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

DALL-E 2 generates images of Kermit The Frog in various films

twitter.com

Post reply on HN
Print view View original

181–190 of 198 posts

Re: DALL-E 2 generates images of Kermit The Frog in various films

#181

Post by gwern »

simiones wrote: post #133
silicon2401 wrote: post #34

Earlier quoted context omitted.

Are you aware that this is how human artists work, too? They make countless works that aren't comparable to their top works. Even more so if you count the practice when they were just starting out. I think some people just refuse to believe that real art can be generated without humans and selectively look for things that confirm their pre-determined conclusion. Witnessing this always feels like witnessing someone be…

There is a crucial difference: the human artist does this choosing themselves. If DALL-E just spits out 1000 images and then a human goes through them and picks the best 2-3, and those are good - it's impressive, but the human was still a crucial part of the process. On the other hand, if DALL-E were to generate 10 billion images, and choose the best 2-3 itself and give those as output, and if at least one of those 2…

> On the other hand, if DALL-E were to generate 10 billion images, and choose the best 2-3 itself and give those as output, and if at least one of those 2-3 would be consistently great, then DALL-E could be indeed considered to be creating (good) art.

It's worth noting that the OpenAI samples for DALL-E 1 used CLIP to rank generated samples, and got a big boost from that. For many model architectures, you can run them in reverse to do 'image -> caption', and 'score the caption' quality: if 'the caption is bad', that indicates your image was screwed up and low-quality (introduced by Cogview). DALL-E 2 doesn't use either approach, or finetuning on user choices like InstructGPT, and I dunno if OA is going to implement any of these, but there is a wide universe of techniques applicable here to improve quality and we should keep that in mind (https://www.gwern.net/Forking-Paths) if we are going to make any assertions more sweeping than "this specific model, at this very instant, with this particular interface, is only at this level of quality".

Re: DALL-E 2 generates images of Kermit The Frog in various films

#182

Post by treesprite82 »

orbital-decay wrote: post #157
dougmwne wrote: post #115

Earlier quoted context omitted.

These are fun discussions because words like "artistic creativity" have a colloquial meaning that could only apply to humans since the dawn of humanity. Now you have an image of Kermit in Wall-E. I have never seen or conceived of an image of Kermit in Wall-E. Let's assume that adorable robot Kermits do not exist in the training data to be spit out like a search algorithm. The image is new, it did not previously exist…

>So it sees like the only difference between the "Not creativity" that Dall-E is doing and "Real Creativity" that humans do is tht humans are the ones doing it? The differentiator is whether the result is worthy to look at for humans, that's all. In case of the OP, you forget that the human had to predict that the combination of two would be interesting for other humans, and then construct the prompt, possibly select…

> you forget that the human had to predict that the combination of two would be interesting for other humans, and then construct the prompt, possibly selecting the best pictures. That's who did most of the work here

If the Twitter user claimed that the text prompts themselves were generated by asking GPT-3 for "An interesting sequence of text prompts to feed an image generation AI" or something like that, I would have believed them.

Presumptively, I imagine it's harder to create a model that generates images matching a certain human-language input prompt than to create an image generation model with no language component and have it pick its own scenarios internally. I don't think the former is done to palm off "most of the work" to humans, but rather because people want an easy way to see create own ideas so there's more demand for it.

> You have to re-train it from scratch every time you want it to remember something truly new, there's no feedback loop to do that.

As far as I'm aware, this isn't true. Deep learning is perfectly compatible with fine-tuning an existing model using new data. OpenAI/MS have been doing this with Codex to improve it based on Copilot telemetry and code from new languages/libraries.

Re: DALL-E 2 generates images of Kermit The Frog in various films

#183

Post by tdehnel »

adamsmith143 wrote: post #139
tdehnel wrote: post #111

Earlier quoted context omitted.

Sorry to be this guy but that is not creativity. It’s using what already exists, not conjecturing something new. Contrast with real creativity (what people can do but machines currently cannot) where you conjecture something completely new. For example, Copernicus conjecturing the idea that the Earth revolves around the Sun. No machine learning model would have gotten there because it would have been trained on a bun…

>Sorry to be this guy but that is not creativity. It’s using what already exists, not conjecturing something new. Such a cute point of view, completely wrong but cute. Please go find the original images of Kermit in Blade Runner and WallE that were just copied here. >For example, Copernicus conjecturing the idea that the Earth revolves around the Sun. No machine learning model would have gotten there because it would…

The smugness of your reply annoyed me. But I feel somewhat satisfied that you’re just misinterpreting what I said.

I never said copying existing images, I said it’s using what already exists for inspiration. That is not the same as creativity.

I want to see what Dall-E comes up with when you ask it to create something new. Maybe “a new type of animal” or “an apartment building made to withstand radiation”. Basically anything that it can’t use existing images and ideas to create. Hell, I’d like to see someone ask GPT-3 for a prompt that Dall-E would fail at and pit them against each other.

The point is a human could come up with those trivially. I think this system would struggle. Because it’s not capable of creating anything new, only combining things that already exist.

Re: DALL-E 2 generates images of Kermit The Frog in various films

#184

Post by vouwfietsman »

tdehnel wrote: post #174
vouwfietsman wrote: post #154

Earlier quoted context omitted.

This is a common fallacy. Call it moving goalposts, no true Scottsman, the AI effect, whatever. The behavior is as follows: an argument over whether an ill-defined attribute is possessed by a computer is defended or attacked with useless semantics since nobody can agree on what any of the words mean anyway. Creativity, intelligence, consciousness. It doesn't matter what you say, you cannot define these concepts with…

This single quote from Wittgenstein might just be the dumbest statement ever made. How would we ever make progress if we did not discuss what we don't understand, and critique and debate new creative ideas?

By not taking the quote literally, but trying to understand it in its context.

The quote is a judgement about what can be logically achieved through discourse, and what cannot (and should not). He draws this boundary to separate the things we can reason about and the things that fall beyond reason and so cannot be reasoned about at all. If you cannot reason about something, you should not bring it up in (philosophical) discourse.

He does this to settle philosophical debates around, for instance, mystical subjects that were fought over to exist or not exist. The conclusion for him is: we do not have the logical tools and concepts to reason about the problem, so we should stop wasting time talking about it.

It's not about lack of knowledge or intelligence, its about acknowledging the limits of reason and discourse, and forgoing any conclusions made beyond those limits. It's a pursuit of truth.

At least, this is how I understand it, I'm not a philosopher.

Re: DALL-E 2 generates images of Kermit The Frog in various films

#185

Post by shrimp_emoji »

pphysch wrote: post #136
tdehnel wrote: post #111

Earlier quoted context omitted.

Sorry to be this guy but that is not creativity. It’s using what already exists, not conjecturing something new. Contrast with real creativity (what people can do but machines currently cannot) where you conjecture something completely new. For example, Copernicus conjecturing the idea that the Earth revolves around the Sun. No machine learning model would have gotten there because it would have been trained on a bun…

Humans don't generate new ideas from nothing. Your "real creativity" doesn't exist. Everything is dependent on what came before, and therefore derivative to some extent. Copernicus got his idea after gathering a lot of data, explicitly and implicitly, training his internal model of the world.

Even when it may appear that a human brain "generated something from nothing" (that is: you completely fail to account from where it could have possibly derived its output), you can always fall back to "genetic memory". Even a newborn has a lot going on that was "pre-programmed" by genetics, like a form of hardcoded training. :p

Re: DALL-E 2 generates images of Kermit The Frog in various films

#187

Post by tdehnel »

vouwfietsman wrote: post #184
tdehnel wrote: post #174

Earlier quoted context omitted.

This single quote from Wittgenstein might just be the dumbest statement ever made. How would we ever make progress if we did not discuss what we don't understand, and critique and debate new creative ideas?

By not taking the quote literally, but trying to understand it in its context. The quote is a judgement about what can be logically achieved through discourse, and what cannot (and should not). He draws this boundary to separate the things we can reason about and the things that fall beyond reason and so cannot be reasoned about at all. If you cannot reason about something, you should not bring it up in (philosophica…

It’s not controversial to say we have to stick to reason and logic in the pursuit of truth.

But there is nothing about understanding the difference between human and machine intelligence that is forbidden by any reason or logic that I am aware of.

Certainly there is a lot we don’t know about that topic. But unless that knowledge is specifically blocked by logic or the laws of physics, it is knowable. The only thing missing is the knowledge required to understand it.

The only source of knowledge is conjecture and criticism, and discussion with others is a great tool for that process. The more the better.

So unless you believe we know all there is to know about a topic right now, there is always the possibility that some new knowledge could help us get closer to the objective truth about it. And it’s guaranteed that new knowledge will come from creative new ideas and criticism of existing ideas.

Saying something is unknowable is akin to believing in a mysterious god creator, or fairies. But more importantly, it is a way to stop or limit the progress of our understanding of the universe. It’s like when parents say “because I said so”. It shuts down discourse and understanding and leads to closed societies.

Re: DALL-E 2 generates images of Kermit The Frog in various films

#188

Post by vouwfietsman »

tdehnel wrote: post #187
vouwfietsman wrote: post #184

Earlier quoted context omitted.

By not taking the quote literally, but trying to understand it in its context. The quote is a judgement about what can be logically achieved through discourse, and what cannot (and should not). He draws this boundary to separate the things we can reason about and the things that fall beyond reason and so cannot be reasoned about at all. If you cannot reason about something, you should not bring it up in (philosophica…

It’s not controversial to say we have to stick to reason and logic in the pursuit of truth. But there is nothing about understanding the difference between human and machine intelligence that is forbidden by any reason or logic that I am aware of. Certainly there is a lot we don’t know about that topic. But unless that knowledge is specifically blocked by logic or the laws of physics, it is knowable. The only thing m…

You're entirely misunderstanding me and the quote, as they both agree and align with what you're saying completely.

Again, the quote is not about avoiding or ignoring valid logical discourse, criticism, hypotheses, or any other logically coherent verbalization. The quote says that if the words are ambiguous, don't use them in logical arguments. Meaningless here does not mean hypothetical or conjectural or wrong, but simply ill-defined. This interpretation leaves plenty of room for everything you have described and are afraid would be criticized by the quote.

So, to get back to why I brought it up: creativity is an ill-defined concept in the context of logical reasoning, and so should not be used in logical reasoning as if it was an attribute that we can empirically or theoretically judge any entity to be in possession of.

This is why all replies to the above comment get lost in semantics: what is creativity? This is ill-defined, and hence based on your personal preference you may arrive freely at any conclusion you wish. Hence using the word "creativity" during a logical argument or truthful statement is guaranteed to produce conflict when multiple interpreters are present. Multiple logical interpreters can never be in conflict about purely logical reasoning, this is why we can have math.

Again, this is NOT saying that anything is unknowable, it is simply saying that whatever creativity is, or is composed of, is not clearly defined. This means it cannot be knowable or unknowable until it is clearly defined, and specifically only if done so in a rigorous enough definition to use it during reasoning.

To get back to your last point: doing what I described above is specifically to further the progress of our understanding of the universe, and specifically to attain new knowledge and understanding, because it cuts down on the discourse that does not contribute to that goal. It's the difference between arguing about whether machines have creativity and building DALL-E.

Re: DALL-E 2 generates images of Kermit The Frog in various films

#189

Post by FeepingCreature »

simiones wrote: post #142
FeepingCreature wrote: post #110

Earlier quoted context omitted.

For copyright purposes, the tweeter "made it themselves."

That remains to be seen. If I take a photograph of a still from the Matrix and print it, that's not the same as me photo-realistically drawing the same still from memory, which is itself not the same as me photo-realistically drawing it while looking at the still itself. Copyright law is way more nuanced than you think, especially around fair use.

I think this is already established. https://www.smithsonianmag.com/smart-news/us-copyright-offic...

Re: DALL-E 2 generates images of Kermit The Frog in various films

#190

Post by FeepingCreature »

bogwog wrote: post #121
FeepingCreature wrote: post #109

Earlier quoted context omitted.

Being bad for artists and the environment is not illegal. If you look at a movie poster, your brain does not need to be released in the public domain. Even if you sketch it from memory.

> Being bad for artists and the environment is not illegal. Yet we have copyright laws and environmental protection laws to protect both. > If you look at a movie poster, your brain does not need to be released in the public domain. Even if you sketch it from memory. Because I’m not a machine. I’m contrained by physics, whereas these models are not. Copyright laws were made to protect artists from IP theft. If I make…

> Yet we have copyright laws and environmental protection laws to protect both.

Yes, but we don't have copyright laws to protect the environment. So harm to the environment is not a copyright argument.

> Because I’m not a machine. I’m contrained by physics, whereas these models are not.

I don't even know what that means.

> But a painting, a book, a song, etc are easy to steal, especially with technology.

The AI does not merely duplicate training samples. That said, effort is also unrelated to copyright.

Post reply on HN