You are viewing a single comment's thread from:
RE: Machine Learning...and the Apocalypse?
Unfortunately, I think stirring and fine-tuning is popular to boost scores on common datasets. Which is fine if you are making a model to just predict that dataset, but may end up perform worse against more general applications (like reality). Lazy programming and poor analysis will end up leading us to the apocalypse. Good think sport prediction models won't become sentient. I hope.
Good luck on your data science journey!
That's actually one thing we did talk about in our classes. Not necessarily that it's poorly done intentionally, but rather companies/businesses are asking their employees with a background in programming or BI to start and run their data science division...with the assumption that the transition is seamless (and more importantly will cost them less money). But what they don't realize is that they're setting it up to fail from the start (or at best, not going to reach its full potential), because the employee goes in there, (through no fault of their own, they just need the training), hacking away at the data and potentially overfitting.
Another thing we talked about too is that sometimes, the companies don't even know what they need. My director was telling me some of his students weren't getting certain jobs because companies were seeking candidates with deep knowledge and understandings of machine learning, deep learning, AI, etc...when all they really need is just someone who understands it to the point of being able to implement it. In others words, seeking candidates with PhDs, when really, a bachelors with some ML experience would have sufficed.
I just think a lot of companies want to get on board with it, as fast as possible, and not really thinking ahead of time how they want to utilize it.