About my own interest in the current generative AI craze

I just wanted to clear up some confusion about my own interest in the current AI “craze”.

1. First, let me point one thing: so far I have not done any fundamental work in the generative AI, i.e. I have not been engaged in building or training of generative AI models. There are some Bio ML projects that I am involved with that may lead in that direction, but I don’t consider myself primarily a researcher in this field.

2. My primary interest in generative AI stems from one of my long-term “other” interests: the so-called “future of work”. This is something that I have for a long time thought about, explored, read on, and wrote about. It’s a topic that’s close to my heart, as it directly affects my everyday work and life, and what I want to accomplish professionally. My career path has been anything but straightforward, and most of the time I had to either improvise or find alternative ways into the professional life. Sometimes way, way, way out of the box.

3. In that regard, I believe that generative AI tools have an enormous potential to transform how we work and build our professional standing. The little that is already possible to do with them is amazing. And what is just beyond the horizon will make all of our heads spin. I have much more to say on this topic, but this is not the place to expand on it. I am collaborating on a small ebook on this theme that will come out in a few weeks, so hopefully I’ll cover much more of that material there.

4. One particular area where generative AI is making an outsize impact is creative work. I have a creative side that has for the most part languished. It’s virtually impossible to do anything creative with so many other demands on my time. I am seeing a lot of opportunity to finally start dabbling in my creative pursuits more.

5. The impact of the current generative AI on tabular data ML is so far minimal and largely indirect. I’ve seen works that write better SQL queries, some solid but basic Copilot code generation, and some decent generic ideas for feature generation. But nothing that is having an outsize impact - yet. Part of it is probably the same old disinterest in these sorts of problems that’s prevalent in most of the ML world. And I don’t have any sense if anything profoundly impactful will change here any time soon, but I think that it would be possible to make a more advanced AutoML solution with relatively straightforward implementation of the existing and the soon-to-be released tools.