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

161–170 of 661 posts

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

#161

Post by jefftk »

jefftk wrote: post #117
mmh0000 wrote: post #89

Earlier quoted context omitted.

You mean one of Google's domains? # whois appspot.com [Querying whois.verisign-grs.com] [Redirected to whois.markmonitor.com] [Querying whois.markmonitor.com] [whois.markmonitor.com] Domain Name: appspot.com Registry Domain ID: 145702338_DOMAIN_COM-VRSN Registrar WHOIS Server: whois.markmonitor.com Registrar URL: http://www.markmonitor.com Updated Date: 2022-02-06T09:29:56+0000 Creation Date: 2005-03-10T02:27:55+0000…

While appspot.com is a Google domain, anyone can register domains under it. It would be similarly surprising to see an official GitHub blog post under someproject.github.io

Fun fact: appspot.com was the second "private" suffix to be added to the Public Suffix List, after operaunite.com: https://bugzilla.mozilla.org/show_bug.cgi?id=593818

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

#162

Post by spywaregorilla »

jvalencia wrote: post #122
jdashg wrote: post #95

Earlier quoted context omitted.

Additionally, if you optimize for most-likely-as-best, you will end up with the stereotypical result 100% of the time, instead of in proportional frequency to the statistics. Put another way, when we ask for an output optimized for "nursiness", is that not a request for some ur stereotypical nurse?

You could simply encode a score for how well the output matches the input. If 25% of trees in summer are brown, perhaps the output should also have 25% brown. The model scores itself on frequencies as well as correctness.

Suppose 10% of people have green skin. And 90% of those people have broccoli hair. White people don't have broccoli hair.

What percent of people should be rendered as white people with broccoli hair? What if you request green people. Or broccoli haired people. Or white broccoli haired people? Or broccoli haired nazis?

It gets hard with these conditional probabilities

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

#163

Post by Imnimo »

tines wrote: post #42
daenz wrote: post #7

>While we leave an in-depth empirical analysis of social and cultural biases to future work, our small scale internal assessments reveal several limitations that guide our decision not to release our model at this time. Some of the reasoning: >Preliminary assessment also suggests Imagen encodes several social biases and stereotypes, including an overall bias towards generating images of people with lighter skin tones…

This raises some really interesting questions. We certainly don't want to perpetuate harmful stereotypes. But is it a flaw that the model encodes the world as it really is, statistically, rather than as we would like it to be? By this I mean that there are more light-skinned people in the west than dark, and there are more women nurses than men, which is reflected in the model's training data. If the model only gener…

>If some particular demographic shows up in 51% of the data but 100% of the model's output shows that one demographic, that does seem like a statistics problem that the model could correct by just picking less likely "next token" predictions.

Yeah, but you get that same effect on every axis, not just the one you're trying to correct. You might get male nurses, but they have green hair and six fingers, because you're sampling from the tail on all axes.

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

#164

Post by LudwigNagasena »

not2b wrote: post #29
ceeplusplus wrote: post #14

Earlier quoted context omitted.

The ironic part is that these "social and cultural biases" are purely from a Western, American lens. The people writing that paragraph are completely oblivious to the idea that there could be other cultures other than the Western American one. In attempting to prevent "encoding of social and cultural biases" they have encoded such biases themselves into their own research.

It seems you've got it backwards: "tendency for images portraying different professions to align with Western gender stereotypes" means that they are calling out their own work precisely because it is skewed in the direction of Western American biases.

You think there are homogenous gender stereotypes across the whole Western world? You say “woman” and someone will imagine a SAHM, while another person will imagine a you-go-girl CEO with tattoos and pink hair.

What they mean is people who think not like them.

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

#165

Post by joshcryer »

young_unixer wrote: post #124
not2b wrote: post #29

Earlier quoted context omitted.

It seems you've got it backwards: "tendency for images portraying different professions to align with Western gender stereotypes" means that they are calling out their own work precisely because it is skewed in the direction of Western American biases.

The very act of mentioning "western gender stereotypes" starts from a biased position. Why couldn't they be "northern gender stereotypes"? Is the world best explained as a division of west/east instead of north/south? The northern hemisphere has much more population than the south, and almost all rich countries are in the northern hemisphere. And precisely it's these rich countries pushing the concept of gender stere…

The bulk of the trained data is from western technology, images, books, television, movies, photography, media. That's where the very real and recognized biases come from. They're the result of a gap in data nothing more.

Look at how DALL-E 2 produces little bears rather than bear sized bears. Because its data doesn't have a lot of context for how large bears are. So you wind up having to say "very large bear" to DALL-E 2.

Are DALL-E 2 bears just a "natural consequence of natural differences"? Or is the model not reflective of reality?

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

#166

Post by contingencies »

pxmpxm wrote: post #116
curiousgal wrote: post #64

Earlier quoted context omitted.

> it will be used in ways we haven’t anticipated Oh yeah, as a woman who grew up in a Third World country, how an AI model generates images would have deeply affected my daily struggles! /s It's kinda insulting that they think that this would be insulting. Like "Oh no I asked the model to draw a doctor and it drew a male doctor, I guess there's no point in me pursuing medical studies" ...

Postmodernism is what postmodernism does.

Love it. Added to https://github.com/globalcitizen/taoup

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

#167

Post by theptip »

cellis wrote: post #129
ShakataGaNai wrote: post #48

All of these AI findings are cool in theory. But until its accessible to some decent amount of people/customers - its basically useless fluff. You can tell me those pictures are generated by an AI and I might believe it, but until real people can actually test it... it's easy enough to fake. This page isn't even the remotest bit legit by the URL, It looks nicely put together and that's about it. Could have easily put…

Inference times are key. If it can't be produced within reasonable latency, then there will be no real world use case for it because it's simply too expensive to run inference at scale.

There are plenty of usecases for generating art/images where a latency of days or weeks would be competitive with the current state of the art.

For example, corporate graphics design, logos, brand photography, etc.

I really do think inference time is a red herring for the first generation of these models.

Sure, the more transformative use-cases like real-time content generation to replace movies/games, but there is a lot of value to be created prior to that point.

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

#168

Post by contingencies »

umeshunni wrote: post #121
Ar-Curunir wrote: post #65

Earlier quoted context omitted.

Except "reality" in this case is just their biased training set. E.g. There's more non-white doctors and nurses in the world than white ones, yet their model would likely show an image of white person when you type in "doctor".

Alternately, there are more females nurses in the world than male nurses, and their model probably shows an image of a woman when you type in "nurse" but they consider that a problem.

@Google Brain Toronto Team: See what you get when you generate nurses with ncurses.

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

#169

Post by userbinator »

user3939382 wrote: post #23
daenz wrote: post #7

>While we leave an in-depth empirical analysis of social and cultural biases to future work, our small scale internal assessments reveal several limitations that guide our decision not to release our model at this time. Some of the reasoning: >Preliminary assessment also suggests Imagen encodes several social biases and stereotypes, including an overall bias towards generating images of people with lighter skin tones…

Translation: we need to hand-tune this to not reflect reality but instead the world as we (Caucasian/Asian male American woke upper-middle class San Fransisco engineers) wish it to be. Maybe that's a nice thing, I wouldn't say their values are wrong but let's call a spade a spade.

Indeed. As the saying goes, we are truly living in a post-truth world.

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

#170

Post by colinmhayes »

curiousgal wrote: post #149
colinmhayes wrote: post #140

Earlier quoted context omitted.

Yes actually, subconscious bias due to historical prejudice does have a large effect on society. Obviously there are things with much larger effects, that doesn't mean that this doesn't exist. > Oh no I asked the model to draw a doctor and it drew a male doctor, I guess there's no point in me pursuing medical studies If you don't think this is a real thing that happens to children you're not thinking especially hard.…

> If you don't think this is a real thing that happens to children you're not thinking especially hard I believe that's where parenting comes in. Maybe I'm too cynical but I think that the parents' job is to undo all of the harm done by society and instill in their children the "correct" values.

I'd say you're right. Unfortunately many people are raised by bad parents. Should these researchers accept that their work may perpetuate stereotypes that harm those that most need help? I can see why they wouldn't want that.
Post reply on HN