When I first started using GenAI for setting up my LLC, one of the use cases I wanted to try out was to use GenAI to create a logo for my business with a tag line. This was a mighty struggle and a painful process. Here is a link to my experiment back then.
Well fast forward 8 months (to the date) and I came across various posts talking about Flux for image generation. I was not sure what to expect but this time I went it with lower expectations. I also came across a Flux AI agent from agent.ai and decided to give it a try. I was testing out 2 things - how do Dharmesh Shah AI Agents work and also if the underlying Flux generator was any good.
So my first prompt, I gave it the context that my wife a writer and I started an LLC and wanted a logo. So this is what it generated -
For some reason image of an AI consultant is always a brain and the author was represented by a nib of a pen (for those that have used ink pens). This was ok. So I asked it to generate a logo and gave it specific colors (orange and grey) and asked it to have the name of the company in it. This was not something I could do before. The results were better -
Getting the text accurate in the image had been a big challenge and getting it to generate colors had been a challenge too.
I then asked it to forget the colors and generate a different logo with the company name and a tag line and it came up with this.
So what is the point of all this -
If AI can't do something we need to keep reminding ourselves that it cannot YET do that task. If the need is important enough, someone is working on figuring it out.
Being able to abstract some of the underlying complexity by using Agent.ai was also pretty cool. While this is a simple example, it shows how GenAI is going to consumed at scale in the future. All the complexity of prompt engineering is going to be hidden away from the end users.