A bit of fun from ChatGPT and DALL-E

ChatGPT and DALL-E are great tools and unless you spent the last years in the vacuum of deep space, you know what they are. I use ChatGPT on a regular basis professionally and I started using its image generation features, which rely on DALL-E. The tools are great and quite capable but as always with generative AI, the users shall never forget that those great tools have zero understanding of what they generate.

One use-case for me is getting quick illustrations for blog and linkedin posts, especially for the new website my team and myself are running, which is all about aviation news - you can check it at www.hotelfoxfox.com.

Why use AI there? Because my personal collection of aviation pictures is limited, I can't afford a professional illustrator at this stage and I don't want to steal random pictures on the internet. Copyright still exists, right? So ChatGPT and DALL-E are the next best solution. But as mentioned above, you have to be careful, even if the pictures are "just" for an aviation news website. Here is a little collection of funny outputs, for your entertainment.

First of all, a query about aviation news with a cockpit them resulted in this strange 5 engines, single pilot cockpit...

Second, is this nice 747... The AI got the model right but the configuration must be quite an emergency. The main landing gear is extended but the nose landing gear is still retracted. It also feels like the right-wing flaps are lower than the left wing, but this could be a question of perspective. But what do you think of those clouds that looks a lot like jet engines? Scary...

The next one is the result of a prompt asking for an illustration of an airline's growth, as their results were pretty good. I almost published it before noticing that the left wingtip is missing. Oh, and the three engines combination too ;-)

But once again, don't lose the perspective and the big picture here. How cool is it to have software that generates pictures just on the basis of a quite generic prompt? You can even let it decide on the style if you have no inspiration. What I miss the most is an ability to re-work the generated pictures. So far, ChatGPT always comes with different ones, even with prompts like "Generate the same picture with ...". 

Will illustrators die because of people like me using ChatGPT and DALL-E? Certainly not. First, I suppose they can use the same tools to do basic designs and then work them to final products, or get inspiration. And second, my use case is for cheap pictures that I need on a DALL-E basis (pun intended), and I could never afford paying for a human graphic designer, not to mention that it would put quite a load on them.

So thank you for this, and for keeping me on my toes checking your results.

One more? Ok... here is a really nice air taxi landing in front of the Eiffel Tower in Paris, with the second Eiffel Tower in the background... Very Parisian... even the style of the buildings on the right hand side are so Paris...