Personally I’ve been having a great time with chatGPT, incorrect answers and all. Last night it walked me through reading the UART breakout schematic for the PINE64 PineNote despite me not looking at a circuit diagram since middle school. I was able to integrate its answers with my own reasoning about the diagram, a couple internet searches to confirm the existence & purpose of labeled ICs, and what I had read with a search of the community chat history. Now I feel confident enough to place orders on a USB C breakout board, a UART dongle, and cables to connect the right pins together. Building this understanding took an hour or two when it would have been (at least) a full day’s work if I didn’t have something to quickly answer my very basic questions and follow-up questions.
I am absolutely loving this early stage of the system’s development where it has not yet been turned into a lucrative behavior modification machine. This is idle speculation but as systems like these mediate more and more of our access to knowledge (they absolutely will replace google search as we know it) the FOSS movement will become more and more identified with cataloguing & indexing curated human-written technical works. Our computing environment must stay fully scrutable to those of us who care.
I agree that the cataloguing and organising of existing knowledge, which is already important, is going to become yet more important as a result of the existence of models like chatGPT. chatGPT speaks authoritatively on lots of topics, but its output is not always accurate, which I would suggest means that double-checking information against reliable sources will become paramount.