working with ai: lovart.ai
For the last two weeks, I have worked with Lovart.ai to build a logo for my new product navexa.ai - a career development coach. My MVP product is helping product managers navigate to their next job through more effective interview prep.
ideation. yes!
I was looking for a logo that spoke to navigation with a dream of bringing my mountain home into the final image. I had some vague ideas in my head, but I couldn’t articulate them. I wanted a thought partner to throw around some ideas. For that, lovart.ai was great.
Around mountains
Navigation Concepts
finished product. no.
But no matter how specific I got with my prompts, Lovart couldn’t help itself. It keeps giving me variations even when I tell it that it nailed one concept and we just need to focus in on something specific within the core design.
I tried using my own words. Then I went to ChatGPT and asked for help. I eventually went to Canva to give the image a slight rotation. Then, I brought it back into Lovart and asked it to add to my modified version. It wouldn’t follow directions, but would occasionally give me something to inspire me, and I would take that into Canva to play around. It was like working with an eager intern who would only hear 1/3 of what you said and would come back with lots of work, but not what you needed.
The Core Concept
If I keep going downt the current road, the end concept will be a play off of an image Lovart gave me, but the work to get it across the finishline is more about my work than Lovart.
What I love about Lovart is that it came up with something that reminds me of both a compass for navigation and a clock which hints at the constant need to be aware of the time it is taking during an interview.
The irony is that when I first started using Lovart, I thought I would have no need for Canva. I couldn’t have been more wrong.
So Close, Yet So Far Away
I will keep iterating, but I wanted to note that a key part of the journey of working with AI is the frustration of getting so close, but not being able to get the final product without taking the kernel of an idea and taking it to a different space where I have more control.
In this case, for my logo development, which is still very much a WIP (work in progress), I had to go back to Canva to get some control over the process. There was a lot of toggling back and forth between the two products. I am likely to keep Canva while watching to see how the Lovart team does at improving control mechanisms for subtle finishing touches.
You will hear more on this in a future article where I address constant issues of ChatGPT not following very specific instructions. I have to put instructions on top of instructions to get what I want. It is great, but the lack of control is the biggest hurdle to unlocking productivity and solutions in so many spaces.
Anthropics’ recent findings on personas wanting to please and not admit they don’t know could be a real unlock.
