My thoughts on Gen AI, part 1

ChatGPT, Claude, Gemini, DeepSeek, Qwen - they are all incredible right now. It's easy me to accept that this is how things have always been, but we're only seeing about 4 years of public projects in the generative AI space.

I don't know how things will shake out or progress in the future. Maybe there are limitations that we will hit. Or maybe we're marching quicky toward an AI singularity event that changes our society beyond imagination.

I do have some initial thoughts:

• The generative AI models are incredible. I've been using them broadly and it has greatly sped up my output.
• GenAI allows you to become a relative expert* so quickly. For example, I have the capability to build a machine learning model on a set of data, but I don't ever do it so my experience is little. Taking the theoretical knowledge and using it with GenAI allows me to build practical applications.
• One example is building plugins for my TRMNL device. If I were to go through all the documentation, I am confident that after a week of tinkering I could build a working plugin. But with Claude I was able to build a plugin in an hour.

That said, I'm nervous when I say that GenAI allows you to become a relative expert so quickly. There could definitely be some Duning-Kruger effect.

Also, the knowledge is not transferable. What I mean is that if I build a plugin with the help of AI, it doesn't enhance my experience to build a plugin. I will not be any quicker building one on my own after building one with AI (with the exception of having a realworld example to work from).

The takeaway for me is that genAI is an incredible tool, but the foundations need to be strong. There is a saying in data analytics: garbage in, garbage out. It refers to when bad data is used as part of analysis.

With GenAI, the data pipeline coming in is my own brain, so if I don't constantly learn I won't be able to use GenAI to it's potential.