AI is improving significantly, and the last year I’ve noticed huge improvements, for example in image generation. The opportunity to use AI has become more prominent across tools I’ve been using for years, such as WordPress.
I just wrote a post on WordPress about my Dutch partner applying for a visa for our visit to Australia. I clicked on the “generate image”.
Note that I did not write this prompt, it pulled it from the content of the post.
Spot the mistake/s? I find it a bit like the game I used to play as a child comparing 2 images which had a few (deliberate) differences. But here, I’m spotting the differences between the article I have written, the auto generated prompt from the WordPress AI, and my own brain and knowledge/experience.
Visa versus passport: So the prompt says Australian visa” – but here this is clearly a passport
Spelling of Australia: incorrect on the title
So these mean I cannot use this image. I thought I’d try again, by simply changing the word “Australian visa” to “Dutch”, with the logic that perhaps it will change the passport to a Dutch one….
Indeed, that’s exactly what it tried to do…with very interesting results, but again, an unusable image.
So I tried again – I reduce the prompt to just one short sentence.
This result – again amusing, but unusable. What on easy is the windmill doing in the air. And what happened to the right hand part of the Australian flag?
Overall, AI is still learning. It’s fun to experiment, and teaches us to be very clear in our communication and prompting – but even then, it can result in mistakes.
It’s certainly improving though and in many cases now, it’s getting harder to tell which is an AI generated image.
In the end, I simply found an old photo I had made and used this:
Zestee 