I like that. Which AI did you use?
Oh! I used ChatGPT 4o.
1 Like
Yes, big difference.
It gets better! I asked it to change the picture so it looked like a paper bag that contains a bunch of burgers. I wanted something taller to fit a physical book cover.
Within 5 minutes, a few variations, I came up with this:
2 Likes
I changed the topic title because the just released enhancements to Gemini and ChatGPT image generation deserves a more general description.
I’ve been experimenting with ChatGPT4.o’s recently enhanced image generation. It is getting remarkable. Here is a new prompt I used and the resulting image:
Prompt: create an image of a window. It is raining outside. There are drops on the window pane. There are flowers and trees outside. Inside is a table with a cup of coffee and book. The coffee is hot with steam. Make it photo realistic. Make it in a format I can use as the start page image in the Safari browser.
Result:
2 Likes
I tried the same prompt in Grok as an experiment. Two results for comment and comparison. I think this use case for AI is very impressive.
“Create an image that looks like a black and white sketch. Show a conference room with men and women sitting around it. Put a TV screen at the front of the room with a slide showing”
1 Like
That is really good, did you add something about “photorealistic” for the second image?
I feel sorry for the (non-existing?) guy who has to sit by himself at that small separate table… 
1 Like
He’s probably with the visiting accreditation team observing the dynamics of the board meeting.
1 Like
One more just for fun
Prompt: Create a Miyazaki-style image of a man sitting in a sport coat with a button-down shirt, sitting in a leather chair next to bookshelves. He is holding a cup of coffee and has an open book on his lap. He has brown hair and is wearing glasses. He is approximately 60 years old. Make the scene look like a warm study.
Added “photo realistic” for the second image.
3 Likes
Maybe slightly OT, but if something like that guy in the corner is ruining an otherwise great result, I just use a photo editor to remove it.
(For me, that’s Photoshop and just a few clicks with the AI remove tool or AI generative fill take care of it easily.)
I did not add anything. Grok produced two images. I am new to all this but I am amazed at the interesting and useful output using these models, including the variety depending on the model.
I agree. I think with some practice, I can use AI to generate images I can use in Keynote slides. Just one more example, my daughter gave me a prompt so I could generate and image for her. Here is what it did–she loved it.
2 Likes
On picture #2, does anyone see the possibly non-human at the table?
1 Like
In the first image I can see: a man with three hands; Siamese twins conjoined at the head: a man with three fingers on the right hand; a man sitting half way into the desk, with his forearm bent back to front as it disappears into the blouse of the women behind; and almost everybody else wearing mittens or glove puppets.
Striking, perhaps, but surely not worth throwing human illustrators out of work for.
1 Like
Continuing down the rabbit hole, I updated the prompt, using the second image, with the android like person, to be:
“Update this image so that the second person on the right side of the table appears to be a woman instead of a man.”
Attached is the new images generated. Interesting that the hands appear to be the give away; they are not well done. But having said that, on first glance, pretty good.
I am also throwing out there that I have no expertise on the proper way to phrase prompts. I expect a more experienced person, could refine the image appropriately.
True. It’s helpful for us non-illustrators to conjure up a few images that we’d never pay an illustrator for.
I am writing a manifesto to complement my new book, The Bottleneck Detective, and I pasted the text and my book cover into ChatGPT and it came up with some options.
They weren’t what I wanted, but they helped me validate that my idea was worth it, and I’ve no passed the text and the examples to my assistant who will work on them in Canva, using the original book cover files as a starting point.
She’s keen. And I enjoyed my play with ChatGPT.
I probably would have given up without chatGPT’s help in testing the ideas out. My assistant would have had no work to do and my readers would have missed out on my snazzy new manifesto!
Did you know that the real artist considers this an insult?
3 Likes
I’m sorry if this appears unnecessarily blunt, but personally, I will not knowingly give money to an author or company who uses generative AI, either for illustrations or for text. Even in the case of non-commercial material, then the use of Gen AI without attribution makes me less likely to trust the author.
That doesn’t mean that I don’t treat the work of human authors/illustrators sceptically – I do — but the use of Gen AI without acknowledgement adds a clear warning sign that the author is prepared to cut corners and/or deceive their readers.
Beyond the obvious quality and trust issues, to me the ethical issues involved in using a tool almost certainly trained on stolen copyright data outweigh any momentary convenience to the author. Should the AI companies be forced to pay adequate recompense to the authors/creators they have stolen from – which seems vanishingly unlikely given the kleptocracy currently in charge – then fine, though the environmental concerns will persist.
It may be that we’re too late to put this particular genii back in the bottle, but that doesn’t mean we should all give up, does it?
3 Likes
@rob Thanks for passing this along. It was an excellent read. In fact, I’ve saved the article in DEVONthink for future reference.
When I asked AI to create an image in “Miyazaki-style,” I didn’t know what that was nor had I heard of Studio Ghibli. That is not surprising given that I’m not on social media. I ran across the phrase a couple of days ago somewhere, I don’t recall where, and decided to see what happened. Until I’ve had time to more carefully consider this matter, I’m not going to ask AI to create anything in the Miyazaki-style.
The rise of AI presents both intellectual and ethical challenges. Since everything AI produces is ultimately derived from human input, one could argue that any AI-generated content potentially infringes on copyright. More troubling, it could also be seen as diminishing the value of human artistic and literary expression.
Does this mean we should never use AI to help create written content or images? The answer may not be so straightforward. I recall reading—perhaps here or on the Focus forum—about a book on creativity that suggested all creative work is, in some sense, borrowed from others (forgive the rough paraphrase). According to this perspective, nothing is entirely original. Every writer or artist draws on what they’ve learned from or been influenced by. Everything is in some way derivative.
By that reasoning, one could argue that AI does something similar. It doesn’t copy works outright; rather, it draws “inspiration” from a wide range of sources and assembles (I think that is a good description) something new—an image or piece of writing that, while shaped by past works, is not a direct replica.
For example, I recently saw an image of a seafaring captain that I found compelling. Instead of copying it for a blog article, I took time to consider its key elements and then wrote a prompt to create a new image—one inspired by the original, but not a direct copy. The result was this:
I used that image with a blog article, and added the caption, “AI Generated.”
I would consider this “original art” in the sense that I described the desired elements, and the AI assembled the image based on that description. To me, this doesn’t feel much different from studying a piece of art for inspiration and then creating something new from it myself.
Finally, for what it may be worth, I have disclaimers on my blog site as follows:
AUTHORSHIP
Unless otherwise indicated, all articles are authored by Dr. Barrett Mosbacker.
HUMAN-CREATED CONTENT POLICY
All content on this blog is created by humans unless explicitly stated otherwise. Any use of AI-generated content, such as images or text, will be clearly noted in the respective post or section.
As I explore and experiment with AI tools, I will indicate if and how AI was used in the creation of any content to ensure clarity and integrity.
All of this is going to take time to figure out. Hopefully, the courts will help define the parameters of what is and is not infringement. While what is legal does not always define what is ethical, the law can provide guidance.
Honestly, at this early stage, I’m still figuring it out. These are simply some meandering thoughts over my morning coffee. That said, I’ve decided not to request images in the Miyazaki style, but I will continue to explore and use AI in ways I believe are authentic and ethical.
3 Likes
I think we all are. @rob and @brookter arguments are valid.
I believe the issue is in what copyright is. Or, better yet, the reason why it exists. I think OpenAI has trained its models in copyrighted material, that’s for sure, and has thus caused damage to the copyright owners.
But there’s a bigger, darker history here: AI not only threatens copyrighted work. It threatens all work, I think Miyazaki will have a case in a court law, but if you happen to be a worker in a contact center you will be replaced by an AI. Or if you are a taxi driver. And you will not have copyright laws to protect you.
4 Likes