I launched the site way back in September 2015. As with most things in my life, the CSP Blog was not the result of some carefully crafted plan, such as to corner the online market on signal-processing instruction, create a side-hustle, or manage my brand, whatever that might mean. It was a lark. I wanted my wife to start a blog or website as a place to share her writing with the world. “Look, dear, it really is super easy to create your own website,” I said to her, after writing a post or two. And it really is easy.
But then the CSP Blog took on a life of its own. Or, better said, it took over my life. Certainly it took a lot of my time, and still does.
I picked wordpress.com because of equations. Typesetting equations is easy on wordpress.com because it is based on LaTeX, which I already knew quite well. On the downside, I can’t convince wordpress.com to increase my allotted storage because I’m at the top of the cost hierarchy. Offering to pay more money is met with indifference to my plight. But I manage.
For those first easy-going controversy-free years, I just wrote math posts, because CSP is a branch of signal processing, and signal processing is math. Then the image-processor machine-learners started looking up from their image classification problem to see what else they could conquer. So I got into the business of “post-publication reviews,” which are sorely needed, apparently, across all fields of science and engineering because peer review and publishing are in crisis. That led to my collaboration with Dr. Snoap and our contribution to the study of generalization (lack thereof) in deep-learning networks. But don’t forget that I also do post-publication reviews on signal-processing papers (correntropy, anyone?)
Then modernity came around and kicked me in the teeth, along with everybody else. The trainers (shorthand for the supervised-learning machine learners) decided that not only did they not need or want to know anything about “domains,” or how problems were tackled and solved using science and math, but that everything of yours is actually theirs. They will take it, copyrighted or not, and then they will pass it off as their own. This is what happened:
Continuous growth in page views (a reader-engagement proxy) until 2023. And in late 2022, the world started freaking out over ChatGPT.
Coincidence? Maybe.
So now the CSP Blog is 10. Adapt or die, I suppose. I do have some remaining advantages over AI. I always answer your questions in the Comments sections of all the posts, and I think my answers are better than any answers you are going to get out of an LLM. I think I still have an advantage over the LLMs in terms of creativity–see the Desultory CSP series of posts. I doubt the LLM would think of applying CSP to the human genome. I won’t be obsequious, fawning, or sycophantic. I very rarely hallucinate.
But I will adapt too. In the next ten years I hope to launch and flesh out two new subtopic threads on the CSP Blog: Communications ToolKit (CTK) and Mathematical Foundations of Signal Processing (MFSP). The former will detail communication signal types, error performance, error-control coding (including turbo coding and the Viterbi algorithm), channel estimation, synchronization, etc. The latter is inspired by my post-publication review of a paper by signal-processing professors that appear to hold signal-processing mathematics in contempt. It will attempt to define and study numbers, sets, convergence, continuity, linear algebra, differential and integral calculus, differential equations, and any other mathematical topic that is relevant to signal processing and that I kind of understand (so that leaves a lot of topics off to the side).
Why would I want to do all that, and also continue the CSP posts and post-publication paper reviews, in the face of the LLM onslaught? I value human intelligence (HI), is all. My promise to you, dear human reader, is that I will continue to promote HI and I hope you will join me.
Happy Birthday CSPB!
