Random Data Science Thoughts
Starting points
Over time, my curiosity and my ADD combine to lead me to some interesting rabbit holes. One of those got enhanced the other day. Over my career, I’ve written thousands of lines of SQL code, so I was thrilled the other day when I found an email from Posit discussing a new project ggsql that suggests that you can build visualizations compatible with Wilkinson’s Grammar of Graphics using only SQL. I thought, as have others, of how to combine that with duckdb to create a analytical and visualization system that allows you to stay in SQL. For me, that’s a plus. I can write SQL in my sleep. I know what you’re thinking; I’m sure that some of my SQL looks like I really did write it in my sleep, but that’s not what I meant. I may try to code something in the future using a Jupyter Notebook and play with both to see what they can do for me. I mean, if you can write SQL and you can do something productive with this, then it would make sense since SQL is easier to learn than most other languages. That’s something that our new LLC could look into developing in depth.
On another note, between my CompTIA work and my regular paid gig, I’ve spent a lot of time in the AI space. I’ve recently found an interesting series of articles on Medium. The author, Akimitsu Takeiuchi writes a lot of AI articles and they have a Buddhist philosophy bent to them. I think that’s what got me fascinated with his work. One of his most recent pieces, AI Companions Don’t Need Agency to Create Attachment is a good warning read. We’re working with systems that are designed to sound human, create dopamine effects, and stimulate continued interaction to maximize revenue. I think that it says a lot about us as people more than it says about the systems that we build. I also am really interested in his memory management techniques when working with an AI. I would like to adapt it to a local AI deployment rather than being tied to Claude.
Projects
I have a pretty hefty electronic to-do list right now. I have a few book reviews outstanding. I just got a new one with Manning. I have done a prior review of this book. This would be the final proof review. Personally, I wish I got credit in the book for it, but I don’t think that comes with the process. We will see what happens next.I have a new website to work on for our LLC. Dunsworth, Mann, and Associates, LLC I know that it isn’t live yet, but when it is, the link will take you there.
I also ran across an interesting article about time series forecasting and how models are not tasked correctly, by the author’s opinion. In Why Most Time Series Models Fail Before They Start, Fatih Tüzen discusses his procedures to dive into the data and ensure that forecasting models behave as expected and give honest outputs. I think that I want to try the technique with 9-1-1 centre volume data. I believe that it will be a challenge to the workflow, but that’s what I want. He wrote it the code in R, so I’m really stoked to see that. It reminds me that if I had nothing else to do, I wouldn’t mind trying to port timecopilot to R if it were possible.
Other Thoughts
So, about that LLC I referenced above. My wife and I figured that we would start an LLC so we can offer analytics, research, and other statistical processes to 9-1-1 centres around the country and, wishful thinking, around the world. We’ve been discussing it for a while, but there was always a reason to delay. Things have come together and now we’ve started the venture and I’m hoping that someone reading this may believe that their centre could use our services.