There is a tendency to picture data scientists as hipsters. The consciously cool kids chasing after the latest algo the way brooklynites chase after the latest band.
There is nothing wrong with hipster style, even bankers recognize that colourful t-shirts and impressive beards correlate with technical skill.
However, falling into the hipster self-image trap is dangerous if it impacts your focus. As Peadar Coyle points out chasing what is “in vogue at the moment … [can] make you use the wrong tool for the job” or “obsess over model performance without realising your job is to add value”.
The distinction that matters is not hipster vs. button-down style. What matters is whether you are an artisan content to craft abstract models at the lab bench or you are a practical industrialist who takes machine learning out into the messy world to delivery real value.
Part of dealing with our messy world is communicating with the stakeholders whose adoption will make or break your value creation. In the model performance vs. system utility trade-off we should fall on the side of utility.