Recently I have been working with LLMs and it's really heavy on building natural language text queries to get it to do what you want. It's mainly about injecting blobs of text to make it change its mind. Also what I find funny is that when you ask it to first think through the execution plan and you also append "please", then the results are much (much!) better.
I think it's a good thing that natural language takes over too.
It's another basket of skills that we can offload from human understanding. Let the AI we create learn the technicals and get good at interpreting our own language.
It's the same way a calculator works. Takes care of the nitty gritty work for us based on our inputs. Calculators can easily perform complex tasks that we don't need to understand.
It makes coding, and thus creating, more efficient and accessible to everyone.
reply