Is it correct to state that in 5 to 10 years we’ll see Julia as the default for new Data Science projects? (ML and statistical inference). I’m not a biggie on switching tool sets just because something is becoming “hot”. I like to start using something when it’s boring and battle tested, the youngsters can do the bleeding. But it seems like the likely candidate if something is going to displace the Python and R ecosystem?
I’m a huge Julia proponent (and before, a huge Python proponent over Matlab), but I would be careful about claiming that Julia will be the coming standard. Python has a crazy amount inertia and excellent projects still in the pipeline.
I think Julia will eventually win because writing your code in one language which isn’t C or FORTRAN is extremely productive and leads to much more composable libraries. The progress that’s been made on, for example, deep learning frameworks is impressive given the lack of massive investment from FAANG. I hope this leads to it dethroning Python, but it might not. If anything will, I think it has the best chance.
I’d suggest you give it a try anyway because its really not a hard language to learn. The ecosystem itself is quite good. The fact that you can write fast code without C extensions leaves you less dependent on the ecosystem, too.
> Is it correct to state that in 5 to 10 years we’ll see Julia as the default for new Data Science projects?
No, it is not correct to make this claim. Nothing is going to de-throne Python in the next five years and it is EXTREMELY unlikely that anything will de-throne it in 10 years. Any language that replaces Python in these tasks will need to be significantly better, and Julia just isn't that. Incremental improvement in a few areas that reek of premature optimization is not going to be a compelling argument for the masses.
The language that de-thrones Python has not been invented yet, and it will probably need some sort of hardware-coupled advance to have a chance (e.g. if the next big leap in mass-produced hardware were to drop 4K cores into a cheap SoC then a simple scripting language that handled internal data and execution concurrency might take over.) Julia is nice, but if anything you are probably going to see more migration from MATLAB and similar older dead-ends to Python over the next five years than you are to see migration from Python to Julia.
Probably yes - but 10 years is more likely than 5. Python really gained steam when it was 15+ years old. Julia has two years since its 1.0 release.
But it will eventually take over Python, unless something third comes and takes the cake before it. Julia is simply much better: More consistent, better designed, faster, more flexible, more extendable and with better tooling.
But is say most people can just wait. If Python is working fine for you, and it's not going anywhere the next 10 years, why not just wait? At that point Julia will be more mature with better learning resources and a better ecosystem. You can always just pick it up then.
