Generated by Ideogram using the prompt “White Typography, "The Lord is a refuge for the oppressed, a stronghold in times of trouble." - Psalm 9:9 Scripture Quote Over Background Photography of the sun rising over a Mountain., photo”

Basic AI News As of September 2023

As you may already know, AI technology has been exploding recently (not literally). With ChatGPT hitting one million users in five days, and currently sitting at about 100 Million, AI seems to be growing too quickly to keep up, but I’m here to help.
First, I’ll briefly summarize how Generative Text models like ChatGPT work. Have you ever been sending a DM to someone, and you got suggestions to finish your sentence or predict the next word? Well, that is essentially what ChatGPT does. When you give it a prompt, it calculates what the first word is likely to be, and then the second, and so on. This is an oversimplification, but it gives you a good idea of the basics. This is why it sometimes makes things up. This is called hallucinating. It doesn’t really know what it’s saying, because it’s trying to give you something that sounds right. We’ll discuss how to stop this in a later issue.
Generative images work very differently. Most of them work off of Diffusion Models. Sometimes an image file can be damaged or corrupted, and we see that as static or noise all over the image. Well, these diffusion models basically were taught to de-noise those images. Then they took it a few steps further. What happens when you de-noise something that is just pure noise? That’s how they work. Your prompt tells the model what is supposed to be under all the noise, and the model cleans up the random set of pixels to be more and more like your prompt until it’s done. This isn’t the only type of model for making images, but that is how most of them work.
AI is still advancing rapidly. Image generation just took another leap forward with the release of Ideogram AI, the first image model that can generate consistent text and good image quality. Other models could write text before, but their images were practically useless.

ChatGPT is a useful tool, but you need to know how to use it right. That is where prompting comes in. You can go to ChatGPT and type in, “Write a poem about pizza,” and it will, but your results will likely disappoint. A better prompt would be, “Write a 250-word professional poem about pizza in the style of William Shakespeare.” Feel free to try both of those out at Sign up and compare results. If you don’t want to come up with a prompt yourself, there are tools to improve your prompts before sending them. Bye.