Hate to say it, but I'm using "AI" chatbots (mostly copilot) more and more. Not to do work for me and I'm not uploading stuff to it - I use it just to help me remember stuff I didn't do for a long time.
Like specific configs, commandline one-liners, jumpstarts for scripts etc.
And it makes me wonder: because of my experience, I see a lot of mistakes, insecure recommendations, weird shortcuts it does and can correct them on the fly, but what about people who do not have this experience to compare with at all? For example, when using some niché tech, questions to the chatbot will deem much worse results because it was not learnt on it - but I can very easily overcome it by asking questions about similar, more widely adopted, tech - and improvise from received answers. Would a junior person with no experience do it? And even worse, I'm afraid of the day when I start finding more and more software run and developed by inexperienced juniors whose only learning path would be based on conversations with chatbots 🙁
Because no matter how good would be the training set for any model, these chatbots are not sentient, they don't understand - they're just remixing stuff and returning most probable word after word after word. I'm also afraid that after some time, it will introduce a large "stagnation" by not recommending stuff developed recently, because it has not yet any information about it, which will further amplify said stagnation.
=> More informations about this toot | View the thread | More toots from kayla@eilhart.cz
text/geminiThis content has been proxied by September (3851b).