Post #2667671

Another colleague asked me for my views. He basically agreed with me, but worries he'll be left behind if he doesn't use it (even though he agrees that even if it produces working code, it will be garbage in structure and unmaintainable, and maintenance is by far most of what we actually do in software engineering, not developing new code). I commented that I'm actually an extremely stubborn person. I've got hard moral lines that I will not cross, consequences be damned. I'm like this about lots of aspects of my life. Looking back, I've got really, really practiced at boycotting things I consider unethical to the point where it's second nature. So, I explained, *even if the fucking thing turns out to be necessary, I'm still not going to use it!*. He liked that, and said to stay strong. It might be wishful thinking, (it's probably wishful thinking) but I do believe that sooner or later the consequences of the shit code will come back to bite us, and we'll turf the whole endeavour. Hopefully sooner. Our problem isn't that we need to produce more shitty code faster, we need to produce better code. And the thing fundamentally cannot do that, because it's just a statistical autocomplete for what symbols look like they should go in the right place; it can't actually reason the logic of a problem. The hell with it. #LLMs #AI #GrandTheftAutocomplete