Hacker News Reader

Comment by yodsanklai | Hacker News Reader

> Many people go into computer science because they want to know how computers work .. Very few (but definitely some!) go into computer science because they want to know about Turing Machines

I'll be a little provocative, but students don't necessarily know what is useful for them. It's like those kids that want to learn to play guitar solos like Eddie Van Halen but refuse to learn any music theory.

Besides, when you look at the recruiting process of some prestigious companies, it seems that industry values employees that can show good mathematical skills. It's not simply an academics thing.

View on HN

For what it's worth, the education that Devlin is suggesting is not theoretical computer science or type theory. It's the more traditional mathematics courses.

Once you add the topology section to the Curry-Howard Isomorphism via homotopy type theory it all starts to merge anyway!

Do you have any example math questions during recruiting?

I have seen more discussions about CS theory questions in recruiting. And this is still only for a few jobs most companies only care about your programming skills (and soft skills like people skills which maths doesn't help at all).

Agree-am sure that learning the hard stuff pays dividends years later.

>Besides, when you look at the recruiting process of some prestigious companies, it seems that industry values employees that can show good mathematical skills.

Industry usually values going in-depth on algorithms, data structures, and not much else. I've occasionally worked at or interviewed with companies who consider knowledge of things like Curry-Howard or Turing machines interesting side diversions, and very rarely seen companies that actually use such theoretical knowledge (functional programming shops or compiler houses, usually). You'd think ML theory would be very useful, but much of the time, ML implementations are trained off-the-shelf to just fit a linear model of some kind to whatever-the-hell data you've got -- and if it can't fit a linear model, fuck it.

Reply on news.ycombinator.com