Post #1894399

I want to express our heartfelt appreciation for this remarkable writing about LLMs https://www.scottsmitelli.com/articles/you-dont-have-to/ And it truly is remarkable writing. I think that goes regardless of your views on LLMs. The allegorical story of Lyle, doing a somewhat repetitive task, not getting projects of the ground or to the finish line, focused more on doing the thing rather than the "product" deeply relates to us as I'm sure it does to many. The guilt also of a life of unfinished projects with nobody to see them strikes at us constantly. I don't think the author's position on models differs too much from our own either. There are tasks like looking up small snippets or examples that are — for better or for worse — executed better through many models. As long as you verify these you're not fundamentally sabotaging your ability to think. I also think a lot has to be said for our position on the cost externalities. These current models, even open weight ones which do not fundamentally compare, will never run on consumer hardware. I will never be able to own this tool and quite frankly I am not interested in returning to the era of timesharing for everything I want to do. I do think the writing is appropriately harsh at people who have fundamentally incompatible core beliefs on LLM usage and its acceptability and I regret that this may potentially tear about friendships that didn't have reason to explore that difference of beliefs before and may not have even had insight into it. It is a long read but please read it. You will feel rewarded for it.