Random Musings

ChatGPT is more trouble than it's worth some times

I’m anticipating an economic downturn in the first quarter of next year and have been spending some time over the last month deciding what should be liquidated in anticipation of that.

Don’t ask me why. I can’t teach what’s in my squirrel brain. Please DO NOT take investment advice from me, because I don’t want that on my conscience.

This post is about the limitations of various AI searches. First, I asked ChatGPT to give me a list of all US Recessions, their starting months, duration, and end months.

For some reason, it wouldn’t give me data prior to 1858 even though there were at least 13 recessions between the ratification of the US Constitution and 1858. The data was interesting though with January leading the way for a total of 5 recessions out of the 35 provided.

Then I asked it to give me the percent change in the S and P 500 in the month of January going back to 1950 (screenshot below.)

Again, the data was limited and ChatGPT told me that it could only take me back 30 years. This is despite the fact that the source it quotes in its answer (StatMuse) has all the answers I wanted.

(Chosen at random, January of 1955 had a decline of 0.33 percent.)

I’m not irritated, because deep down I know I’m better off.

As is often the case, the answer to a question is far less important than the process used to get to that answer.

Manually inputting data into a spreadsheet will make me ask questions I didn’t think of and give me better analysis on the backend. I was just curious what ChatGPT would say and I found some shortcomings.

I did not test this out on other platforms which have their own disadvantages ranging from bad data to drowning baby seals in the water I just wasted for this post.