2 years ago, when ChatGPT was had only been released to the public for a few months, I decided to put it to the test. Could it program a whole orchestra season? Even though it was the early days of consumer AI, I was surprised it managed to do it, even with the odd issue. Now, with new models available and 2 years of improvements, I decided to see if it could do any better.

The prompt I gave was the same as before:

Oh, and I even asked it to make the thumbnail for this blog when it was done.

Last time, I was blown away by the fact it actually came up with something workable. It gave some interesting repertoire choices, concerts were probably around 90 minutes, and it included concert names, family and movie concerts, and a concert “celebrating women composers”. However, it was also pretty unforgiving in brass sections for some concerts, and having one concert for women composers meant excluding them from every other concert. It also wasn’t a smooth process, with ChatGPT’s 3.5 model crashing several times while creating the season and slowly writing out its response.

2 years on and using ChatGPT 4.0, results were similar but a LOT smoother. Zero crashes, and an almost instant answer was given. As well as the title of the concert, there is also a one line synopsis. Again there is a concert celebrating women composers that then excludes them from the rest of the season, despite the synopsis for this concert saying that it is “Highlighting the contributions of female composers, showcasing their musical voices in a predominantly male-dominated field”. It rightly found an issue, but didn’t address it. Also a concert called “The Spirit of Africa” with no composers from Africa is unthinkable.

By using a less open ended prompt and further tweaking, I’m certain the response could be improved.

Below is the full unedited response from Chat GPT. I’ll let you decide if it’s done a good job or not, and I’d love to hear your thoughts!