DeveloPassion's Newsletter - Prompt engineering

Hello everyone! I’m Sébastien Dubois, your host. You’re receiving this email because you signed up for DeveloPassion’s Newsletter. Thank you for being here with me ✨

If you enjoy this newsletter, then become a supporter ❤️

Another week, another newsletter! I hope that you all had a great one 🤩

This week I spent quite some time playing with stable-diffusion and exploring its possibilities. My mind is still blown by the fact that the dataset is < 10GB. I also can’t wait for the general availability of the 1.5 version.

I started exercising again, and managed to hurt my back already. Go me 😂. On other news, my son Raphaël is now 1 year old. Time flies!!

I really need your help. If you get value from this newsletter, please consider becoming a supporter. You’ll be able to join the curious minds private community and meet like-minded people while enabling me to keep creating content. If you prefer, you can also buy one of my products on Gumroad or buy me a coffee tea.

Alright, let’s goooo! 🚀

During the last few months, my interest in AI image generation has grown a great deal. It started with me reading about DALL-E, and realizing how much potential this has for creativity, and was renewed with the release of stable-diffusion under an Open Source license.

Compared to DALL-E, stable-diffusion has very important advantages:

Since the release of stable-diffusion, there has been an insane amount of innovation from the open source community. User interfaces have been released, tutorials, modified versions, guides on how to disable filters, etc.

It’s really an understatement, but the pace of innovation there is astounding. And it will only keep accelerating, as more and more people discover it. People are already experimenting with video.

As a content creator, my main interest in the technology is in adding images that look a bit different from the usual stock images from Unsplah and the like. With stable-diffusion, I can easily generate images that match the content and provide a more “immersive” experience for my readers.

As a creative person, it’s also really fun to use this technology to create art and mixing different tools, techniques and genres together.

As a parent, the technology is also fascinating, both for fun and for education. I intend to spend time with my kids to show them the cool stuff we can do with technology today. I have no doubt that many tools and apps will appear to put this technology into the hands of more and more people.

One key thing to understand with this technology is that the text used to generate images is of the utmost importance. If you just use “cat”, then you’ll get a cat, that’s for sure, but not necessarily something great. Instead, you need to provide long and very clear descriptions of what you want to be generated. This requires some skill and experience. I’m convinced that this skill will become more and more important over time, as this technology matures. It might very well become a new job type in the future. Smart people have already started creating ways to share/monetize useful prompts. I also found some useful guides on the topic, including a list of open source prompts that will surely grow.

If you’re curious to learn more, do give Simon a follow, he’s got really interesting content on the topic. Also, check out the sub-reddit dedicated to stable-diffusion.

There are legitimate concerns about this technology being open source, but I am convinced that it was just a matter of time anyway. OpenAI would have been hacked at some point and their code released into the wild. I’m curious about what the future holds, but the ITSec community is clearly already facing new challenges with OpenAI and image generation AIs like DALL-E and stable-diffusion. For instance, bots using GPT-2 and GPT-3 are already being abused via “prompt injection” attacks. They’re normally “configured” to act a certain way, but well-crafted prompts can modify the behavior of the bots. “Fun” stuff 😂

I didn’t make any progress on the video recording front. Frustrating as it is, I didn’t have much free time. I decided to prioritize health and family, so the business didn’t get much attention.

I’ve started exploring ideas for a new project in the community space. I have good hopes that it might turn into a wonderful calm business. I’ll let you know more once it’s time!

This week I’m going to give one more try to the video recording, and I’m also going to write new guides for the Obsidian Starter Kit.

I’ve passed $6K revenue on Gumroad, which is really cool 🎉

Last but not least, the Google Search Console continues to give me good vibes. The upward trend continues to accelerate: