have something like imagen or dalle, I can imagine something that can produce "tasty" food from random ingredients isn't far off.

Re: Imagen, a text-to-image diffusion model

#536

Post by robocat »

slg wrote: post #400
robocat wrote: post #378

Earlier quoted context omitted.

> I think it is reasonable for the creators to avoid sharing models known to not be smart enough to avoid exaggerating real world biases. Every model will have some random biases. Some of those random biases will undesirably exaggerate the real world. Every model will undesirably exaggerate something. Therefore no model should be shared. Your goal is nice, but impractical?

Fittingly, your comment fails into the same criticism I had of the model. It shows a refusal/inability to engage with the full complexities of the situation. I said "It is reasonable... to avoid sharing models". That is an acknowledged that the creators are acting reasonably. It does not imply anything as extreme as "no model should be shared". The only way to get from A to B there is for you to assume that I think t…

  “When I use a word,’ Humpty Dumpty said in rather a scornful tone, ‘it means just what I choose it to mean — neither more nor less.’

  ’The question is,’ said Alice, ‘whether you can make words mean so many different things.’

  ’The question is,’ said Humpty Dumpty, ‘which is to be master — that’s all.”

Re: Imagen, a text-to-image diffusion model

#537

Post by ismepornnahi »

qz_kb wrote: post #437

I have to wonder how much releasing these models will "poison the well" and fill the internet with AI generated images that make training an improved model difficult. After all if every 9/10 "oil painted" image online starts being from these generative models it'll become increasingly difficult to scrape the web and to learn from real world data in a variety of domains. Essentially once these things are widely availa…

Adding a watermark to all AI generated images should be imperative.

Re: Imagen, a text-to-image diffusion model

#538

Post by jillesvangurp »

agar wrote: post #463
qz_kb wrote: post #437

I have to wonder how much releasing these models will "poison the well" and fill the internet with AI generated images that make training an improved model difficult. After all if every 9/10 "oil painted" image online starts being from these generative models it'll become increasingly difficult to scrape the web and to learn from real world data in a variety of domains. Essentially once these things are widely availa…

The irony is that when the majority of content becomes computer-generated, most of that content will also be computer-consumed. Neil Stephenson covered this briefly in "Fall; or Dodge In Hell." So much 'net content was garbage, AI-generated, and/or spam that it could only be consumed via "editors" (either AI or AI+human, depending on your income level) that separated the interesting sliver of content from...everythin…

He was definitely onto something in that book where people also resort to using blockchains to fingerprint their behavior and build an unbreakable chain of authenticity. Later in that book that is used to authorize the hardware access of the deceased and uploaded individuals.

A bit far out there in terms of plot but the notion of authenticating based on a multitude of factors and fingerprints is not that strange. We've already started doing that. It's just that we currently still consume a lot of unsigned content from all sorts of unreliable/untrustworthy sources.

Fake news stops being a thing as soon as you stop doing that. Having people sign off on and vouch for content needs to start becoming a thing. I might see Joe Biden saying stuff in a video on Youtube. But how do I know if that's real or not?

With deep fakes already happening, that's no longer an academic question. The answer is that you can't know. Unless people sign the content. Like Joe Biden, any journalists involved, etc. You might still not know 100% it is real but you can know whether relevant people signed off on it or not and then simply ignore any unsigned content from non reputable sources. Reputations are something we can track using signatures, blockchains, and other solutions.

Interesting with Neal Stephenson that he presents a problem and a possible solution in that book.

Re: Imagen, a text-to-image diffusion model

#540

Post by barneygale »

robocat wrote: post #378
slg wrote: post #301

Earlier quoted context omitted.

>Could skewing search results, i.e. hiding the bias of the real world Your logic seems to rest on this assumption which I don't think is justified. "Skewing search results" is not the same as "hiding the biases of the real world". Showing the most statistically likely result is not the same as showing the world how it truly is. A generic nurse is statistically going to be female most of the time. However, a model tha…

> I think it is reasonable for the creators to avoid sharing models known to not be smart enough to avoid exaggerating real world biases. Every model will have some random biases. Some of those random biases will undesirably exaggerate the real world. Every model will undesirably exaggerate something. Therefore no model should be shared. Your goal is nice, but impractical?

> Your goal is nice, but impractical?

If the only way to do AI is to encode racism etc, then we shouldn't be doing AI at all.

Post reply on HN