AI Progression (ChatGPT Till Today)

When ChatGPT first hit in November of 2023, I was on board.

Here’s my first post about it. At that time, people didn’t really know the difference between an AI chatbot and Google. So searching for answers was one of its main use cases. However, ChatGPT didn’t have access to the live internet. It had a snapshot of it from a year prior. So you could ask it about the Ben Franklin’s role in the American Revolution, but you couldn’t ask it what time the Eagles played on Sunday.

It was a response engine, not a generative tool.

I, however, was already experimenting with it generatively for both writing and coding. Although it felt amazing, it wasn’t very good and had a very small context.

For example, I could ask it to “generate me a bash script that will execute these 4 python scripts and put all of the output in a CSV table.”

It would give me something that I could copy out into my editor and then run it myself. It wouldn’t work right, so then I’d fix it and keep going. It was a helper. An intern who needed micromanagement.

Over the next year and a half, this was how I worked with it. Over that time, it got better and better. Larger context so it could understand larger problems and write more code and create longer technical documents. Better at writing code that works and fixing bugs when given the failing output.

It also gained access to the internet, which greatly improved its ability to help me with everyday things like “find me a train route from Frankfurt to Nuremburg on March 12.”

Then along came the agent with access to your whole computer.

Now I could say to it: “Write a bash script that executes these 4 python scripts, put all of the output into a CSV table, and you run and test it until there are no bugs. Sort the CSV table on test results and date, verify that the table is correct, and then generate an Excel file with a bar graph of Pass/Fail for each week. You can find my credentials for the Jenkins server in ~/code/.env.” I don’t even have my code editor open.

That’s a mind-blowing advance. Because AI is a useful intern. You’re the boss, and you can tell it to go do things and you trust that it does it.

Next, I’ll talk about where we are today and what’s coming next…