Fuck you with all the Coursera courses where we had to write 3 lines of code and the rest of the function was already there. Guess what, I have to write the WHOLE thing myself in my job and those 3 lines were the easiest part. Nobody hires me because I can calculate the dimensions of a fucking convolution layer on the paper.
Fuck your Kaggle competitions where you couldn’t win anything if you didn’t have a few personal GPUs and ensembled tens of heavy models. Who the fuck is really gonna use that model anyway? Fuck your stupid score-boosting methods that were no better than dark magic.
Fuck your research venues where professors have 5+ accepted papers for EACH conference! All the labs are turning to fucking paper factories and pressuring students to publish as it’s like laying eggs. Who the fuck can keep up his knowledge with this pace for publications? Fuck your review process that’s slightly better than random for most of the papers. People don’t change a single fucking word and re-submit it to another equally prestigious conference and get accepted. Fuck your “reproducible research” where the paper’s code repositories are unreadable at their best and nothing works anymore if you change one line (IF you manage to run them at all). Fuck your papers appendix section where the authors put all their lousy proofs hoping that nobody reads them.
A huge fuck to you deep learning for all the stupid knobs and handles we had to tune just to squeeze out 0.001% accuracy. Let’s change the learning rate and wait for 10 fucking hours to see how it goes… now let’s change the number of layers and wait another 10 hours. Fuck your novel model architectures that are only SOTA when you don’t calculate the standard deviation because the training process is taking a fucking century. Fuck your research papers trying to invent everything again using deep learning because it’s so hard to do a literature review. Deep learning for compiling source codes, deep learning for graph shortest paths, deep learning for databases, who the fuck is going to use that anyway…
Fuck you deep learning for trying to replace all the classic ML methods that were totally fine, for making people believe everything is solvable if you go “deep” enough and there is no need to learn clustering or even regression.
Oh, I almost forgot, fuck your stupid NLP puppets that no small institution can train.
Fuck you reinforcement learning with all the hype and the least impact imaginable. I’m fucking tired of seeing AlphaGo in the first slide of any AI presentation. Fuck your agents trying to beat DoTA and LoL. Ok, you can win any game you want, just don’t waste all the GPU for a single headline in the news. Nobody likes playing against a super fucking hard AI and enjoys losing over and over again.
Fuck you data science, the sexiest job of the 21st century. There is nothing sexy about your hopes of applying cool methods to data being destroyed by finding out there is no data to work on and spend the rest of your fucking career building shitty pipelines and dashboards out of nothing. Fuck you data science for promising a good career switch; anyone with the slightest interest in data science took a bunch of shitty online courses and they are applying to all the roles containing the word “data”. People don’t even know the difference between a data scientist, data analyst, and data engineer but they love “exploring data” so they apply to all of them and wait for their dreams to be crushed sooner or later.
Fuck you for all of your useless buzzwords, your harmful trends, and your false promises.
I’m just playing ML, you know I love you.
Thanks for reading :)
This post was inspired by a similar post about the startup world.