‘Insanely more ambitious’ than Google knowledge graph
November 30, 2013 7:31 AM   Subscribe

Not satisfied with Rule 110 - the first Turing Complete cellular automata (as proven by Matthew Cook), Wolfram Mathematics has announced a system that is “Making the world computable..." A combination of natural language recognition, data description and computational language, the goal is to allow a person to define what they want, not how to get it and manipulate it. Wolfram describes the process as follows: "A human defines what the goal should be, and a computer does its best to figure out what that means, and does its best to do it..." Of course, the rest is left as an exercise for the reader...
posted by BillW (37 comments total) 26 users marked this as a favorite
 
They've already used it to figure out how to monetize it. The answer was to monetize it by using it. To monetize it.
posted by clvrmnky at 7:49 AM on November 30, 2013 [5 favorites]


What could possibly go wrong?


Human: "Computer please solve problem X." (X being a social/political problem like peak oil.)

Computer: "Calculating ...
...
...
...
...
Solution obtained: kill all humans."
posted by oddman at 7:51 AM on November 30, 2013 [10 favorites]


Auggh my eyes have rolled out of their sockets.
posted by XMLicious at 7:54 AM on November 30, 2013 [15 favorites]


Has the Wish spell from D&D taught us nothing?!
posted by GameDesignerBen at 7:58 AM on November 30, 2013 [7 favorites]


paul kedrosky called the PR: "One of most entertainingly hubristic product semi-announcements ever" :P hopefully it can live up to watson!
posted by kliuless at 8:00 AM on November 30, 2013 [2 favorites]


Auggh my eyes have rolled out of their sockets.

Well, the man did invent a WHOLE NEW KIND OF SCIENCE... seems legit.
posted by jeisme at 8:01 AM on November 30, 2013 [6 favorites]


This amazing hype is beyond any self-promotion that's ever come before it; in fact, it's going to change the way that we think about hubris entirely. "Those other world-changing people aren't changing the world enough," said Wolfram, while bathing in a tub of money that, while large, is only a fraction of the size of the swimming pools of money harnessed by Google. "They just aren't as ambitious," he added. He then demonstrated with an entirely uncooked example, by typing in some words and making the flags of south america appear. "In conclusion, I am the king of the moon!"
posted by kaibutsu at 8:04 AM on November 30, 2013 [62 favorites]


The best bit about the PR is the ' How could I have been so stupid as not to recognise my own genius?' shtick. As for what all this actually is... Well, Mathematica was useful, his science was not, and I don' t know about Wolfram Alpha. He's combining the IoT and AI, both of which are sacks with plenty of vipers. I'll wait for someone else to take a peek before sticking my dick in.
posted by Devonian at 8:07 AM on November 30, 2013 [8 favorites]


I'll wait for someone else to take a peek before sticking my dick in.

Can we use better language than this? Gross!
posted by oceanjesse at 8:23 AM on November 30, 2013 [6 favorites]


I find it ironic that Mathematica has a godawful programming language....
posted by kaibutsu at 8:27 AM on November 30, 2013


Apologies, it was inappropriate. The alpha-male testosterone OD from the PR got under my skin.

Please read 'nose' instead.
posted by Devonian at 8:59 AM on November 30, 2013 [2 favorites]




Seems like a cool idea
posted by stbalbach at 9:05 AM on November 30, 2013 [3 favorites]


The Mathematica language isn't that bad, it just looks bad. It's Lisp with function notation instead of s-expressions. It's homoiconic and everything is a list. The head is the function name and the tail gives the arguments. At some point it had ML-esque pattern matching added to it, so as functional languages go it's a bit more forward-looking than Lisp. But it was first created before the modern appreciation for clean, orthogonal syntax emerged and it looks more like Perl than Haskell.

The linked article is certainly eye-rolling. This thing is vastly overstated. Mathematica is a handy tool for manipulating data, and binding that to a large body of curated information is an obvious good. Releasing Mathematica free for use on the Raspberry Pi is an act worthy of praise. However, making the leap from here to sentience is laughable. I'd point out that "sentient" is Koetsier's word here, not Wolfram's.
posted by rlk at 9:32 AM on November 30, 2013 [3 favorites]


I agree we need to stop taking Wolfram at his press releases. Though I do find WA useful from time to time, I was never a lover of Mathematica. About 1/3rd of the way through ANKOS I realized it was just Neil Stephenson having a joke over on us, there was no ending, and I threw it across the room.

But the idea of a computing paradigm where there is a described set of data and a goal, and the computer figures out how to best accomplish the goal, is not new. The question is, has he found a way to make it generally more useful than past incarnations?
posted by jeffamaphone at 9:34 AM on November 30, 2013


My old high school had a really good mathematical education: they offered the usual high school things, but they also had calculus II and calculus III and linear algebra and offered it as a structured, high school class to high school students. In addition, the grade requirement was somewhat relaxed, so juniors could take it. But every student had to take a math class every year for all four years. So a class was made for seniors who had already taken college sophomore math to fool around in, and for some reason it was sponsored by Wolfram Research, and we fooled around with Mathematica all day in it.

It was there that I learned to map, and to reduce, and to fold, and so on. It was not actually until I read SICP that I learned the basic basis of Mathematica's syntax and formalism is in Lisp. Basically, I abandoned Mathematica summarily then, except maybe for the actual purposes it was inclined to. It's not a particularly general-purpose programming language, and it's not a programming language particularly suited to AI, even if Wolfram claims assiduously that it is.

At heart, Wolfram is fundamentally a physicist, so he isn't motivated by the same motivations that the computer scientist is. You can practically hear this screaming out at you when he talks of complexity as if it were a good thing, something that is absolutely necessary for the power of Mathematica and Wolfram Alpha. I've used both of those extensively, and they're powerful, and they're very fragile, as tools go. I know both APL and Perl, in addition to regular expressions, and therefore I can tell you quite precisely the queer feeling that you have when you write code that you damn well can't read: you get the same feeling when you write Mathematica, and I bet you'll get the same feeling with the Wolfram language, if it isn't just identical to Mathematica.

A concrete example: their machine learning algorithms are called, "Classify" and "Predict", and seem to be designed to be non-transparent. Would they be stupid enough to back it with something that doesn't classify non-linearly separable things, like an individual perceptron? No. Will literally any algorithm they use to back it have significant drawbacks? Yes. Random forest, there's a drawback. Semisupervised deep ANN, there's a big drawback. Do they tell you what that model is? Not currently, hopefully not never, but that may fuck up your day someday.

Mathematica has ANN libraries, to be sure. There's got to be a random forest implementation out there somewhere. But the cavalier attitude of Mathematica towards complexity has led me to just fuck off and go use Lisp if I want to do symbolic computation.
posted by curuinor at 9:38 AM on November 30, 2013 [20 favorites]


I like the part where he breathlessly explains what an API is and the reporter passes it on like it's a really great idea that Wolfram just came up with. Is there some kind of active antipathy to competence in the world where journalism goes under a name like Venture Beat?
posted by RogerB at 10:11 AM on November 30, 2013 [5 favorites]


This isn't artificial intelligence or sentience, but it could form the basis for fantastically enriched sandbox worlds. I hope he gets together with Rockstar Games.
posted by stonepharisee at 10:19 AM on November 30, 2013


Around 2000, Matthew Cook published a proof that Rule 110 is Turing complete, i.e., capable of universal computation, which Stephen Wolfram had conjectured in 1985. Cook presented his proof at the Santa Fe Institute conference CA98 before the publishing of Wolfram's book. This resulted in a legal affair based on a non-disclosure agreement with Wolfram Research. Wolfram Research blocked publication of Cook's proof for 2 years.
posted by Obscure Reference at 10:23 AM on November 30, 2013 [7 favorites]


Metafilter: In conclusion, I am the king of the moon!
posted by JHarris at 10:28 AM on November 30, 2013 [3 favorites]


So basically, Wolfram|Alpha is this huge "curated" knowledge engine. And the Wolfram Language is essentially an otherwise unremarkable computer language with a bonus trojan-horse API by which they get to monetize all the information they've accumulated in W|A.

Now, the issue that I have is that if your program is doing something useful, that means it will be accessing this data in a novel fashion that you might not want your competitors to be aware of. But since the back end is W|A, then this is by nature a two-way street. They will know, automatically, what you are doing with your data and what kind of associations you're calling into play. It's like, imagine you're the Derpistan Government, you build a satellite to spy on the US, but in order for it to perform its basic mapping functions, it has to continuously use the Google Earth API. There would be no way for the satellite to hide what it was doing from Google, nor from the NSA. Similarly, how trustworthy would any assurances from Wolfram (& Hart) Research be that it was not engaging in some nefarious All Your Datum are Belong to Us scheme?

Still, I give Wolfram credit for realizing that somewhere between Google, Facebook and LinkedIn, there was still an untapped reserve of personal information waiting to be exploited.
posted by xigxag at 10:58 AM on November 30, 2013 [8 favorites]


But can it beat Watson at Jeopardy?
posted by Blazecock Pileon at 11:08 AM on November 30, 2013


A very readable review of A New Kind of Science: A Rare Blend of Monster Raving Egomania and Utter Batshit Insanity
posted by delegeferenda at 1:02 PM on November 30, 2013 [6 favorites]


The last time I used WA ... ever ... I asked it to tell me what is the volume of ice covering Antarctica. I got a weather report.

End-Of-Line.

At least Cyc (after decades of effort) isn't continually hyping its inefficacy. Maybe Wolfram ought to just build himself a big pyramid, or a 100,000-year clock, or devote himself to a "Space Z" project. The level of his "greatness" has already been amply demonstrated.
posted by Twang at 1:21 PM on November 30, 2013


I'll give the system five minutes before somebody asks it for the answer to Life, the Universe, and Everything and ties it up for the next seven and a half million years.
posted by Strange Interlude at 1:40 PM on November 30, 2013 [3 favorites]


It would simply reply ALL GLORY TO THE WOLF-FATHER.
posted by JHarris at 2:09 PM on November 30, 2013 [2 favorites]


The Wolfram Answer™ is protected by copyright and intellectual property restrictions specified in the Gospel of Wolfram™ to which you agreed before asking the Question. The Answer™ may not be republished, verbally summarized, or discussed in a non-Wolfram Language™ without the express written permission of The All-Seeing Wolf-Ram™.
posted by RogerB at 2:16 PM on November 30, 2013 [2 favorites]


Maybe now Wolfram will finally be able to fund a working implementation of "undo" in Mathematica -- despite all the impressive features of Mathematica, there is a long list of bizarre, very basic usability problems like that.

Feynman's letter to Wolfram comes to find: You don’t understand "ordinary people." To you they are "stupid fools."

However, the idea of organizing calculations as notbeooks that mix prose, graphics, and equations is highly underrated. I wish more people with organize their calculations that way, with an eye for making the calculation comprehesible to someone else, even if that someone else is one's own future self with imperfect memory.
posted by TheyCallItPeace at 2:24 PM on November 30, 2013 [10 favorites]


It's a database. You'll access Wolfram Alpha's database.

It's a custom built database that gets much of its data by doing other queries for you. Those queries use data sources that don't conform to any particular database design, but Wolfram Alpha has been made to interpret them anyway.

That makes it a more general software service than you would expect of a database. You can use it to construct graphical views on the data, and do whatever else Wolfram's team comes up with. And it has this very elaborate query language that has a lot of ways to hide complexity. (This is sometimes a good idea.)

It looks to be a very useful database. The self-modifying code thing means there will be a lot of security problems, but that's probably fine if you're only a user. Unless all code written in the Wolfram language is self-modifying, in which case I hope you like bugs.
posted by LogicalDash at 2:35 PM on November 30, 2013 [2 favorites]


You can tell this is a serious article because it contains a picture of a woman in her underwear overlaid with a screen of code.
posted by medusa at 2:47 PM on November 30, 2013 [1 favorite]


The Wolfram Answer™ is protected by copyright and intellectual property restrictions specified in the Gospel of Wolfram™ to which you agreed before asking the Question. The Answer™ may not be republished, verbally summarized, or discussed in a non-Wolfram Language™ without the express written permission of The All-Seeing Wolf-Ram™.

The expressive power of The Language makes all natural languages seem like the barks and yaps of dogs. And just as calculus could never be translated into growls and woofs, the answers to questions posed in The Language are only conceivable in The Language. Quite frankly, after learning The Language you will have a hard time seeing pre-Wolfram people as full human beings, because their mode of thought will be more akin to that of the lower animals than to your own.
posted by Pyry at 3:01 PM on November 30, 2013 [5 favorites]


Computer, what is the meaning of the universe?

[...compute compute compute...]
   42 
Computer, what is the meaning of the internets?

[...compute compute compute...]
   Rule 34
posted by jfuller at 3:08 PM on November 30, 2013


TheyCallItPeace: "Feynman's letter to Wolfram comes to find: You don’t understand "ordinary people." To you they are "stupid fools."

I've never used Mathematica or WA, but, given Feynman's letter and the fact that this guy held up the publication of the proof for so long pretty much tells me that I'm probably not missing out on much.

As for his new world-changing stuff, I'm not holding my breath.
posted by InsertNiftyNameHere at 3:26 PM on November 30, 2013


On the other hand, most of them truly are stupid fools.
posted by five fresh fish at 6:08 PM on November 30, 2013 [1 favorite]


You can tell this is a serious article because it contains a picture of a woman in her underwear overlaid with a screen of code.
posted by medusa at 2:47 PM on November 30 [+] [!]


... under the paragraph heading "Sentient code and intelligent objects". Ugh.
posted by mukade at 8:48 PM on November 30, 2013


INSUFFICIENT DATA FOR MEANINGFUL ANSWER.
posted by Nomyte at 9:45 PM on November 30, 2013


One presumes that just like Alpha this will oversell and under-deliver.
posted by ob1quixote at 10:39 PM on November 30, 2013


« Older ASCII fluid simulator   |   Indirect fusion's nothing less than HiiiPoWeR Newer »


This thread has been archived and is closed to new comments