Live data from Hacker News

Imagen: An AI system that creates photorealistic images from input text

imagen.research.google

221–230 of 233 posts

Re: Imagen: An AI system that creates photorealistic images from input text

#221

Earlier quoted context omitted.

I’ve listened to a number of his podcasts, I’ve never gotten that impression at all. He comes across as a well versed, legit scientist. And there is lots of other evidence to back that up, not just my impression from a few podcasts, publications, tenured professorship at Stanford, etc. But in this case, does it really need extensive scientific study? Just think about it and look around. The thing our entire biologica…

Almost all of his diet/neuroscience/health claims (outside of commonly known things, like "exercise is good for you") are at best a gray area, and at worst just flat wrong. That is because we really don't know much about the brain and its relationship to human experience and behavior - and so anyone claiming to know how to improve your life with "neuroscience" is a con artist. Same with diet, which he also harps on c…

Is there something personal here as well? “Con artist” seems like a strong choice of words.

It’s interesting that I don’t really see that at all.

Maybe you’re a more frequent listener, I’ve just listened to a few where he is a guest and a few more than that of his solo episodes on his podcast which I particularly like because they just come across as a professor giving an extended lecture.

I don’t get any hype man, wannabe guru, snake oil vibes at all. He does use some of the Tim Ferriss type verbiage of like, optimize, and what not. (Not saying Ferriss is snake oil, btw) But I’d say that’s just par for the course in that genre of podcasting, it never seems to impede on valid information. He speaks like a scientist.

I’d imagine he’d be the first to say we know very little about the brain, but what do we mean by that? Things like qualia and consciousness, sure. But that doesn’t mean we can’t begin to parse out some things, like the things in his wheelhouse that he talks about, the effects of light exposure, positioning of the eyes, different types of breathing, etc.

It’s often easier to tear down than put up. Do you or anybody else you know do better at avoiding the criticisms you have here about the communication of research?

To some points- I’ve heard him talk about chemicals affecting the brain, which is a given right? Chemicals effect the brain in various ways. I’ve also heard him emphasize extreme caution and prudence in experimenting with chemicals at levels outside those one typically encounters day to day and often outright discourage it.

Light alcohol consumption is detrimental I’m no expert here but it is easy to imagine wishful thinking leading to the pro column.

I do notice marked difference in quality of sleep with even light consumption.

Do the pros outweigh the cons? Best left to a personal calculus I suppose.

You can "hack your brain chemicals" with supplements to achieve some desired effect on your mind

Is this a controversial statement? Have you conversed with anyone who has had a cup of coffee smoked a joint, ingested psilocybin or LSD?

The way he talks about the dopamine system and how too much redlining can throw the system out of whack seems well in line with what one observed in many other systems, biological and otherwise.

My life has been changed as everything changes your life to some degree however big or small. I’ve gained knowledge condensed from someone else’s life’s work in the same that occurs in reading a book.

This is a “force multiplier”(genre term) I should remind myself of more often.

I should spend more time reading books.

Re: Imagen: An AI system that creates photorealistic images from input text

#222
post #175

Is anyone aware of people doing the same for NSFW images? They can easily wipe out an entire industry. No models to pay and to check for legal ages, infinite possibilities: just write the pic you want, the massive body part you want, how many genitals are involved and boom. You have your image.

[0] Was front page recently. [0] (OBVIOUSLY NSFW) https://news.ycombinator.com/item?id=32572770

Thanks!

Re: Imagen: An AI system that creates photorealistic images from input text

#223

Earlier quoted context omitted.

Generating videos and 3D models is _much_ more difficult than images. You can’t just train off videos from the internet in the same way, because they don’t have sufficient text labels to understand them like CLIP does.

Oh but they have sound which can be annotated much faster/more efficiently. You also potentially have screenplay but the amount of training data is probably too less and sparse. FWIW, I don’t think the AI systems will generate a whole video by itself - it’ll be some form of image to image generation where an artist will render a rough sketch of the scene and the AI will fill in the details, frame by frame.

I mean at this point it’s really not a question of if, but when - right?

Re: Imagen: An AI system that creates photorealistic images from input text

#224
These things keep showing up latetly in various forms. A day or two about it was creating image of a porno star. But always a sometimes good sometimes crappy web app you can test it out, but usualy it is very limited or not possible.

The porn star one said it had used up its resource, and this one I cant figure out if i can run it. I get spreadsheet from the linkm but I am unsure how it helps me.

As a guy who does not know a lot of "AI" are these things just showing they "Hey I trained this then to do ...... / I have a big data model"?

Is there a standard way to share it?

If you start out with 10.000 penguins to teach a model about penguins. Once the model has worked through it, do you have any more use of the images separately?

How large are the models?

Re: Imagen: An AI system that creates photorealistic images from input text

#225
post #121
post #114

Earlier quoted context omitted.

right, because I can't be trusted to not be racist

Indeed, you can't be trusted not do anything bad with it. Who's going to vet each and every user? Who's going to check every time it's used for the coming 10, 20, 30 years?

If someone is racist to me online, I'll just turn off my screen

Re: Imagen: An AI system that creates photorealistic images from input text

#226
post #189

Earlier quoted context omitted.

That’s a pretty uncharitable interpretation of my post. You shared an example of a single mitigation that you personally find ludicrous (without explaining why). And then I’m supposed to throw up my hands and go “I guess it’s pointless to try and be less racist”? Bias amplification is a real issue. https://www.theverge.com/2016/3/24/11297050/tay-microsoft-ch... Tay might still be around if Microsoft gave a thought to…

> a single mitigation That is the only mitigation used in DALL-E 2, which up until recently was the only publicly available text to image model. > I’d prefer not to have this awesome technology tainted out of the gate as a tool for racists and pornographers Why is it your business what people do with the model? If people want to be racist they can already do so, they don't need a shitty model that doesn't work half a…

I just voiced my support for their decision. Just as you are voicing your displeasure with it. I don't see how we are different.

Re: Imagen: An AI system that creates photorealistic images from input text

#227
post #74

Every single time I see an article about a new AI model that has a section called "societal impact" I know immediately they are not releasing the model, the training set, nothing... It seems to be the kind of bullshit statement that those companies put in place of "we paid $500k training this model and we're not giving it for free to anyone".

> “While a subset of our training data was filtered to removed noise and undesirable content, such as pornographic imagery and toxic language, we also utilized LAION-400M dataset which is known to contain a wide range of inappropriate content including pornographic imagery, racist slurs, and harmful social stereotypes. Imagen relies on text encoders trained on uncurated web-scale data, and thus inherits the social biases and limitations of large language models. As such, there is a risk that Imagen has encoded harmful stereotypes and representations, which guides our decision to not release Imagen for public use without further safeguards in place.”

Re: Imagen: An AI system that creates photorealistic images from input text

#228

Earlier quoted context omitted.

If the seed remains the same, the difference between similar prompts is actually continuous. Here's a slew of images (1 through 5) I generated all from the same seed and same prompt sans a word or two: https://www.instagram.com/p/Chg60Fou6xB/

Thanks. Those are amazing by the way

Thank you mate. Spent a lot of time, even though the AI did all the _real_ work :-)

Re: Imagen: An AI system that creates photorealistic images from input text

#229
post #175

Is anyone aware of people doing the same for NSFW images? They can easily wipe out an entire industry. No models to pay and to check for legal ages, infinite possibilities: just write the pic you want, the massive body part you want, how many genitals are involved and boom. You have your image.

[0] Was front page recently. [0] (OBVIOUSLY NSFW) https://news.ycombinator.com/item?id=32572770

Looks like there’s a lot of road ahead for this, but def. A starting point

Re: Imagen: An AI system that creates photorealistic images from input text

#230
post #215
post #138

Earlier quoted context omitted.

Well… midjourney used stable diffusion (with an additional guidance model I believe, not just prompt engineering) for their beta model which they already closed down again… it’s back to their old far inferior model.

Why did they close it down?

The rumours are that it was too good at generating nudity for their comfort, and in particular that some users may have combined that with younger subjects.
Post reply on HN