I work for an AI startup, Gluino. We are literally building an AI system to help people do work. Still, I’m a bit of an old-fashioned human. I carry a pen and paper with me everywhere. One of the things I wrestle with is how to use AI responsibly and to the best of its ability to help me.

In this series of posts, I want to discuss how I’ve used AI in the past, how it has helped me, and how it hasn’t. I will also mention some interesting uses of AI that I have seen or heard about, hoping to inspire responsible use of what we call AI.

The Gemini Commercial

During the 2024 Paris Summer Olympic Games, Google released a commercial for its AI product, Gemini. In the commercial, a father uses Gemini to write an appreciation letter to an athlete from his daughter.

There was a bunch of feedback on the commercial.

The issue is whether or not an AI should be responsible for writing something like this—for writing the words of a kid instead of having the child write them. Are we losing something authentically human by using an assistant to do work that we should be doing?

I agree with most of this in general. AI shouldn’t be used to do human things like show appreciation for something or someone. AI, by nature, doesn’t have emotion.

Where I differ is my belief that AI could be used to create a “framework” for a letter if you don’t know where to start. Asking the AI, “What should I put in my appreciation letter?” is much more useful than asking it to write the entire letter for you.

How I’ve used AI

There are two prominent examples of how I’ve used AI. One was very successful, and the other was a big failure, possibly due to how I wrote the prompt.

Acroyoga Teacher Description

I teach acroyoga at the YMCA. I’m their first teacher, so they needed a description for the job they were hiring me for if they ever needed to replace me in the future.

I asked an AI to generate a job description. I started the prompt with “Write me a job description for an acroyoga teacher.” What I got back from that was about 70% correct. That’s awesome. I went through and updated the content quickly and got the job description out very quickly.

This was a very successful use of AI.

Student Recommendation

One of my former cheerleaders asked me to write a recommendation for them for a cheer coaching position in their hometown.

I asked the AI to generate a recommendation letter, hoping that I could go in and edit it. What I got out was predictably full of platitudes without specifics; I hadn’t given it any to use.

I tried to modify the letter to be used, but in the end, I had to start from scratch. What the AI generated was an excellent example of what I didn’t want to write.

I might have gotten better results by asking about the format of my letter or if there were specific areas I should target, but generating the whole letter wasn’t successful.

I was listening to a podcast this week from Hacked where they were talking about the hype surrounding AI. One of the hosts mentioned having a lawyer friend who sends in the transcripts of testimony in his cases and asks the AI to find all the inconsistencies. This saves the lawyer a ton of time and should be easy for a computer to identify.

This seems like a very good use of AI, especially if you can tell the AI to err on the side of a false positive instead of a missed positive.

The challenge here is that if you have to catch all the inconsistencies absolutely, depending on AI might not be possible. But combining large data with specific targets sounds like a great use for a computer, and using something that can understand some of the subtleties of the language is pretty awesome.

Summary

I don’t want to call this a conclusion because I’m still investigating the use of Artificial Intelligence as we use it today. These specific examples highlight that there are good uses of AI and uses that might not be appropriate. We haven’t even talked about using AI to write college essays or anything else that could be considered academic cheating.