Hacker News new | ask | show | jobs
by vkk8 1564 days ago
Data science is an oversold profession. When studying it, you learn really cool things; Bayesian models, reinforcement learning, AI, all the trendy stuff. In reality, most of the data scientists end up spending almost all of their time interacting with horribly designed, decade(s) old databases and the analysis part of their job is mostly just plotting the mean as a function of time and doing logistic regression over and over again.
3 comments
jstx1 1564 days ago
Btw this isn't a problem specific to data science, it's how education works a lot of the time. Getting a computer science degree, you'll learn a lot of cool things but then a lot of your day as a developer will be spend gluing framework APIs together. In a medical degree people learn tonnes of interesting and advanced stuff but lots of doctors spend all of their day entering data into a computer, writing the same prescription for antibiotics over and over again and telling people to rest.

I feel like data science gets called out disproportionately for something completely normal.

link
boa00 1564 days ago
I can relate to this. The process of studying ML algorithms, how and why they work from the mathematical standpoint was really fun. But the process of building models themselves seems quite boring to me
link
goodpoint 1564 days ago
Spot on. "Data science" and "Machine learning" are indeed hyped up terms.

Most of the work is basics statistics, some SQL and poking at a ML library until you get the output you wanted.

Unsurprisingly, the day to day work can be very underwhelming.

link
princeb 1564 days ago
i spent some time in "data science", back when it was just called "research", in a corporate environment.

95% of your value as a "data scientist" is figuring out how to squeeze diamonds from tons and tons of steaming piles of sh*t, and that kind of work is basically just filtering, collating, collecting, organizing from disparate data sets with different fields, data types, formats, some of them with varying degrees of accuracy. it's not fun, but very few other people are doing it, and the result is valuable to a lot of people, so it pays well.

link