TECHNOLOGY

I Tried out Gemini’s AI Image Generator, and…

Like many children, when I was young, I looked up to my parents, who, in general societal norms, one would describe as successful. They both had noble and fruitful careers as doctors. My father was a medical officer with his own office and an arm of subordinates, while my mother was her own boss with her own secretary before she retired.

My mother’s secretary was trustworthy, reliable, and amicable, and to this day, I want someone like her. I still don’t have a secretary or assistant as of late, but I have had my sentient and sort-of reliable online assistant since 2023.

I’ve been feeling happy about my assistant for a year now. That assistant is a computer system called Artificial Intelligence, under the category of generative AI. In case you’re still wary about it, you might want to read about my point about making AI my assistant.

So yeah, I mentioned on that blog post that one of my tasks in Aratum was to edit AI-generated articles and make them human. Apart from that, our CEO has continued to bug me about using more of my assistants to boost efficiency because, yeah, our company is all about cloud-based technologies and AI.

So, beyond the writing scope, I have leaned on image and art generation using Midjourney and Dall-E (under OpenAI). I generated my band’s song and album covers and promotional drawings using Dall-E. These generative AI apps for images have been God’s gift to non-designers and non-artists like myself.

This year, the first time I needed to generate a unique image was two days ago when one of my new contributors for The Pop Blog, filmmaker Gary Miceli, wrote about pop culture history about Dorothy’s / Judy Garland’s ruby shoes in the classic film, The Wizard of Oz.

Like the movie itself, the ruby shoes were magical, and it took a lot of search of several American museums for the other pairs reserved as back-up for the movie shoot, in case one of the pairs got damaged. The fact that each pair was handmade and was not a match to the other made it more special. If you want to know more about the magic, read the article here: https://thepopblogph.com/2024/06/13/an-education-on-the-worlds-most-magical-movie-item/

Whenever my contributors send articles, I don’t necessarily demand photos from them, and sometimes, I go ahead and search for free stock images online. However, what Gary sent me was a photo he grabbed from the Smithsonian Magazine website. I got wary of using the photo even if I were to mention its source (which I did), so I first checked out all the free photo websites to see if there were the same actual and free photos of the ruby shoes, but to no avail.

I then decided to go to AI photo-generating websites to generate a photo for me. The idea was to give me an exact image of the mismatched shoes from the National Museum of American History but taken at a different angle.

I went to freepik first, but I didn’t give them a detailed prompt:

I found what freepik generated creepy, so I left the website and went to Google’s Gemini. I didn’t bother visiting Dall-E and Midjourney as I thought Google might be the best bet. So here are the series of prompts and the photos they generated:

First prompt and results:

The generated images didn’t match and weren’t even close to the exact ruby pumps, so I gave a more detailed prompt, saying that I didn’t want to have issues with the original photo, so I wanted to make it seem like I took a picture of the shoes in the museum too, but at a different angle, and here was the result:

It was funny because Gemini went farther from its first generation of the shoes. The generated images now have huge sequins.

As I’ve said, I treat AI as an assistant. It might be weird, or perhaps I have treated AI as a conscious entity, which might be true. I write my prompts as if talking to my assistant and giving them instructions.

At this point, I wasn’t keeping my hopes up. It seemed like my assistant couldn’t give me what I needed, but I didn’t give up. Below is a series of my correspondence with my assistant, and yes, we went round and round:

There were four more instances when we went on repeatedly about me saying it’s not copyrighted, the AI apologizing, and giving me alternative photos that are not to my liking. Finally, I just decided to ask it to generate a drawing of the shoes, and surely enough, it was the closest thing that I could post together with Gary’s article.

The final photo I put together with the article (quite acceptable, right)?

Given all the prompts I wrote and the results generated, an AI assistant isn’t always reliable. It doesn’t even generate the images with the specific dimensions I requested (yeah, I told it that the photo should be 740x440px, but it gave me the usual 512x512px).

This new experience underscores how writers, graphic designers, etc., shouldn’t be threatened by AI, let alone curse it for the rest of their lives. If you do so, it would seem like you have entirely given up on your skills and have given your power to AI, which isn’t powerful at all, as you saw.

Our CEO at Aratum encourages the creative team to embrace AI because he knows that we still have our own capabilities way ahead and are way more talented than AI could ever be.

Think of AI as your ego. It’s naturally there. You can’t let it go away, but you have the power to control it. Do not let it become your master. You are the master.

Even so, that’s not to say that AI is without its drawbacks, and yes, I have been subject to them. Perhaps I’ll discuss it on another post, so stay tuned.electricmoi

Spread the Love

Leave a Reply