It feels like we're supposed to default to asking AI for help these days
One thing that is really starting to bother me is how it feels like we’re supposed to default to asking AI for help these days instead of asking an actual human. Every once in a while, when I ask someone a question, instead of the person answering me themself, they tell me “I asked AI and got this” and forward me some AI-generated answer. I find this frustrating for two reasons, the first being that it feels like a condescending insult to my intelligence. I am perfectly capable of typing my own questions into a chatbot and reading the answers it produces. But the second reason is that if I ask someone a question, it’s because I want to hear directly from that person. I’ll often deliberately ask multiple people the same question so that I can get multiple sides of the same story.
I actually think it’s okay to consult AI. LLMs have access to way more training data than any human could ever digest in a lifetime and AI assistants can be great for finding things like files or links. Using an AI tool might lead you to read some document you wouldn’t have found otherwise or absorb some viewpoint you might not have heard otherwise. But in my opinion, AI tools should just be one member in a full chorus of voices. It shouldn’t replace using your brain and being observant and resourceful and it shouldn’t replace talking to people who are doing the same. I’m concerned that what is happening right now is that instead of adding to the chorus, we’re shrinking it, and that we will shrink it and shrink it until the only voice left is AI.
There’s this sort of narrowing of voices and perspectives that happens when we collectively mediate our work through AI. We have a tendency to believe that machines are infallible and treat them as a voice of truth, and we keep delegating more and more complex things for them to tell the “truth” about. Case in point: lots of policies are ambiguous and open to interpretation, and different people handle them differently. But these sorts of interpretations are slowly on their way out, because people are now using AI to decode those documents, and AI responses are slowly becoming the canonical interpretations.