Since some time back AI (Artificial Intelligence) has been discussed a lot, at least among my friends and social circles. Some people are experimenting with various types of AI that are completely based on text, which you can chat with or ask to write different types of texts.
Text-based AI
ChatGTP seems to be one of the more popular text-based AIs, but if you search for ChatGTP, for example, a variety of different results will come up. From models that are completely free, to those you need to purchase a subscription to use.
Opinions differ on exactly how effective AI is at answering more complex questions correctly. Jonathan Choi, a law professor at the University of Minnesota, says that AI could pass a law exam with a grade of C+, which is most likely a better result than I would achieve.
In Sweden, Jakob Heidbrink, a lecturer in law at the University of Gothenburg, has put AI’s abilities to the test. According to Jakob, ChatGPT writes linguistically like a “not particularly talented first-year student”, and the answers contain both factual errors and circular reasoning.
It is very possible that at least some of the differences lie in the language, and that ChatGPT has simply been trained more and for a longer period of time for American conditions.
From Text To Image With AI
Personally, I’m more interested in AI that generates images from text (text2img). I’ve always envied those who are good at painting and drawing, whether they work with pen and brush or digitally. I’ve always wished I had the artistic flair, but unfortunately I can barely put together a stick figure. But with txt2img AI, completely new possibilities open up for those who, like me, have a desire to express themselves with images but simply can’t.
As with text-based AI, there are a variety of txt2img AIs, and one of the most popular and developed is Midjourney. Previously, you could try Midjourney for free, but after it became increasingly popular, it has switched completely to paid subscriptions.
The quality of the images depends entirely on how good you are at describing in text what you want the AI to do. This description is called a “prompt” and can vary from one or two words to very long and complex descriptions. The image above was made by me and with the most basic description. Feel free to check out my own gallery, where I post some of my best or most interesting art.
If you don’t want to pay for a subscription to Midjourney, there are other options. While Midjourney works as an app through Discord, Stable Diffusion has a website where you can try generating images for free. There is also a mobile app that is free, but you have to watch an ad every time you want to generate an image.
Use AI Locally On Your PC
For now, Stable Diffusion is free to use on their website, but it only works as a demo. If you want to access everything Stable Diffusion has to offer, you can install a standalone version on your computer. However, this requires a fairly decent computer.
To be able to install and run Stable Diffusion locally on your computer, you need a graphics card with at least 6 GB of VRAM (preferably more) and at least 10 GB of space on your hard drive. Read about how to install Stable Diffusion on your computer here –> Stable Diffusion on your PC
Both Midjourney and Stable Diffusion use SFW (Safe For Work) filters, which means there are certain types of images you can’t generate. Midjourney uses a blacklist of words among other things, while Stable Diffusion seems to filter results. Exactly which images are blacklisted can vary between programs.
If you are a bit technical and feel comfortable with simple programming, you can change or completely remove the filter, if you choose to install Stable Diffusion on your computer.
If you liked what you read, consider to visit my Patreon and show your support for my work.
Don’t forget to sign up on my newsletter, to keep yourself up to date!