knowledge-based programming
March 1, 2014 6:08 AM   Subscribe

 


Software development's ultimate clown car.
posted by Foci for Analysis at 6:25 AM on March 1, 2014 [10 favorites]


I wish he wouldn't gesture with his hands like that.
posted by Obscure Reference at 6:39 AM on March 1, 2014


It's like the mechanical automata of the 17th century - not trivial at all, nicely adorned, mostly useless, and one day we'll have the real thing and it will be pretty good.
posted by dhoe at 6:59 AM on March 1, 2014 [6 favorites]


This will change the world just as Wolfram Alpha has!
posted by Poldo at 7:17 AM on March 1, 2014 [6 favorites]


I'll say this much about Wolfram Alpha -- smart high school kids use it to leverage their understanding of math and other subjects in ways that are pretty neat.
posted by srt19170 at 7:27 AM on March 1, 2014 [12 favorites]


I learned how to use Mathematica about 8 or 9 years ago now. It's amazing how far it has come in that time. The language looks to be pretty much the same syntax, just with a whole bunch more models built in already.
posted by Roger Dodger at 7:28 AM on March 1, 2014 [2 favorites]


Grrr I feel strongly about this but I don't know how to express it. I do absolutely agree that this direction is great. I've used Mathematica for 20 years and it's damn handy. I love the Raspberry Pi. I teach CS students how to scrape the web, manipulate information, and produce useful outputs using Python and its many libraries, and I'd even push this as a gen-ed requirement at my big state university.

But if Wolfram keeps framing this type of activity as a new thing or FFS a new kind of thing, then it's just going to keep producing eye-rolls. Facebook didn't succeed because they impressed us with how transformative and different it was. It succeeded because it blended naturally with people's lives.

I really want people to understand that they are empowered to integrate the technologies that surround them, to serve them better. To do that, we can't stress that there's something new to be practiced and mastered. It must instead be presented as a natural extension of their everyday use of that tech.
posted by rlk at 7:55 AM on March 1, 2014 [31 favorites]


My thoughts exactly, rlk. Or: If Mathemtica is so wonderful, why do I keep reaching for my spreadsheet?
posted by RobotVoodooPower at 8:01 AM on March 1, 2014 [3 favorites]


The hype seems to be a strawman. He's not saying it will change the world or that it is some revolutionary tool that will make existing tools obsolete. It's just a cool tool. If you don't like it don't use it. Personally I think looks incredible, but no idea how much it will cost or how easy it is to create useful (not just cool) applications.
posted by stbalbach at 8:02 AM on March 1, 2014


This is the sort of thing that will seem almost magically prescient when it works, which it will over a large terrain of cases which the developers have anticipated and an impressive chunk of terrain which they haven't specifically thought through. But it will also fail hilariously in many other cases, often indistinguishable to humans from the cases where it seems to work, and you will never know going into a problem with it which set your problem is in.

"Who is the king of the Moon?" shouldn't be all that difficult for something so powerful to correctly parse.
posted by localroger at 8:09 AM on March 1, 2014


Introducing a new closed source language today is a risky move.
posted by idiopath at 8:33 AM on March 1, 2014 [4 favorites]


Is it closed source?
posted by xmutex at 8:36 AM on March 1, 2014


The specification for CDF is open-ish, but you need closed-source tools like Mathematica to create CDF files. Not dissimilar from the early days of PDF.
posted by Blazecock Pileon at 8:47 AM on March 1, 2014 [1 favorite]



"Who is the king of the Moon?" shouldn't be all that difficult for something so powerful to correctly parse.


I'm fairly well read and educated (PhD + postdoc), and without context, I don't have any idea what you mean by that question.
posted by 445supermag at 8:50 AM on March 1, 2014 [1 favorite]


Actual incident:

I had a student who I was helping with a problem, and we got to the point where he needed to do a calculation, something like "(1.731 - 1.5)/0.25". I left him to work that out and help another student. When I came back to him he had his answer. It was "0.924 x".

I boggled. x? There wasn't an x in the problem. Where had it come from? I knew that he wasn't a very hard working student, and could see that he hadn't brought a calculator like students were required. Turns out he had thrown it into some search engine and ended up at the Wolfram Online Integrator, which had evaluated his formula and integrated it dx.

This is no slam against the quality of Wolfram's tools, but it really brought home to me that the easier they are to use, the more likely they are to be used by people who have no idea what they mean.
posted by benito.strauss at 8:58 AM on March 1, 2014 [16 favorites]


I have been working towards waiting to finally watch this promotional video for the past 30 years.

How does a person become comfortable using their last name for most of their software products, programming languages, and even some select proprietary data formats? Working for Wolfram sounds much more gratifying if your last name is Wolfram. I wonder if they subsidize last name changing for their employees.
posted by oceanjesse at 9:01 AM on March 1, 2014 [5 favorites]


How much of the stuff shown in the video is actually new? I don't have access to an installation of Mathematica at the moment, but it doesn't seem like that big a step up from what recent versions of Mathematica have been capable of.

How does a person become comfortable using their last name for most of their software products, programming languages, and even some select proprietary data formats?

Eww, yes. When I worked on a small software project for my high school together with a friend, he started referring to the path description language I came up with for the project as “SPS” in the source code, which was short for “[my last name] PostScript”. Almost a decade later, that name still makes me cringe.
posted by wachhundfisch at 9:31 AM on March 1, 2014


I think the big new thing is the promised frictionless hosted deployment / service creation. It's a lucrative move if it pays off. The endgame is companies with key parts of their infrastructure running on Wolfram©®™ hardware. This is basically setting up their dominos for future rent seeking / gouging.

I was tired when I commented above. The closed source language is old, the novel thing here is the pay-as-you-go deployment trap revenue plan. Someone could of course use the quasi-open spec and try to reverse engineer an alternative, but who knows how that would turn out.
posted by idiopath at 9:45 AM on March 1, 2014 [1 favorite]


If Mathemtica is so wonderful, why do I keep reaching for my spreadsheet?

I know people who do word processing tasks in Excel.

How does a person become comfortable using their last name for most of their software products, programming languages, and even some select proprietary data formats?

I think that is explained in Strunk & White, or maybe Bartleby.
posted by charlie don't surf at 10:03 AM on March 1, 2014 [4 favorites]


This looks really awesome. But I'm not a real Mathematica user now, although I often reach for the ole Wolfram-Alpha.
posted by zscore at 10:13 AM on March 1, 2014


So, it suggested a word cloud for something when I tried alpha... and for the life of me I can't figure out how to get a word cloud out of the Gettysburg address.

(No, not this address: 1195 Baltimore Pike,Gettysburg, PA)

Google on the other hand, came up with this
posted by MikeWarot at 10:27 AM on March 1, 2014 [1 favorite]


Your degree is obviously not in MeFi studies, 445supermag: Who is the king of the Moon?
posted by eeriegongs at 10:48 AM on March 1, 2014 [1 favorite]


This is no slam against the quality of Wolfram's tools, but it really brought home to me that the easier they are to use, the more likely they are to be used by people who have no idea what they mean.

And the more likely they will create people who have no idea what they mean. I was a lazy mathematics student that was well-endowed with mathematics technology. It's extremely tempting to just give up on the subject when you hit the limits of your automated tools, because there's just too much to catch up on by that point.
posted by cosmic.osmo at 11:13 AM on March 1, 2014


But if Wolfram keeps framing this type of activity as a new thing or FFS a new kind of thing, then it's just going to keep producing eye-rolls.

I agree with almost everything you say, but I have to admit that a lot of this really does look pretty new to me. I mean, of course one could technically do all these things with existing programming languages. Technically. But at least based on these canned demonstrations, I'm seeing ridiculous amounts of power. Also, I don't fully understand how knowledge is so deeply integrated into the language itself, except to say that I personally haven't seen anything like it. Built-in support for semantics in an actual programming language could be very exciting, if that's what this is. Perhaps I'm getting ahead of myself, because my impressions are based solely on this video.

I definitely want to put some research into this, because it could actually be pertinent to my work. I have often got the impression that Wolfram has this tendency to put together some very nice tools and ideas, but make excessive claims of novelty. Alpha has always struck me as more of an interesting toy than a useful tool, and his notion of cellular automata struck me as an intriguing reframing that didn't contribute much. Mathematica, on the other hand, is great software that has served generations of scientists.
posted by Edgewise at 11:20 AM on March 1, 2014


It's odd that he calls it a language, which has a precise meaning in computer science/engineering. It's more of a really, really neat library backed by a massive online knowledge database, right?
posted by spiderskull at 11:21 AM on March 1, 2014 [2 favorites]


(I should add that you can do symbolic variables in quite a few languages, including Python if you design your classes cleverly)
posted by spiderskull at 11:23 AM on March 1, 2014


So, it suggested a word cloud for something when I tried alpha... and for the life of me I can't figure out how to get a word cloud out of the Gettysburg address.

(No, not this address: 1195 Baltimore Pike,Gettysburg, PA)

Google on the other hand, came up with this


Well for me this was the top result on Google.

And I'm not clear but it seems you're comparing whether something can generate a new thing to whether it can find it, assuming someone else has already generated it and put it online. They seem like very different tasks.
posted by RobotHero at 11:28 AM on March 1, 2014


"Who is the king of the Moon?" shouldn't be all that difficult for something so powerful to correctly parse.

I'm fairly well read and educated (PhD + postdoc), and without context, I don't have any idea what you mean by that question.


Maybe I'm not educated enough to confuse myself on this issue, but I really can't think of any context where "There is no king of the Moon" isn't the most appropriate response.
posted by localroger at 11:57 AM on March 1, 2014 [2 favorites]


Stephen Wolfram is the original grey goo, turning everything he touches into more Stephen Wolfram.
posted by benzenedream at 12:07 PM on March 1, 2014 [8 favorites]


I liked this demo so much that I did something I hadn't done yet this year, since there's no need any more. I emailed this link out to several people.

As a tech demo, it is like a music video. No second is wasted. It is fantastic.
posted by sieve a bull at 12:42 PM on March 1, 2014


I mean, of course one could technically do all these things with existing programming languages. Technically. But at least based on these canned demonstrations, I'm seeing ridiculous amounts of power.

The "rent-seeking" seems to come from joining a potentially expressive (if closed) language and development environment with backend data curated by Wolfram's company.

Basically, when using this, you're paying (through license fees for Mathematica or other components) for someone in a Wolfram office somewhere in the world to sit down and catalog data — to assign words for colors and other descriptors to world flags, for instance, or to write routines that join social network data with geographical locations or other metadata — in such a way that the functional language can quickly retrieve and operate on metadata to calculate an accurate result.

Curation of data isn't free. That's why the open/closed nature of this doesn't really faze me. This is a big deal in genomics, for instance. Here, biological experiments are expensive, time-consuming and hard, and they need lots of specialized, intelligent labor to massage data into something that can be consumed by equally-specialized bioinformaticists and their oft bespoke software. I see no problem with this costing a reasonable amount of money, in the cases where a public consortium cannot achieve the same results.

I can't see an open model supporting an equivalent to Wolfram Language until a closed-source tool is demonstrated to be successful and important enough, such that people would want to actively sponsor and contribute to an open-source project that would basically copy Wolfram's language, metadata and data. Perhaps that will happen one day, soon.

I don't know how useful this is to me in my work, where Wolfram's tools and databases are less useful, but I can see it useful to people in other similarly information-rich fields. If this project is closed, at least it is a private company that is wagering risk in its creation.
posted by Blazecock Pileon at 12:47 PM on March 1, 2014 [4 favorites]


Wolfram wolfram Wolfram wolfram wolfram wolfram Wolfram wolfram.

Really. "The Wolfram language." The man's egomania tends towards infinity over time.
posted by JHarris at 1:16 PM on March 1, 2014 [7 favorites]


Blazecock Pileon: My worry is about more than the closed language / datasets. Combine this with their cloud deployment option ("turn any program into an API instantly!") and you have a situation where:
  • the language definition and implementation is controlled by a single vendor
  • translation to another language is difficult
  • replacement of the upstream provided APIs / datasets is difficult
  • your API hosting options are limited to the same company that controls the language and the datasets
By developing your new API against a stack like that, you are building a pipeline from your bank account into Mr. Wolfram's.
posted by idiopath at 1:17 PM on March 1, 2014 [3 favorites]


"Facebook didn't succeed because they impressed us with how transformative and different it was. It succeeded because it blended naturally with people's lives. "

But then how come its interface makes me wretch?
posted by sutt at 2:10 PM on March 1, 2014


Basically, when using this, you're paying (through license fees for Mathematica or other components) for someone in a Wolfram office somewhere in the world to sit down and catalog data — to assign words for colors and other descriptors to world flags, for instance, or to write routines that join social network data with geographical locations or other metadata — in such a way that the functional language can quickly retrieve and operate on metadata to calculate an accurate result.

Yes, I agree that this is the boring part. The potentially interesting part, to me, is whether or not it provides a standard for, let's say, my company, to express its semantic data in a way that can be used with this language. I hear what you're saying about the closed nature of the platform, but I still don't to what degree it's closed (since I haven't done the research yet). If you're saying it does not provide an API for expressing my data, then it really does fall short of its potential, and in fact create a new market for more open imitators. If the idea is as sound as it looks, and as closed as you seem to suggest, that even seems inevitable.
posted by Edgewise at 2:23 PM on March 1, 2014 [1 favorite]


What's exciting to me is how much of this seems possible to implement freely (not encumbered by Wolfram's Company Store mentality) with, say, iPython/Sage and the Wikipedia datasets/namespace. The pieces are already there.
posted by kandinski at 2:35 PM on March 1, 2014


What's exciting to me is how much of this seems possible to implement freely (not encumbered by Wolfram's Company Store mentality) with, say, iPython/Sage and the Wikipedia datasets/namespace.

The devil is definitely in the details. As is always the case, the challenge is really about making all those pieces talk to each other. I'm not saying it's impossible (obviously), but it is a huge project.
posted by Edgewise at 2:54 PM on March 1, 2014 [1 favorite]


Wikipedia lacks ontology, much less coherent application of an ontology within a genre of data. The structure and content of one movie actor's wiki page can differ entirely from that of another, for instance.
posted by Blazecock Pileon at 3:24 PM on March 1, 2014 [4 favorites]


Wikipedia lacks ontology

And yet is more popular than Blazecockopedia
posted by Reasonably Everything Happens at 8:08 PM on March 1, 2014 [1 favorite]


Curation of data isn't free. That's why the open/closed nature of this doesn't really faze me.

This is incredibly true. If this language is just a platform for sufficiently well-curated semantic data, that in itself could create tremendous value.
posted by Edgewise at 9:02 PM on March 1, 2014 [2 favorites]


And yet is more popular than Blazecockopedia

And, further still, neither of us are good sources of data for what this language uses. Fancy that.
posted by Blazecock Pileon at 10:41 PM on March 1, 2014


I notice that KingOfTheMoon|Alpha has APIs for a number of open programming languages. I wonder how long those will last once this language launches.
posted by Omission at 1:40 AM on March 2, 2014


Wolfram Research only accomplishes anything because they exploit their connections with academia to chronically underpay their employees. Math and physics PhDs with software development experience routinely accept jobs with Wolfram that pay a piddling $50k without much advancement potential.
posted by jeffburdges at 1:55 AM on March 2, 2014 [4 favorites]


Math and physics PhDs with software development experience routinely accept jobs with Wolfram that pay a piddling $50k without much advancement potential.

Presumably they do not do so at gunpoint.
posted by sweet mister at 5:33 AM on March 2, 2014


A Rare Blend of Monster Raving Egomania and Utter Batshit Insanity

(this is the funniest book review ever posted on the internet)
posted by bukvich at 5:52 AM on March 2, 2014 [7 favorites]


Presumably they do not do so at gunpoint.

By that logic there is no such thing as a bad employer.
posted by Omission at 6:48 AM on March 2, 2014 [2 favorites]


(this is the funniest book review ever posted on the internet)

Wow. Yes. Yes it is. Thanks for the link. I stupidly read most of ANKOS and I had never seen this review.
posted by bh at 6:58 AM on March 2, 2014


If this language is just a platform for sufficiently well-curated semantic data, that in itself could create tremendous value.

I looked at the Financial library and basically the only thing it has is historical stock price data (daily closing price).
posted by stbalbach at 7:27 AM on March 2, 2014 [1 favorite]


An ex-academic's marketable skills are intently related to but differ wildly from their academic skills, so newly ex-academics have little clue what price their skills actually warrant on the market. You can often keep them cheap if you stick them in a podunk town with few other employers. If I were a tech head hunter, then I'd pursue Wolfram's younger employees with extreme gusto, absolute gold mine of underpaid talent.
posted by jeffburdges at 9:14 AM on March 2, 2014 [1 favorite]


Does your company offer autographed photos of Stephen Wolfram in the break room? Checkmate, recruiters.
posted by benzenedream at 9:20 AM on March 2, 2014


(this is the funniest book review ever posted on the internet)

As a side note, the author of that, Cosma Shalizi is very smart, a good writer, and has a good critical eye about data and statistics. For example, "So You Think You Have a Power Law — Well Isn't That Special?". I've enjoyed following his blog in the past.
posted by benito.strauss at 9:21 AM on March 2, 2014


cosmic.osmo: "This is no slam against the quality of Wolfram's tools, but it really brought home to me that the easier they are to use, the more likely they are to be used by people who have no idea what they mean.

And the more likely they will create people who have no idea what they mean. I was a lazy mathematics student that was well-endowed with mathematics technology. It's extremely tempting to just give up on the subject when you hit the limits of your automated tools, because there's just too much to catch up on by that point.
"

Oh god, this. I am glad that I learned calculus before Wolfram Alpha because at least when I was having to get help from my roommate I figured out how to do stuff and some of it stuck. Whereas the ODE class I took, which was essentially based in Mathematica, has not been retained at all. Part of that is that I use calculus every day and stuff from differential equations less often, but I do think there's some merit in being forced to do math problems by hand over and over until they're second nature for someone who's going to work in something related to mathematics.

I also think that a lot of people who slog through a couple of semesters of calculus get to a point where they're want to throw up their hands and say "fuck it, I'm never taking an integral again" or "I'm never going to analytically solve a large system of equations again" because hey, that's what mathematica and Matlab are for, right? Which is a different kind of problem, but if you run into it I think it eventually becomes self-correcting to the extent that you quickly run into those tools' limitations (either they're hard to get working for your particular problem or are super slow computationally and it's just easier to program in the analytic solution to your problem...)
posted by dismas at 9:56 AM on March 2, 2014 [1 favorite]


CRS On Getting Tenure: "For myself, I feel — relieved, even pleased. Relieved and pleased, but not triumphant. I benefited from a huge number of lucky breaks. I know too many people who would be at least as good in this sort of job as I am, and would like such a job, but instead have ones which are far less good for them... I even have some fear that this has deformed my character, making some ordinary kinds of happiness insanely difficult... I worry about what I have given up for it..."

congrats, i guess?
posted by kliuless at 4:49 PM on March 4, 2014 [1 favorite]


« Older The Staunton Lick   |   If Stable and Efficient Banks Are Such a Good Idea... Newer »


This thread has been archived and is closed to new comments