23
Heard a dev at a coffee shop say LLMs are just fancy autocomplete and it got me thinking
I was grabbing a latte downtown last Thursday and overheard two guys arguing about AI. One guy said LLMs are nothing more than fancy autocomplete and can't actually reason. That made me step back and look at how I use these tools at work. I realized I treat them like a brainstorming partner, not a thinking replacement. Has anyone else found that mindset shift actually made their output better?
3 comments
Log in to join the discussion
Log In3 Comments
emma_clark1mo ago
Hold up, I gotta push back on this a little. @jessica_ross38, I hear what you're saying but I think we're giving these things way too much credit. Calling it "fancy autocomplete" is spot on, but that's actually the problem. Autocomplete is predicting what word comes next based on tons of text it's seen, not based on any real understanding. When I asked an LLM to "show your work" on a simple math problem last week, it gave me a totally wrong step-by-step but a correct answer by pure luck. That's not reasoning, that's pattern matching. Relying on it as a brainstorming buddy can lead you down wrong paths if you're not careful, and I've seen people get so used to its suggestions they stop thinking critically. It's a tool, yeah, but treating it like a partner blurs the line between your own thinking and its guesses.
5
jessica_ross381mo ago
That's exactly how I see it - LLMs are "fancy autocomplete" but that's not an insult. It's more about what we do with the autocomplete, you know? The whole brainstorming partner thing clicked for me too. When I stopped asking it for answers and started asking it for options, my work actually got better. It's like having a rubber duck that talks back (but sometimes gives terrible advice).
4