Hacker News Reader

Comment by nomel | Hacker News Reader

> I can only guess either safety from misuse

I still don't understand what this would mean. Where are all of the terrible things that were supposed to happen, now that Stable Diffusion is available?

We've been able to create completely photorealistic fiction for decades now. See any movie with CGI for an example of whole worlds, and people, that don't exist. The bar has gradually been lowering (see the amazing CGI that YouTubers do these days), and now maybe there is a bit of a step function down, but being able to make things that aren't real isn't remotely new. I don't understand the fear.

View on HN

While I think the fear might be overexaggerated, being able to make realistic fake content with such ease means it's harder to know what's true and what's not. Plus this has been the claim of OpenAI from the beginning. It's possible the true objective is to keep it private to leverage for money and this is just their excuse.

And how often does this happen with Photoshopped images that aren't immediately disproven?

My grandmother once emailed my family frantically after she saw a picture of the Abraham Lincoln statue defaced with graffiti. Obviously that was a Photoshop, and in this case, even a bad one, but clearly fake images and content make it harder to discern truth

> means it's harder to know what's true and what's not

The danger for society is in not already knowing that is the case, since it's relatively trivial, without AI, to make fake content.

Well sure, I think that's dangerous too. I think more people should be skeptical of the images and content they consume in addition to it being a problem that truth is harder to discern.

Indeed, this talk OpenAI does is basically security through obscurity, and it's holding us back. Look at how often people make noise with screenshots of tweets or emails that never happened. You don't need photorealism or fancy machine learning for that, and it creates a lot of problems! If they weren't pretending that all we need is to put some yellow tape around machine learning, maybe there would be some interest in solving this type of stuff properly. But you don't need "AI" for that. You just need public awareness and some basic, pre-existing cryptography knowledge.

Photoshop is a pistol, this thing is a gatling gun.

We are already living at a gatling gun level of content, from any individuals perspective into the web.

We are way past that. If anything, trivialising and democratising the generation of fake content, will educate the masses. Perhaps, bring back the trust for science, which we need a lot.

Like every new technology that came before, it’s a tool that will be used for good and for nefarious purposes. Some tools are powertools. There will be new benefits, but be prepared for a new raft of modern problems.

> means it's harder to know what's true and what's not.

Obligatory XKCD:https://xkcd.com/2650/

>> I can only guess either safety from misuse

> I still don't understand what this would mean. Where are all of the terrible things that were supposed to happen, now that Stable Diffusion is available?

Mainly people making porn (e.g. stuff like deepnudes). It seems like a lot of work has gone into into preventing that (e.g. filtering porn out of training data, having porn-detection models to block porny output). There's also been a lot of talk about political fakes, etc, but I'm not sure how likely that is to actually happen at this point. I think one of the "selling points" of limiting access to DALL*E was that they could revoke access to people who they deemed to be misusing it.

Someone else will come along that doesn't have the same arbitrary limitations, it's a battle you're bound to lose.

Agreed. Unfortunately, it's just a matter of time before someone develops a similar open source project that someone uses for unsavory purposes.

A problem is that training huge model costs $$. https://twitter.com/EMostaque/status/1563870674111832066

I wonder is it possible to distribute the task for volunteer's idle GPU (like BOINC)

The porn industry is much more than $600k. If there's a business model/need for deepfake porn somebody would do it eventually.

Porn industry will happily pay but I suspect that they won't open source it.

Stable Diffusion allows anyone to make kiddie porn with a half-second of curiosity/effort. Maybe you didn't know about that, maybe you think it's NBD, but in any case, that is the tire fire which aspiring AI majors want to avoid.

Pen and paper can do the same. Or Photoshop. Anyone can draw anything! OMG, stop the paper factory.

>Anyone can draw anything!

I'm pretty sure one of the primary arguments for Dall-E and Stable Diffusion existing is that there are lots of people who can't draw anything.

Stable diffusion can make very realistic looking images (probably videos soon) that is accessible to anyone.

Actually, anyone can't draw anything, especially the closer you get to photorealism. Also drawing by humans cannot be done in parallel very well. With an AI you can get an unending stream of pictures.

I can't quite decide if that's good or bad.

On one side there's no problem to like CP if it's AI generated and no real kids are harmed.

On the other side it may desensitize the public to this type of content.

It’s been a week, there is going to be an explosion of believable fake items that are going to be used to lure people in to even more unbelievable conspiracy theories than there currently exist. Your average conspiracy nut didn’t have the skills or know how before, but they sure do now.

Also you’re probably not seeing all the pedo content that people are already generating for themselves.

> be an explosion of believable fake items

I think that's a good thing. A slow creep is dangerous, because people may not notice, and would be tricked. An explosion of insanity makes everyone realize they can't believe everything they see on the internet.

It will be all of the above so it doesn’t really matter since it’s all going to be indistinguishable from the real thing.

Which has been the case for the last few decades. Nobody should be trusting anything they see on the internet.

It might end up having the opposite effect. When all sides of an argument can bomb each other with AI generated content it could poison the well and make people less likely to believe things they see.

Instead they'll pick and choose whatever to believe the conforms to their world view or comfort, and dismiss the rest as 'fake news'.

We'll get more Q-Anon than Wikipedia.

You have to wait a little bit more, until HD video synthesis is possible on a mid-range GPU. Then on a mid-range smartphone.

(see the amazing CGI that YouTubers do these days)

Any favorite examples?

Here's one (that also goes into the details of how it was done): https://www.youtube.com/watch?v=0Cz8CjLq0fQ

Thanks! Are you currently hacking on any interesting hardware projects, if you don't mind me asking?

CGI sucks. Monty python had the best special effects.

Reply on news.ycombinator.com