Dear tech companies: Please stop putting words in my mouth
Tech’s endless attempts to apply neural networks to writing are an utter failure to understand humanity and address real problems
If there’s one thing that’s stayed consistent throughout my entire time in the tech industry, it’s been the relentless effort on the part of engineers to fabricate problems in order to justify spending time on programming challenges they find “interesting.” Just because a problem exists in the programming world does not necessarily mean there is a real-world analog for that problem. Yes, writing algorithms can be fun, much in the same way that solving a Rubik's cube can be fun, but just as a solved Rubik’s cube has little value in the real world beyond the mental stimulation of solving it, it’s silly, and potentially problematic, to assume that solutions to novel engineering problems must necessarily have real-world applications.
One such problem that, for some reason, seems to appeal strongly to engineers is sentence completion. I suspect this has something to do with the fact that neural networks have allowed for more robust pattern matching than was ever before possible, and that this trend is simply programmers curious to apply these algorithms to anything they can think of, but regardless of reason…
