3 ms·

The fact that you get access to DALL-E 3 "for free" if you're already subscribed to ChatGPT Pro is going to give MJ and other competitors a serious run for thei

The fact that you get access to DALL-E 3 "for free" if you're already subscribed to ChatGPT Pro is going to give MJ and other competitors a serious run for their money. Also being able to reuse a seed to emulate a InstructPix2Pix architecture is a game changer.

trompetenaccoun 3y ago
Maybe I'm not their typical target demographic but OpenAI's products are completely useless to me, the way they've neutered them and are restricting the output. For images I'd rather run Stable Diffusion locally. You own what you produce with it too (with some caveats). GPT-3 was cool before they came out with the chat version but it's been all downhill from there.
- willsmith72 3y ago
  I don't find the the quality bad at all, and find them extremely useful. So far it's my daily assistant for software architecture, business plans, marketing content, debugging, summarisations, and many more. Have you used gpt4?
- Spivak 3y ago
  The filter categories for OpenAI's moderation API are hate, harassment, self-harm, sexual, sexual/minors, and violence. Is it really the end of the world that DALL-E is rated PG-13? There's no limit to the available models to generate adult content, but having one that doesn't makes it embedable in other applications.
- garblegarble 3y ago
  >You own what you produce with it too As I understand it, this is incorrect - all of these outputs are the creations of a machine, and as such are not eligible for copyright protection
  - DharmaPolice 3y ago
    Surely a manual modification of the image would be a fairly easy way around this? Anyway, that may not matter - I think what people are looking for here is the absence of being sued by someone else, not the ability to sue someone else. If I generate an image to put in my companies annual report, I'm not worried about someone else copying the image somewhere else.
  - SamBam 3y ago
    There was a widely-misunderstood lawsuit recently that lots of people interpreted this way, but it wasn't what the lawsuit concluded at all. "Creations of a machine" is meaningless. Machines are tools we use to be creative, even if we're getting some things out that we do not expect. When a 3D program renders light into a scene, does that mean the image is not copyrightable?
    - objectivetruth 3y ago
      Which lawsuit are you referring to? I'm pretty sure GP is referring to the US Copyright Office guidance on copyrighting AI output in the US: https://copyright.gov/ai/ https://copyright.gov/ai/
      - SamBam 3y ago
        I was referring to the recent case against the artist Stephen Thaler, which people interpreted [1] as saying that works created by AI could not be copyrighted. This wasn't what was shown, though. Rather, the judge was ruling about a strawman argument framed by Thaler, where he explicitly stated that he was not the creator, and that he was listing the AI as an "artist for hire." I don't see where in your link it says, as GP did, that "creations of a machine [...] are not eligible for copyright protection." The guidance says that the author must be human, but that "In the case of works containing AI-generated material, the Office will consider whether the AI contributions are the result of “mechanical reproduction” or instead of an author’s “own original mental conception, to which [the author] gave visible form.” The latter is copyrightable. "This policy does not mean that technological tools cannot be part of the creative process. " 1. e.g. https://boingboing.net/2023/08/21/federal-judge-says-ai-generated-artwork-cant-be-copyrighted-because-of-monkeys.html https://boingboing.net/2023/08/21/federal-judge-says-ai-gene...
ranit 3y ago
Where on ChatGPT console can you access DALL-E 3? I am ChatGPT Pro subscriber, but all I can access, additional to GPT-4, are beta features (plugins and Advanced data analysis).
- SamBam 3y ago
  I also don't seem to have access. It seems that it's being slowly rolled out to subscribers, and maybe you and I are on the tail end.
  - ranit 3y ago
    I just got it!
- Auracle 3y ago
  For me it’s available in the app but not the website, although I can access the conversations that have generated images on the website.
GaggiX 3y ago
You can use Dalle 3 completely free on the Bing AI website.
- qingdao99 3y ago
  And it's far better in my opinion - ChatGPT actually creates its own prompts from what you give it and feeds those to DALL-E (you can see this by clicking on the images it returns and reading what the actual prompt was), although you can disable this by just firmly telling ChatGPT not to modify your prompt. Additionally, the image generation itself is quite different. ChatGPT's DALL-E seems to create much more stylized images - much harder to get plain shots that don't heavily embellish your description.
  - TeMPOraL 3y ago
    OpenAI justified that in a paper the other day, saying that DALL-E 3 performs better on longer, more detailed prompts describing all aspects of the image in rich language - so they put GPT-4 in front to expand the typical user's short and vague prompt, so such users get nicer results by default. My own observation: this kind of hack is possible only since/with GPT-4 - it takes an LLM this powerful to reliably extend and enrich arbitrary user input into much longer prompt, that's coherent, consistent, and a plausible (to human) interpretation of the original input. Now this may fan the flames on the "is it or is it not" AI discussions, but: you could almost say that GPT-4 is engaging in creative process here.

trompetenaccoun 3y ago

Maybe I'm not their typical target demographic but OpenAI's products are completely useless to me, the way they've neutered them and are restricting the output. For images I'd rather run Stable Diffusion locally. You own what you produce with it too (with some caveats). GPT-3 was cool before they came out with the chat version but it's been all downhill from there.

willsmith72 3y ago
I don't find the the quality bad at all, and find them extremely useful. So far it's my daily assistant for software architecture, business plans, marketing content, debugging, summarisations, and many more. Have you used gpt4?
Spivak 3y ago
The filter categories for OpenAI's moderation API are hate, harassment, self-harm, sexual, sexual/minors, and violence. Is it really the end of the world that DALL-E is rated PG-13? There's no limit to the available models to generate adult content, but having one that doesn't makes it embedable in other applications.
garblegarble 3y ago
>You own what you produce with it too As I understand it, this is incorrect - all of these outputs are the creations of a machine, and as such are not eligible for copyright protection
- DharmaPolice 3y ago
  Surely a manual modification of the image would be a fairly easy way around this? Anyway, that may not matter - I think what people are looking for here is the absence of being sued by someone else, not the ability to sue someone else. If I generate an image to put in my companies annual report, I'm not worried about someone else copying the image somewhere else.
- SamBam 3y ago
  There was a widely-misunderstood lawsuit recently that lots of people interpreted this way, but it wasn't what the lawsuit concluded at all. "Creations of a machine" is meaningless. Machines are tools we use to be creative, even if we're getting some things out that we do not expect. When a 3D program renders light into a scene, does that mean the image is not copyrightable?
  - objectivetruth 3y ago
    Which lawsuit are you referring to? I'm pretty sure GP is referring to the US Copyright Office guidance on copyrighting AI output in the US: https://copyright.gov/ai/ https://copyright.gov/ai/
    - SamBam 3y ago
      I was referring to the recent case against the artist Stephen Thaler, which people interpreted [1] as saying that works created by AI could not be copyrighted. This wasn't what was shown, though. Rather, the judge was ruling about a strawman argument framed by Thaler, where he explicitly stated that he was not the creator, and that he was listing the AI as an "artist for hire." I don't see where in your link it says, as GP did, that "creations of a machine [...] are not eligible for copyright protection." The guidance says that the author must be human, but that "In the case of works containing AI-generated material, the Office will consider whether the AI contributions are the result of “mechanical reproduction” or instead of an author’s “own original mental conception, to which [the author] gave visible form.” The latter is copyrightable. "This policy does not mean that technological tools cannot be part of the creative process. " 1. e.g. https://boingboing.net/2023/08/21/federal-judge-says-ai-generated-artwork-cant-be-copyrighted-because-of-monkeys.html https://boingboing.net/2023/08/21/federal-judge-says-ai-gene...

ranit 3y ago

Where on ChatGPT console can you access DALL-E 3? I am ChatGPT Pro subscriber, but all I can access, additional to GPT-4, are beta features (plugins and Advanced data analysis).

SamBam 3y ago
I also don't seem to have access. It seems that it's being slowly rolled out to subscribers, and maybe you and I are on the tail end.
- ranit 3y ago
  I just got it!
Auracle 3y ago
For me it’s available in the app but not the website, although I can access the conversations that have generated images on the website.

GaggiX 3y ago

You can use Dalle 3 completely free on the Bing AI website.

qingdao99 3y ago
And it's far better in my opinion - ChatGPT actually creates its own prompts from what you give it and feeds those to DALL-E (you can see this by clicking on the images it returns and reading what the actual prompt was), although you can disable this by just firmly telling ChatGPT not to modify your prompt. Additionally, the image generation itself is quite different. ChatGPT's DALL-E seems to create much more stylized images - much harder to get plain shots that don't heavily embellish your description.
- TeMPOraL 3y ago
  OpenAI justified that in a paper the other day, saying that DALL-E 3 performs better on longer, more detailed prompts describing all aspects of the image in rich language - so they put GPT-4 in front to expand the typical user's short and vague prompt, so such users get nicer results by default. My own observation: this kind of hack is possible only since/with GPT-4 - it takes an LLM this powerful to reliably extend and enrich arbitrary user input into much longer prompt, that's coherent, consistent, and a plausible (to human) interpretation of the original input. Now this may fan the flames on the "is it or is it not" AI discussions, but: you could almost say that GPT-4 is engaging in creative process here.