Something from work I noticed today that will unfortunately become the norm.
If you gibbity an answer, you’re regarded as smarter than someone who searched it. The funny part is I’ve actually given it a try, and it’s wrong so often that it actually wastes my time. And, if it’s right and I don’t agree, I can say it’s wrong, and with a 98% confidence level it changes the answer.
Side note, this is exactly why the fascist tech bros are pushing hard to shove this tech down our throats.
I find in my line of work that if you ask it "101-"grade questions, like the kind of questions a beginner who knows next to nothing would ask, you’ll get acceptable (not good, but acceptable) answers from LLMbeciles. As soon as it requires any degree of serious technical know-how, or as soon as context starts to matter, the utility plummets dramatically from “acceptable” to “what the actual fuck are you thinking using this!?”
The problem is that asking those 101- questions of your coworkers gets you acceptable or better answers more quickly, and often, as a bonus, stuff that has examples straight out of your workspace. So LLMbeciles are kind of worthless for beginner questions too.