@acl Can that function plot the social network of your Facebook friends, do an EdgeDetect on it, blend it and tweet it to @wolframtap? No? Sorry, no bug fixes for you!
Also, I've said this before and I'll say it again — It seems like WRI is just peddling vaporware at this point. They announce a bunch of new Wolfram products that are "Coming Soon" and most of them are still in that phase after 3-4 years. Neural network stuff and the new additions in v11 are a joke. They work amazingly well on toy examples, but for any real world dataset, you'll end up using something like keras/tensorflow and python/scala instead of some Entity filled bloatware /rant
@R.M. Sometimes, all I really want is a "no frills" installation where apart from the "core functionality" (whatever that means) I get to pick what and what not to put in. Then, only if I try to use a function that wasn't part of my original choices, I'd get a General::nosup "The function func is not installed in your current version of Mathematica, please load the blah paclet to use it."
There's too much fluff that turns out not to be practically useful (even if it appears to be from the announcement).
I get the feeling that people at WRI don't really use Mma for "real work" anymore, at least not for the kind of things that Mma's original audience would commonly need.
There is not enough work to strengthen the fundamentals, and most new features get practically abandoned after they are added (and way before they would reach a practicall usable quality)
My 2.2 installer (will all the required associated files) is a little shy of 4 MB. So much power in that size. (No fancy graphics, of course.) Now, I am not very certain that the functionality-to-size ratio is still > 1.
To be fair, Wolfram did write a giant book 15 years ago essentially explaining why he thinks focusing development on traditional mathematics is a dead end. I recently wanted to produce a list of books and movies which contain doctors in the title (Dr Jekyll, Dr Caligari, Dr Zhivago, etc), and Mathematica made that super easy.
I don't know if WRI mining gold with those impractical cloud stuff but they should make all of that locally available so one can host,tweak and expose via their own cloud/server or infrastructure. All those channel based ping pong are of no use if it has to be mediated via an omnipresent server in the middle with the glorious name of the creator who,from V5 to V11, fed mostly junk fast food to muscle up his creation aiming for a new kind of Mr. Universe championship in technical computing
The utmost closed mindedness and fixation to gimmick and buzz words are one of the major perils that stagnates growth in this ecosystem.... :( University even stopped the license last year.
@MichaelHale Exactly. And here comes my question: What would I rather like to have? A system that is able to search Dr. X movies or a system that doesn't crash in completely unpredictable ways, has a front end that runs in a completely separate thread so that it doesn't freeze when you save large notebooks or export "some" formats, a system that doesn't crash the underlying ITK image library when you call filters in a certain order, and so on.
I gladly chose the latter and did the Dr X movie hacking for my free-time because this is not real work for me.
@halirutan I think I understand your position, but I have no idea how common that position is among all of Wolfram's customers, or where they are seeing the most growth, etc.
In my particular case, I would even be fine to have a bullet-proof core system that doesn't support image processing because then it is clear that I'm writing the image code by myself. What I cannot handle is a TotalVariationFilter that works on my small party images, but cannot be used for large experimental data. The reason for that is simple: it makes my work unpredictable. I can estimate how long I'm going to need when I implement something in Mathematica.
When a filter crashes and the WRI support asks me if I really have to use so large images what should I tell them? Yes, because I'm doing serious work, I have a shit-load of RAM, and I expect it to work. In this case, I'm then left with implementing the filter myself or calling a library which increases my workload at the very moment when I really don't it.
Exactly this is the reason for the IGraphM package of @Szabolcs. At the point where you start doing serious work, all the nice features turn out to be too sloppy implemented. And I want to point out that the Image Processing stuff in Mathematica works well compared to other features.
@MichaelHale They will do what a company needs to do: Squeeze out as much money as possible. It seems their way of doing it is to attract new users with shiny new hot features.
I don't question their choice. I just think it is the wrong path and I know a handful of highly skilled Mathematica experts that too think Wolfram is a dead end.
If they stopped spreading in so many directions I would be keen to see what happens. I talked to Leonid recently and told him that I wonder what would happen if they took all responsibilities off from Tom Wickham-Jones, who is btw a highly skilled developer and let him work on the new compiler with as much man-power as he needs.
In addition to that, take some of the staff that works on Wolfram connected devices or on similar things no one needs or can do by himself with good working core language.
But your arguments are completely ignoring Wolfram Alpha and how many college students majoring in something non-technical use it to do their math homework and want it to be able to answer questions about movies too.
@MichaelHale Students are not the ones that pay the bills. It's harsh to say, but everything analytical or graphical can be done in Mathematica and if Mathematica is going to be the next movie data base, it is unfortunately not my system of choice anymore.
I need math, I need numerics, I need graphics and visualization. If I'm able to prototype an algorithm and at the same time compile it to C-like speed, that would be a system that can be used in industry.
I honestly have no idea if the people who pay $3 for the Wolfram Alpha app contribute a completely insignificant or a huge amount of Wolfram's revenue.
@MichaelHale I'm not sure either but you don't have to forget what huge sums are needed to acquire all the data and make it online available.
In either case, it's not something I have to question. I only have to decide what I'm going to do with the situation.
We have to appreciate that a Steve Jobs was unique. What he did was not for everyone but he seemed to completely understand his user-base and was able to predict the future.
If the new users of Mathematica that want to play with the cloud stuff or on toy examples of neuronal networks are indeed producing a superior amount of money for Wolfram, then the current way is the correct one.
@halirutan According to Wikipedia, Wolfram Alpha has been used by Microsoft Bing, Apple's Siri, Samsung's S voice, and Blackberry 10. They pay their bills, I assume. Maybe there are users on these platforms that want to be able to look up movies? And why not make the feature available in Mathematica, if they have it anyway.
@halirutan About connected devices: When I was doing stuff in a physics lab as part of my education, we always used LabView. I think it was because it had the best connection to all of the lab equipment. If Mathematica would be able to connect to the lab equipment, I would have vastly preferred this. But at the time there were no drivers for it. My understanding of the devices project is that they are trying to collect those drivers.
@C.E. Yes. You might not be able to control your sensors in the lab, but now you can control a quadrocopter and fly around.
Again my point is, that in its essence the connected devices is a good idea, but is it really useful in its current stage for more that free-time projects.
