| Author sounds like a young person who feels like he's a god among men just for the fact that he's implemented the algortihms and understands the math and engineering behind the libraries most DS's just pip install. Which is weird coming from a generation of devs, where actually doing this work yourself was the norm. As for DS, from what little I've experienced from the field, he sounds right. Most people come in without a mathematically rigorous education, they talk fancy, but what they end up doing is pulling in dependencies from a pre-written library and using those without understanding the theory behind them. They also ignore the fact that 99% of the value in data science is created by taking good data, understanding the domain, in which case fancy algorithms are unnecessary. And the acquisition of said things needs good data engineering, not data science. But more often than not, the credit and prestige goes to folks who pull in fancy ML algorithms and run extensive experiments and build massive ML pipelines, feeding in truckloads of tangentially relevant data. |
I almost laughed out loud when he said he started working as a data scientist in 2019. Five years is not a very long time. And he claims he already had identified the entire field as full of fraud in the first two years of that!
I agree with a lot of the article's points, but the author took a serious credibility hit with me after asserting that two years of from-scratch experience is enough time to evaluate an entire subfield of computer science.