It seems like the majority of the arguments here are about having a Tensorflow wrapper library in your language. O an not in the AI domain space but I find it intriguing.
Yeah, unfortunately symbolic AI is completely overlooked by many of these kinds of articles. It’s a shame because there are interesting concepts in that space that I’m afraid modern ML research is overlooking in preference for continued development of deep learning techniques. I would have loved to have seen a LISP mentioned in this list somewhere.
Or logic programming! Expert systems count as AI too.
I’m curious: does symbolic AI subsume logic programming? Or do they just intersect? From my (limited) understanding, symbolic AI leans on propositional logic to develop higher order systems of logic (First-Order Logic, Temporal Logic, etc).
Here are some Datalog DSLs. Or, maybe you want Bayesian inference? Try a probabilistic programming language.
Maybe it sounded like a good idea in 2019?
Also I know so many people that got on the Swift bandwagon and thought it was the best possible hammer for all possible nail-like objects. (And I’ve seen that theme for so long that I remember when Turbo Pascal was the best language for everything,)
I like Swift. It’s got a lot of really good design baked in, but I can’t honestly imagine why anyone would choose it over anything else unless Mac or IOS are primary targets.
As I mentioned above, I don’t understand why anyone would choose Swift who isn’t planning on developing a native IOS or MacOS application.
Ian (author) mentioned that he nominated Swift because of it’s TensorFlow bindings. Fits @notagoodidea’s observation neatly.