In the tenth grade I asked for a single Christmas present. At nearly thirty dollars it would be larger than normal plus there would have to be an upscaling of my sister's present. I wasn't hopeful, but it actually happened. A Pickett N4-ES slide rule. I spent an enormous amount of time playing with it. A few things come with heavy slide rule use: the ability to do estimates in your head - particularly involving orders of magnitude and an instinctive feeling for how logarithms work. Both of those skills are still with me although I rarely use a slide rule these days.1 There are other tools that you develop a feeling for through a lot of play - Mathematica and Maple come to mind.
We all use tools. They're often connected with and change with technologies although some of the simple good ones never seem to change much. It's interesting to sit down and make a list rating them on a love/hate scale.
Some of us build our own. Engineers excel - a good friend was working on a certain class of problem in computer science when he started at Bell Labs. His director took him aside and suggested maybe he should create a computer language to give himself the tools to work on the other problem. It turned out building and improving the language became his primary focus and a much richer path than the original problem area.m That language is a tool used by millions of programmers every day.
And there are tools that have caused frustrations to millions. Too many user interfaces are great examples of not understanding fundamental issues. Automotive touch panels are a favorite example of mine. They don't provide feedback and require too much attention from the driver. I was delighted to learn cars with these interfaces won't get good safety ratings in the EU beginning next year. Perhaps we'll see a some thoughtful physical interfaces.
In physics Feynman diagrams are important. In the late 40s and early 50s he and a few others revolutionized the field with quantum electrodynamics which led to more general field theories. The most accurate theories in all of science describing much of the universe. Feynman was disturbed by a few issues and wanted something that felt more Newtonian. He came up with a way of describing interactions that could be summarized with what looked like doodles. The approach wasn't the right approach, but the diagrams provided a way to make calculations much simpler and to think about underlying processes. Simple calculations could be done thousands of times faster and more accurately and, more importantly, the technique shifted how people think. Frank Wilczek wrote an excellent description in Quanta.
Of course computational hardware and software of all types have become important tools individually as well as embedded in complex systems. We tend to find those tools that suit our own needs. What you find useful and even delightful may not be on my radar and visa versa. I tend not to recommend much to people when I don't know their use cases - which leads me to the reason for this post.
Several people have asked for primers as well as my take on generative AI (which many simply call AI, although that's inaccurate). I'll skip the primer - you can find those in a hundred other places, but I've had a bit of experience. In the past year I've served on two committees that made recommendations on gAI student use in college. I think we made some good progress, although it's clear there will be changes over time. I've also used several of the tools - free and subscription - and have formed some initial impressions. I'll break them down into useful, take it or leave it, and not worth the effort.
Useful
Non-technical translation. Corresponding with people who don't speak English and reading posts and articles. There are mistakes, but there's enough context to get the gist of what's going on
Scraping text and even equations from images. (the text part is built into OS X and iOS)
Transcription .. some errors, but again context saves you
Entertainment - I've seen people lacking artistic skills have a lot of fun making images. The same for non-writers looking for frameworks to start.
Some of Apple's local gAI looks useful with the emphasis on local, but that's still a work in progress and I have no experience.
So-so
Formatting LaTeX .. it's far from perfect, but save a bit of time.
Translating literature .. something is lost over a good translator. There's also a good deal of sexism and racism
Playing with ideas as a part of learning. I haven't used this much, but ran into it several times in gAI student use. I'm on the fence, but it might move up to the next level.
Not so good
Search - what were they smoking?
Technical programming.. it takes more work to debug and integrate with existing software than writing from scratch.
Summarizing anything technical.. it's often difficult to find the mistakes as it "lies" smoothly
Physics and math problems. It gets well-known examples, but beyond that is wrong
Re-writing pieces.. some love this, I found it a big fail.
Some bad experiences with gAI chatbots
A number of ethical and energy use/pollution/privacy issues. I think there are large risks associated with LLMS . not super intelligence in the guru, but serious risks now.
I'm no longer paying for a subscription or actively using services like ChatGPT. I'll be following along, but I wasn't able to find a tool worth my time on a regular basis .. at least not enough to counter the ethical issues I see. Of course I recognize everyone will have their own cases. And back to the slide rule - to find if the tool really fits, you have to spend a lot of time playing with it!
I find a few other flavors that come under the heading of AI extremely useful! I'm very curious how many people will pay serious money for gAI tools?
__________
1 I'm of the age where the best things I've owned need to go to close friends and old students. The slide rule went to a student of mine under the condition that she learn how to use it. She did and still does.