I agree wholeheartedly with the premise here - learn different languages to expand your mind - but some details have not aged well.
In particular, today I would strongly recommend Python over Haskell for machine learning or natural language processing tasks. Not because Python is a better base language for these sorts of things (it isn't) but because the library ecosystem that surrounds it is so much larger. Python is basically batteries-included for ML/NLP these days.
I think he was referring to the author saying Haskell would be his choice for machine learning: "I don't do a lot of artificial intelligence, natural-language processing or machine-learning research, but if I did, Haskell would be my first pick there too." In the real world, python dominates the machine learning space.
> Suffice it to say no one has really explored doing ML in an ML. :)
Learned Standard ML in college. Never did anything with it in the corporate and don't know anyone who has either. Though I think it is something everyone should learn. One thing I agree with the author: "Standard ML was the first functional language I learned well, so I still remember being shocked by its expressiveness." If you come from C/C++/Java/etc world, just pattern matching by itself is mind-blowing.
Ok, that’s just weird and comes completely out of left field. I just saw ML and assumed it was a reference to the language since Haskell couldn’t possibly be useful in machine learning. Alas, it’s in the second sentence.
The article says "I don't do a lot of artificial intelligence, natural-language processing or machine-learning research, but if I did, Haskell would be my first pick there too." No explanation, the article goes on to explain purity and lazy evaluation that as far as I understand don't have any connection to machine learning.
To be fair, the precedeeing sentence doesn't read like the author is about to give an argument about why Haskell is good for "artificial intelligence, natural-language processing or machine-learning research".
It just reads like he merely expands his preference of Haskell for what he does, and just mention in passing how he thinks it suit him also in those other domains.
In particular, today I would strongly recommend Python over Haskell for machine learning or natural language processing tasks. Not because Python is a better base language for these sorts of things (it isn't) but because the library ecosystem that surrounds it is so much larger. Python is basically batteries-included for ML/NLP these days.