Cody Blakeney of Arcee AI shares anecdote of being told to use 'natively parallel' instead of 'embarrassingly parallel

They are doing this to me at work on Monday

@code_star i save “embarrassing parallel” for difficult parking spots

@code_star obviously people proud to have created a natively parallel algorithm

@code_star "Embarrassingly parallel" is a particularly annoying form of LLM-speak

@nik_levk @code_star https://en.wikipedia.org/wiki/Embarrassingly_parallel

@code_star Nothing wrong with embarrassingly parallel as a phrase; I’ve never encountered natively parallel once in my years of reading

@code_star imagine being called “map-reducible” 😭

@code_star When I’ve used the term to characterize some extant unoptimized implementation, and especially with non-engineers present, I’ve seen enough discomfort in their reactions that I make a point to add that it’s a technical term that specifically applies to optimizations on the table.

@code_star “Embarrassingly Parallel Problem” is a technical term. 🤨

This sounds like the time my mom yelled at me because I claimed “Computers for Dummies” was the only computer book I hadn’t read yet. I had to explain there was literally a book on the library shelf with that name. 😅

@code_star That’s embarrassing

@code_star

@code_star

@soldni banger

@code_star Never heard of natively parallel. Did they mean naively parallel? I've seen that used as a synonym for embarrassingly parallel.

@code_star The sense I get is that people hear it as insulting to the initial implementer… something like “this thing is much slower than it should be, I’m embarrassed for whoever wrote it.”