4 ms·
I like this take on modern tech motivations. The thing that I struggle with is I agree with it, but I also get a lot of value in using AI to make me more produ
by 1dom 2y ago
I like this take on modern tech motivations.
The thing that I struggle with is I agree with it, but I also get a lot of value in using AI to make me more productive - to me, it feels like it lets me focus on producing substance and actions, freeing me up from having to some tedious things in some tedious ways. Without getting into the debate about if it's productive overall, there are certain tasks which it feels irrefutably fast and effective at (e.g. writing tests).
I do agree with the missing substance with modern generative AI: everyone notices when it's producing things in that uncanny valley, and if no human is there to edit that, it makes people uncomfortable.
The only way I can reconcile the almost existential discomfort of AI against my actual day-to-day generally-positive experience with AI is to accept that AI in itself isn't the problem. Ultimately, it is an info tool, and human nature makes people spam garbage for clicks with it.
People will do the equivalent of spam garbage for clicks with any new modern thing, unfortunately.
Getting the most out of latest information of a society has probably always been a cat and mouse game of trying to find the areas where the spam-garbage-for-clicks people haven't outnumbered use-AI-to-facilitate-substance people, like here, hopefully.
- skydhash 2y agoJust one nitpick. The thing about test is that it’s repetitive enough to be automated (in a deterministic way) or abstracted into a framework. You don’t need an AI to generate it.
- closeparen 2y agoWhile I occasionally have the pleasure of creating or working with a test suite that's interesting and creative relative to the code under test, the vast majority of unit tests by volume are slop. Does it call the mock? Does it use the return value? Does "if err != nil { return err }" in fact stop and return the error? This stuff is a perfect candidate for LLM generation.
- 1dom 2y agoI find it helpful for generating automated test suites in the style of the rest of the codebase. When working across multiple projects and clients, it reduces the mental load of having to remember or figure out how tests are expected to work in each codebase. I agree with your theory about tests. The reality of it is most code is garbage - often including my own - and in a lot of environments, the task is to get the job done in a way that fits in with what's there.
- DowagerDave 2y agoAI seems really good at producing middling content, and if you make your living writing mediocre training courses, or marketing collateral, or code, or tests you're in big trouble. I question how valuable this work is though, so are we increasing productivity by utilizing AI, or just getting efficient at a suboptimal game? I for one just refuse to play.