this post was submitted on 20 Nov 2023
618 points (94.4% liked)
Technology
59312 readers
5184 users here now
This is a most excellent place for technology news and articles.
Our Rules
- Follow the lemmy.world rules.
- Only tech related content.
- Be excellent to each another!
- Mod approved content bots can post up to 10 articles per day.
- Threads asking for personal tech support may be deleted.
- Politics threads may be removed.
- No memes allowed as posts, OK to post as comments.
- Only approved bots from the list below, to ask if your bot can be added please contact us.
- Check for duplicates before posting, duplicates may be removed
Approved Bots
founded 1 year ago
MODERATORS
you are viewing a single comment's thread
view the rest of the comments
view the rest of the comments
I have never seen ChatGPT produce images. Is this a feature of 4.0?
Yeah is this linked with dall-e?
It is. The paid version (GPT-4) is integrated with DALLE-3.
This has all the hallmarks of "human pretending to be an AI" rather than actual AI output
I disagree. This is as you say Precisely the type of thing that happens when an image generator is asked to make a chart/diagram, so to me it seems a really wild leap to go from "This looks like exactly what happens when X" to "someone must have designed this to look like what happens when X".
If it were human designed, I think it would be intentionally funny (which realistically would backfire, but anyway...)
(And besides, paid ChatGPT does indeed connect to DALL-E 3 now)
Tbf I thought DALL-E3 was still just available via bing image creator, missed the memo that ChatGPT was hooked up to it too.
Still, for me though it still looks like it's human generated to try and be funny (it's just haha-AI-so-silly isn't groundbreakingly funny any more). It's mostly the information continuity throughout the image that I've not really seen from an image generating AI before (especially when not even prompted for it), and I've had a play around with DALL-E3 so I would expect the ChatGPT version to be equivalent.
Maybe I'm too cynical, but this just reeks of fake to me.
I tried the same prompts as OP, it didn't generate an image at first instance - had to ask it to generate one. This is the image I got:
@[email protected]
Ropy from pituge
ChatGPT takes the liberty of creating a DALL-E prompt that it doesn't feel the need to share with the user. You can, however, ask ChatGPT to share the exact prompt and seed with you to reproduce the image. Here is the actual prompt and seed DALL-E ended up working with:
Prompt: "A step-by-step visual guide on using Optical Character Recognition (OCR) in Microsoft Word. The guide includes steps like opening Microsoft Word, inserting an image into a Word document, selecting the image, and using the OCR feature to convert the text in the image into editable text. The layout should be clear and easy to follow, with each step labeled and illustrated in a user-friendly manner, catering to users with basic proficiency in Microsoft Word."
Seed: 3993182816
To be clear, ChatGPT decided on its own to create and send this prompt to DALL-E in response to my request for tech support.
That's how you know the AI is good! actually.
/s
Why do you think that?
There's a level of continuity in the image you don't get with image generating AI yet.
Also it's littered with "AI getting things slightly wrong* memes
Also also, ChatGPT doesn't output images
It does: https://openai.com/blog/chatgpt-can-now-see-hear-and-speak
Edit: here's one I did now
Ah fair play, I missed that memo, the first two points still apply though
Yep, sure, it's a wild world we live in and this topic is changing fast. Missing this memo won't matter when the next one will be the next generation but generations are only 6 months apart.
I had a lot of fun asking it to draw ASCII art for me... especially if you ask it for corrections about specific aspects of its art