But in the end I don't think it actually enables us to do anything that we haven't been able to do before, so it's probably not as significant as Deep Learning in that respect. To be completely honest, I am not convinced it has huge practical application, as I don't think it will ever be picked up by industry in its current form.
The reason I don't think it will be adopted by industry is that a company's concerns are typically not really aligned with a programming language researcher's concerns. For example, a company probably wants a language that is easy to hire for and easy to train for, and Haskell is definitely neither of these things. Also, from what I've heard, it's very difficult to debug and benchmark Haskell programs, perhaps due to its laziness.
Take all this with a grain of salt though as I've only been learning it for 1 year and read the majority of two books on it; I'm far from an expert.
More generally, learning Haskell (and to a great extent any programming language) is about a different way of doing things. Deep Learning is a thing to do. (It can also be a way of doing other things, but my point is that they're at different levels.)
With that said, your time is limited. Studying Haskell, Deep Learning or Statistics (an option that is criminally underrated) is a choice you have to make, and Haskell isn't necessarily the best use of your time.
If you feel like dipping your toe in the water, though a) install stack https://github.com/commercialhaskell/stack b) try doing this course (http://www.seas.upenn.edu/~cis194/spring13/lectures.html) and _do the homework_. If you find you're enjoying it, you'll have your answer. If you're not enjoying it, it's probably not worth your time.
And for the record, I have no idea what some of the top-flight Haskellers are talking about either. But maybe one day I will.
That's a great question, paulsutter. The great computer scientist Peter Landin in 1965 asked a similar question about the future of programming languages. To answer the question, he imagined a wonderful futuristic programming language with lambda calculus at its core. He named that language ISWIM, and described it in a paper entitled The Next 700 Programming Languages.
Haskell is essentially ISWIM with a Hindley-Milner type system added to it.
https://www.cs.cmu.edu/~crary/819-f09/Landin66.pdf
Haskell is science fiction from 1965. The future that Haskell brings is right now, the 21st century future.
Yes. People will probably still even be writing in C and C++ in 30 years. The same predictions were made about "4GLs" 30 years ago and have not come to pass.
High-level programming has been around for 50ish years, so we probably won't get rid of it for at least another 50.
If the implication is that machine learning will make human programmers obsolete, then no that won't happen. What will happen is that more programmers will be spending more time tweaking, training and debugging/reverse-engineering machine learning systems.
I don't see the connection between Deep Learning and removing humans from the software engineering process, so yes, humans will still be coding in 30 years.
(and some of it will still be done in C, and maybe even in COBOL)
But, if I look at Deep Learning, I see a freight train to the future. If I look at Haskell, I see a lot of... well, debating about... things.. well I just don't see where it's going. I get the sense that it's a gleeful little clever club that exists as an alternative to doing actual work.
And, where's the counterexample, Haskell is a freight train to where? What's the future that Haskell brings?
This is really a sincere question. I've been programming long enough to feel the limitations of existing tools. But, will humans even be coding in 30 years?