Question the Answers? Answer the questions!
May 15, 2009 5:49 PM   Subscribe

Wolfram Alpha is about to go live. Wolfram Alpha is an answer engine which may just change the way we think about search results.
posted by rollbiz (133 comments total) 25 users marked this as a favorite

 
The infrastructure.
posted by rollbiz at 5:51 PM on May 15, 2009


Maybe mr Wolfram could teach his engine and team to tell time? T+1 hour is my limit. Will check it out, but obviously not now.
posted by mr.marx at 5:59 PM on May 15, 2009


This is awesome.
posted by East Manitoba Regional Junior Kabaddi Champion '94 at 6:00 PM on May 15, 2009 [4 favorites]


Um, I don't think they're actually turning it on until Monday.

(The Alpha web site doesn't tell you that, but Google has no trouble finding it...)
posted by equalpants at 6:00 PM on May 15, 2009 [1 favorite]


The screencast is pretty cool. I just hope it's not another Cuil.

(on preview, beaten to the link)
posted by Horselover Fat at 6:02 PM on May 15, 2009


It is live, if not totally public...
posted by rollbiz at 6:05 PM on May 15, 2009 [3 favorites]


Why don't we wait on the post until it does go live - in two days?
posted by lupus_yonderboy at 6:06 PM on May 15, 2009 [2 favorites]


Wolfram|Alpha isn't sure what to do with your input and the "tips" link doesn't work.

A New Kind of Searching!
posted by DU at 6:10 PM on May 15, 2009 [2 favorites]


I thought this was about some kind of Joss Whedon show crossover.
posted by liquorice at 6:11 PM on May 15, 2009 [6 favorites]


Well, I tried two questions. It gave me a good answer to "How many angels can dance on the head of a pin?" but was stumped by "Why did Fox cancel Firefly?"
posted by never used baby shoes at 6:12 PM on May 15, 2009 [2 favorites]


I think I am going to ask it why Cuil was such a miserable failure.
posted by adipocere at 6:13 PM on May 15, 2009 [4 favorites]


OK, I take it back (sorry) - there is a live link.

First question: what is the speed of sound in cadmium?
It gets 100% for that one. Wow!

how many 1cm spheres can you fit into a 10cm sphere?
It got zero for that one.

How many pentagons are there in a dodecahedron?
It got zero for that one too.

The heaviest gaseous element? I couldn't figure out a way to get that information out of it. Come on! (I tried most dense gas element, e.g.)

What is O() of quicksort? what is the diameter of a cesium atom? No?

Finally got a hit again at How many electrons in a boron atom?

(The only reason I'm picking chem/phys is because I got a hit with that right off...)
posted by lupus_yonderboy at 6:16 PM on May 15, 2009 [1 favorite]


Hmm, it's not so good, at least at the moment. I hope it has some more information when they officially turn it on.

works:
"Diamond Bar area code"
"Diamond Bar population"
"u.s. president"

fails:
"Diamond Bar zip code" (really?)
"slayer albums"
"alpha launch date"
"greek alphabet"
"cat species name"

...and now it's off again. Oh, well.
posted by equalpants at 6:24 PM on May 15, 2009


: Wolfram Alpha is about to go live. Wolfram Alpha is an answer engine which may just change the way we think about search results. \

Just like A New Kind Of Science changed the way we thought about complex scientific systems?

I'm not banking on it. I don't think we have to reorient our whole every few years just because one guy learned the ever-more-ubiquitous art of self-promotion.
posted by koeselitz at 6:25 PM on May 15, 2009 [2 favorites]


heaviest gas element
posted by East Manitoba Regional Junior Kabaddi Champion '94 at 6:26 PM on May 15, 2009


Couldn't tell me who I am, if there's a God, or what is the Dollhouse. I feel nonplussed.
posted by gerryblog at 6:26 PM on May 15, 2009


I attended Wolfram Alpha's webinar recently... it really was very amazing and useful, not only in how intelligently it determined context, but also how it presented the data.

The math queries were sweet, but these searches knocked my socks off: "next solar eclipse Chicago", "Bob in US", "BMI 5'6 140 lbs.", and a scatter plot of the GNP of European countries vs. their miles of railroad track.

If you live for facts, you'll love Wolfram.
posted by markkraft at 6:26 PM on May 15, 2009


I thought this was about some kind of Joss Whedon show crossover.

I thought Wolfram Alpha was where Starfleet made their stand against the Borg.
posted by steef at 6:27 PM on May 15, 2009 [3 favorites]


I'm asking it really simple questions ("how big is a lion", "how many people died in ww2", etc.) and the vast majority give a "WA isn't sure what to do with your input" reply...
posted by glider at 6:27 PM on May 15, 2009


I got a bunch of searches:
hamster, brownie, s&p 500, new york yankees, google, Cardinal Wolsley before it crapped out. YMMV. All of the results were, huh, I guess I might have wanted to know that...but if I typed in brownie you think I was really interested in kCal or its mass?
posted by shothotbot at 6:29 PM on May 15, 2009


the demo was really cool, but my serious / joke questions really didn't get me anywhere:

who killed Kennedy?

who wrote leaves of grass?

when do babies learn to walk?

how many americans make more than $100,000?
posted by bottlebrushtree at 6:31 PM on May 15, 2009


Who is X?

Angela Merkel, Tony Blair, Silvio Berlusconi, and Zapatero all get upgraded to head of state. Poor old Sarkozy is still a politician though. Not very up on European politics, clearly.

Trying it on actors is even worse...it reckons William Shatner is Canadian! Pfft!
posted by Sova at 6:33 PM on May 15, 2009 [1 favorite]


BTW lupus, to answer your question I actually posted it now because there's been this live video launch event and it's interesting to see technology being rolled out live like this. It's been incredibily delayed and awkward, but having been on the business side of a few launches that's kind of endearing to me at this point.

Plus the technology looks really cool!
posted by rollbiz at 6:36 PM on May 15, 2009


This is very very cool, I'm most excited about the medical information, which looks great.

I'm not sure what people typing in dumb search queries are really expecting to get. This is a number engine, not a web search tool.
posted by Super Hans at 6:38 PM on May 15, 2009


Even when I did get results, they were pretty limited... I think it'll be a while before they are more useful, even for number-specific things, than google+wikipedia/etc... But even most of my number-specific questions failed.
posted by glider at 6:43 PM on May 15, 2009


Fiddling around with Wolfram Alpha now, which is giving me a different kind of experience than the initial webinar. It basically shows me how much the way you ask it questions matters in getting the right answer. Obviously, a lot of this is learnable... and more of it needs to be tweaked by the site's designers. The thing is, they're getting a ton of good data on the kind of questions people will be asking, and how they'll be asking it, so it should improve noticeably over the next couple weeks, I suspect.
posted by markkraft at 6:45 PM on May 15, 2009


The hype preceding A New Kind of Science and then the abysmality of the book itself followed by the hype preceding this site is not a good sign. But the screencast is pretty cool. OTOH, "webinar".

Net score: -3
posted by DU at 6:46 PM on May 15, 2009 [2 favorites]


Once again, for anyone else who wants to watch, the live feed is here.
posted by rollbiz at 6:46 PM on May 15, 2009


Some math tries:


homotopy of 2-sphere
returns the data you might want, though doesn't give a reference to any relevant expositions or statements of any theorems: the same search on Google gives as the first hit a much more informative Wikipedia page.

That it doesn't know how to handle "Gauss-Bonnet formula" is just weird. Google again brings a relevant Wikipedia page right up.

Neither Wolfram nor Google can get "regulator of Q(sqrt(2))." Not surprised by Google, but a direct computation like this seems the sort of thing that should be right in Alpha's wheelhouse.

Alpha beats Google on the query "12th cyclotomic polynomial," though Alpha gives a graph of this polynomial, something that it seems hard to imagine wanting; I think what's going on is that if Alpha finds something it can parse as a Mathematica command, it runs that command and then gives you some prepackaged set of info about the result. So for instance "discriminant of 12th cyclotomic polynomial" gives you 144, nice as you like. So clearly one thing Alpha is pretty good at is being a natural language interface to Mathematica.

On the other hand, my attempt to pull up J_1, the Bessel function of the first kind, via "Bessel function j1" was a total failure; it seems to have interpreted j as a variable, and j1 as the product j*1. To get what you want you need to query on "Bessel function 1." Meanwhile, Googling "Bessel Function j1" gives you the information you need on the first hit; ironically, a page from Wolfram's MathWorld.
posted by escabeche at 6:46 PM on May 15, 2009 [2 favorites]


What is the speed of an unladen swallow?
posted by jenkinsEar at 6:48 PM on May 15, 2009 [3 favorites]


Humans melt at 940 degrees!
posted by glider at 6:51 PM on May 15, 2009 [1 favorite]


Wolfram Alpha doesn't like your potty mouth.
posted by markkraft at 6:52 PM on May 15, 2009 [1 favorite]


Scoopler. A (fairly new too) real time search engine.

The freshest results, not necessairly pertinent or accurate. Full of twitter. I'm not sure what purpose it serves. Profanity can yield some fun results.
posted by phoque at 6:54 PM on May 15, 2009


Also, asking it who I am gives the flight time from the country which shares my first name to the town in Germany which shares my last name.
posted by escabeche at 6:57 PM on May 15, 2009


This is a bug
posted by East Manitoba Regional Junior Kabaddi Champion '94 at 6:57 PM on May 15, 2009


Ahhh. So it is bullshit.

I searched a dozen obscure musicians, gradually getting less obscure, and got squat. It'll be pretty damned boring trying to use this stupid thing after it goes live only to get “Wolfram|Alpha isn't sure what to do with your input.” for every single search. It's like I've walked up to the desk at the library, and there's a five-year-old child there; I ask for books on metempsychosis, but the kid just says, “I don't know where that is,” and keeps twirling a ribbon around his finger and singing a song.

Anyhow, the revealing moment comes when I just search for “Grateful Dead.” (“Grateful Dead Music” gave me nothing.) When I search for “Grateful Dead” (no quotes) I get:

Development of this topic is under investigation...

This proves that this whole 'our search engine is a brand new form of artificial intelligence!' line is bullshit. You can see why, right? This search engine isn't computing anything; it's a glorified About.com. Behind the curtain, there are still hundreds of people researching the searches people put in and feeding the machine information to give out. If a 'topic' has to be 'developed,' then it sure as hell isn't computed; this just means that, when the almighty Wolfram|Alpha told me that Joseph Conrad was the author of “Heart of Darkness,” it probably only knew that because a guy in a cubicle somewhere in the UK looked it up on f'ing wikipedia.
posted by koeselitz at 7:00 PM on May 15, 2009 [19 favorites]


I've got a question: How is it for finding porn?
posted by mrnutty at 7:05 PM on May 15, 2009


"The Infrastructure"

Hmm. I happen to live across the street from the looming Wolfram Tower (made to seem oh-so-harmless by the inclusion of a TGI Friday's on the first floor...), here in Champaign, Illinois.

But guess what else lives across the street from the Wolfram Corporate Base?

Just sayin'...
posted by washburn at 7:05 PM on May 15, 2009


It doesn't know anything about MDR TB.
posted by These Premises Are Alarmed at 7:06 PM on May 15, 2009


It knows how much wood could a woodchuck chuck but it does not know why six is afraid of seven.
posted by TheOnlyCoolTim at 7:07 PM on May 15, 2009 [1 favorite]


koeselitz, you should see Wolfram's blog article about the launch. The level of self-aggrandizement is unbelievable. What's more, he says that much of Wolfram Alpha is based on "OKS" (the "old kind of science"), and implies that any way in which the system fails to live up to the hype would be the fault of the "OKS" part rather than the "NKS" claptrap.

Sorry if it seems like I'm criticizing the man rather than the product, but this sort of stuff really rubs me the wrong way. I work in computational linguistics, and I would be very excited indeed if Wolfram and co had really discovered amazing new ways of processing linguistic data, and it were all above-ground. But I don't think they have, and it doesn't seem like it is.
posted by Maximian at 7:15 PM on May 15, 2009 [5 favorites]


Reminds me of Dr. Know from AI.
posted by Korou at 7:32 PM on May 15, 2009


Wow. Stephen Wolfram is really fucking smart!

not.
posted by metastability at 7:36 PM on May 15, 2009 [2 favorites]


Wow, is this webcast live? Did I just totally randomly stumbled into it at the right time to see the launch?
posted by loquacious at 7:38 PM on May 15, 2009


For comparison, open up Alpha and Powerset at the same time. In my brief testing Powerset wins big. Where Alpha really falls down is the failure to provide relevant links when it gets confused. If 80% of the time Alpha just says Wolfram|Alpha isn't sure what to do with your input, it's going to pretty useless. At least Powerset tries to send you somewhere with the answer. (Of course, Google does that too, and faster.)
posted by The Tensor at 7:42 PM on May 15, 2009 [1 favorite]


Everyone who is comparing it to google are off - they don't really do the same thing. Google: here is a bunch of links that we magically dug out of the internet. Wolfram Alph: here is some data we magically dug out of ?internet?*

For the certain types of factual information I am looking for this will replace wiki, or relying on goolge to provide data rich information in a timely fashion, which is increasingly an issue. Also: if you are looking at multiple states or stocks or what have you they provide a remarkably consistent and simple interface.

East Manitoba Regional Junior Kabaddi Champion '94 : That's what it says about all former heads of state: See Clinton. Part of the "mr. president" for life bit perhaps?

Only thing I would change: the timeline at the bottom of people pages doesn't always start from the same date - which would be nice when looking at multiple folks at one time.

*is this one of those "just trust us" things?
posted by zenon at 7:45 PM on May 15, 2009


It knows why the chicken crossed the road, but not how many search engines it takes to change a light bulb.
posted by nowonmai at 7:56 PM on May 15, 2009 [1 favorite]


Wow. That interface is... weird. Possibly excruciating.

Hey, if you search Pi it calculates it and offers a button for "more digits".

Hrm, brb. *clickclickclickclickclickclickclickclickclickclickclickclickclickclickclickclick*
posted by loquacious at 7:57 PM on May 15, 2009 [1 favorite]


Using poorly-rendered (but 'designery') images to represent textual results is infuriatingly stupid. I don't fucking care that you really like that font, Stephen.

Especially bad when there's a FOUC -- perfectly nice plain text is replaced with gross slowly-loading images.
posted by blasdelf at 7:59 PM on May 15, 2009 [2 favorites]


Not so good on the basic math questions, either.
posted by nowonmai at 7:59 PM on May 15, 2009


zenon: I typed in "George W. Bush" and it assumed I meant "George H.W. Bush"
posted by East Manitoba Regional Junior Kabaddi Champion '94 at 7:59 PM on May 15, 2009


This would have been a nice conversion (and should have been easy).
posted by Kwantsar at 8:04 PM on May 15, 2009


I keep thinking of the Segway. Hopefully this will be more useful.
posted by various at 8:07 PM on May 15, 2009 [2 favorites]


alright, it aint much now, but this is the future and it is cool.
posted by saul wright at 8:08 PM on May 15, 2009


"Number of wild pelicans?" Nope.
"Wild pelican population?" Nothing!
"Pelican population?" The population of Pelican, Wisconsin is 2676.

Meh.
posted by verb at 8:09 PM on May 15, 2009


I think I broke something.

Yeah, definitely broke something.
posted by loquacious at 8:10 PM on May 15, 2009 [3 favorites]


I like the idea of this search engine, so I'm withholding judgment until they're not being hammered to pieces and obviously having serious errors.
posted by loquacious at 8:11 PM on May 15, 2009


Oh, hey, you can make it do math in the URL string. Those poor, poor servers.
posted by loquacious at 8:15 PM on May 15, 2009 [2 favorites]


Heh.
posted by loquacious at 8:27 PM on May 15, 2009 [1 favorite]


So the first thing I typed in was "sex" Heh. Actually it gave me a lot of information about the word itself, like that it's first use was in 1382, and that it's word frequency is rank is 1173, and it appears 1/11,000 something.

I bet googlers were smart enough to come up with something like this, but likely unable to justify it on increased ad-revenue alone.
posted by delmoi at 8:43 PM on May 15, 2009 [1 favorite]


Using poorly-rendered (but 'designery') images to represent textual results is infuriatingly stupid. I don't fucking care that you really like that font, Stephen.

I get the feeling that the (admittedly annoying) rendered text is to make screen scraping harder.
posted by signal at 8:52 PM on May 15, 2009 [1 favorite]


So far, this has yet to change the way I think about something.
posted by moss at 8:57 PM on May 15, 2009


what does a monkey have sex with
    Wolfram|Alpha isn't sure what to do with your input.

what is the hermitian of the 10 by 10 identity matrix?
    Wolfram|Alpha isn't sure what to do with your input.

what is the transpose of the 10 by 10 identity matrix?
    10^transpose (10 IdentityMatrix[n])

That one is kind of interesting, but not only is not computed, it's also wrong.

what is the transpose of the 10x10 identity matrix?
    10^transpose x^10 IdentityMatrix[n]

Still not parsing the string correctly.

how many co-sets does the set of integers mod 15 have?
    Wolfram|Alpha isn't sure what to do with your input.

Okay, I haven't gotten one useful result out of this thing so far.

If all this thing can do is look up basic statistical information and manipulate it, well, you can do that with Google spreadsheets already, actually.

how old is the creator of mathematica
    Wolfram|Alpha isn't sure what to do with your input.

Weak.

who created mathematica?
    Input interpretation:Mathematica 1.0 Released | people involved
    result: Stephen Wolfram

Which means the last one was actually a parsing failure, rather then a lack of knowledge.

how old is stephen wolfram?
    49 years 8 months 16 days

So it knew, but it couldn't figure out the text. Weak.
posted by delmoi at 9:03 PM on May 15, 2009 [2 favorites]


I get the feeling that the (admittedly annoying) rendered text is to make screen scraping harder.

Which would make it useless for blind people. I doubt they would do it. It looks like the alt tags are all there.
posted by delmoi at 9:04 PM on May 15, 2009 [1 favorite]


mrnutty: I've got a question: How is it for finding porn?

Well, in the interests of... testing this potential research tool, I tried the query “help me find porn.” It gave me this.

Which I take to be the equivalent of, “oh, you want to find... erm. I'm actually busy right now... can't you just use google?”
posted by koeselitz at 9:16 PM on May 15, 2009


*It dosn't know how many pine trees are in new york. (fail)

*It dosn't know who created metafilter. (fail)

*It knows how old Bill Gates is, and it knows how old Stephen Wolfram is, but it dosn't know who's older. (fail)

*I asked it how many sea ports there were in somalia, and if named two of them: Boosaaso and Kismaayo (well that dosn't tell me how many there are, but now we know it's more then two, which is more then we knew before, i guess)

*It dosn't know how many pirates there are in the world

*I got an interesting answer for "how many people in iowa and new york", it gave results for both states and gave ratios and comparisons. But didn't add them for me.

*"how far is the horizon when you are 30,000 feet above the earth?" was something I was actually wondering about the other day, and this thing had no idea.

Overall, this hardly works for the questions I can think up.
posted by delmoi at 9:18 PM on May 15, 2009 [1 favorite]


Apparently, it doesn't know how Babby is Formed.
posted by ooga_booga at 9:35 PM on May 15, 2009 [3 favorites]


Wolfram Alpha is an answer engine which may just change the way we think about search results.

Yeah, whatever...
posted by c13 at 9:40 PM on May 15, 2009


I think it's an almanac.

You can ask it some basic questions -- GDP of Botswana, largest US state, who won the AL Cy Young in 1985 (wait, it can't answer that) and it'll give you lots of nice charts and graphs and pictures. It's a bunch of numbers and facts, like what you'd see in an almanac.

But if it's an almanac, Wikipedia is the World Book Encyclopedia by comparison, and Google is the frikkin' Library of Congress, British Museum, and every truck stop porn video rental shop rolled into one. Maybe the way the data is being presented and crunched is different, but you can get the data elsewhere for free if you're willing to dig.

I mean, it's a nice UI, but it's thin in knowledge compared to the monster that is the web. I even think it's thin compared to the lazyweb.
posted by dw at 9:41 PM on May 15, 2009 [3 favorites]


I'm beginning to have my doubts.
posted by koeselitz at 9:45 PM on May 15, 2009 [2 favorites]


Try asking it about US greenhouse gas emissions. Or the population of Canada. Or perhaps you want emissions per capita. What are the chief exports of Japan. And so on.

On preview: yeah, it's an almanac (except it can do math too). I like it. The information is a lot more readily available than it would be if I had to browse wikipedia or google. The whole point is that I'm probably not willing to dig -- I just want to know a number now, and quick. I think it's going to be immensely useful.

I've used Google for this sort of thing. Like, you can type 'France population' into Google, and it gives you an answer as well as links, and you can do equations and unit conversion and other basic math. Google is limited and unreliable at this sort of task. Wolfram Alpha seems to be built specifically for this sort of task. I think Google should acquire them and merge the outputs.
posted by PercussivePaul at 10:07 PM on May 15, 2009


Here's another example. Canada's greenhouse gas emissions per capita have historically been a little bit lower than the US's, but in recent years Canada's has been climbing and now they are almost equal. I didn't know that; I punched it into Wolfram Alpha just now and the graph showed me. Here

The thing is, it's not perfect; when I tried doing 'per capita' instead of dividing by population it couldn't find the history, and when I tried to compare against world emissions it came back with insufficient data. Also I don't know if I trust the data yet, I want to know if that graph is real. But even as it stands the system is going to be very useful and its potential to improve, as it gets better at finding data sources, is just huge.
posted by PercussivePaul at 10:23 PM on May 15, 2009


Or have you tried using a comma to separate datasets? It compares and graphs them for you. Here's the populations of China, India, Indonesia, and the US. Look at the graph -- I've never seen that graph before, and it taught me something. And now whenever I want to see a graph like that, of just about anything, I type it in and it appears for me. I mean, god damn. Am I the only one that thinks this is cool?
posted by PercussivePaul at 10:33 PM on May 15, 2009


I will say this: It's one of the best online mortgage calculators out there.
posted by dw at 10:35 PM on May 15, 2009 [1 favorite]


I tried to look up how many people in the U.S. speak Spanish. "spanish u.s." or "u.s. spanish" gets you a side-by-side comparison of the U.S. and Spain. "u.s. languages" or "languages u.s." gives you nothing. Plain old "u.s." does give you a list of languages, with percentages, but does not explain what the percentages mean, or where the numbers come from. Suspiciously, Chinese doesn't even make the top 10, apparently being somewhere behind Sea Island Creole English and Navajo. This does not inspire confidence.
posted by equalpants at 10:38 PM on May 15, 2009


Waynes Worlded.
posted by cortex at 10:39 PM on May 15, 2009


Hm. Tried it 3 times. Locked up firefox 3 times.
Moving on.
posted by jouke at 10:56 PM on May 15, 2009


I can easily see this sold as an appliance to big companies, fed with the data from the company data warehouse. I think it would make a very decent amount of money.
posted by dhoe at 11:11 PM on May 15, 2009 [1 favorite]


It may be flawed, but it's funner than a dictionary.

eschaton = crack of doom

eschaton is surrounded by doom

posted by Rhaomi at 11:25 PM on May 15, 2009


Who put the bop in the bop shoo bop shoo bop?

Wolfram|Alpha isn't sure what to do with your input.


Color me disappointed.
posted by GavinR at 11:28 PM on May 15, 2009


I should hope that there is real human monitoring of failed queries, with ongoing revisions to the engine.
posted by General Tonic at 11:30 PM on May 15, 2009


hmm, it says the half-life of carbon14 is 5700 years when the standard value used is 5730 (+/-40) years. Weird that it would be close but not exact for a standard physical property.
posted by Rumple at 11:36 PM on May 15, 2009 [1 favorite]


Simply knowing what other websites the data came from, in the form of WA's 'source list' is not enough. Part of the usefulness of doing research via Google searches is that you see the aggregate of various available sources, and can quickly evaluate multiple sources against one another to get an impression of what the 'real answer' (if there is one) might be. Academic citation standards aside, a similar process applies with Wikipedia. If you are using Wikipedia to get a quick-and-dirty impression of the answer to a question (e.g., what garden vegetables might have high yields, or what charity X actually does with its money), you usually have the citations and external links to tell you exactly where the authors' information came from (and if there are no citations, then you regard the data as highly suspect and/or idiotic). Pretty much all of the successful queries above would be better answered via Wikipedia for that reason, modulo Wikipedia's vandalism/kook problems. The graphing thing is pretty neat, though.

An answer engine's result for a query like 'how many more Buddhists does country X have than country Y' is not going to be very useful without very specific information about where the numbers came from. Both Wikipedia and Google provide that in most cases. But if the answer engine provides the source documents up front, in a form you can easily use, then at best it's a slightly enhanced document-clustering system à la Google.

If I enter a query about Buddhists, and get some answer, I really have no idea whether it's the 'right' answer, and any researcher would be disinclined to trust a single company's figure in this manner. For the armchair researcher, the situation is just as bad, because around the watercooler, you end up citing your factoids with "WA said so". O RLY? YA RLY. Don't get me started on contentious figures such as "how many scientists believe the God of Abraham created the universe" or "how many members does the Mormon church have?". In the latter case, the figure itself is not really contentious, but the church propaganda says one thing and the real-world data says quite another. Who wins the 'answer' spot on the answer engine? Even if you allow multiple winners somehow, the situation is still difficult. Woe unto the answer engine that knows the word 'scientology'...

On another note, WA's user interface is, er, problematic. This is one area Google excels in by any measure I can think of.
posted by Maximian at 11:49 PM on May 15, 2009 [4 favorites]


A: 42.
posted by armage at 11:57 PM on May 15, 2009 [1 favorite]


This is awesome?
posted by sparkletone at 1:03 AM on May 16, 2009 [2 favorites]


where is god?
posted by leibniz at 1:50 AM on May 16, 2009 [1 favorite]


Needs more wolfmen.
posted by item at 1:54 AM on May 16, 2009


I think some people misunderstand the point of this.

It can't answer the narrative kind of questions, like: "should I buy Toyota or BMW?" or "who's older?" because it can only answer numerically.

I see its usefulness in dealing with statistics, historical data, and various small calculations. Highschool homework might need a bit of reform because of this thing.

IMHO, the potential is huge, and I hope they keep on improving it. The priority right now, I think, is the input parser. I'll wait a month and see if it lives up to the hype.
posted by simpleblob at 2:37 AM on May 16, 2009 [1 favorite]


This sounds both frustrating and awesome.
posted by Blazecock Pileon at 2:43 AM on May 16, 2009


Waaaaa a new free service that has potential to be great is in beta test and all I can do is complain about its shortcoming while I didn't pay a dime waaaaaaa. I'm the poster boy for instant gratification waaaaaa. Feed me, service me, hurl some porn this way waaaa.

Sorry but that had to be done.
posted by elpapacito at 2:52 AM on May 16, 2009 [2 favorites]


Bleh, I'm going to wait for Wolfram|Beta
posted by ladd at 2:58 AM on May 16, 2009 [1 favorite]


Metafilter is pretty.
posted by Captain Cardanthian! at 3:07 AM on May 16, 2009


I'm getting some great results, and some weird ones. Trying to find out who played bass in the Sex Pistols, it tells me Sex Pistols is under development. Similar search on Wilco tells me it's an interjection, first used 63 years ago (but doesn't define it as meaning 'will comply').

Intriguing though, and better than I expected. I can see some use for this as an alternative to something like NationMaster.
posted by Infinite Jest at 4:29 AM on May 16, 2009


Reminds me of the presentations Hans Rosling of Gapminder.org did at TED. His ambition was to have all the worlds statistics comparable and searchable – Wolfram Alpha seems to be in the same vein.

Of course the questions of accuracy and sourcing have to be addressed, but just looking at the implementation it seems awesome. I love that it gives you information you might not have thought of asking for…
posted by monocultured at 4:42 AM on May 16, 2009


Are they claiming it's a search engine? Lots of people complain it can't find things, when I don't think it's promising to. I thought it was going to be an advanced version of google calculate, and at that task it certainly impresses.

Dhoe has a great suggestion, though: this would be fantastic with custom data sets
posted by fightorflight at 5:59 AM on May 16, 2009


> If you live for simple-minded math/science facts, you'll love Wolfram.

FTFY. Outside of graphable data (which, as Maximian points out, crucially depends on where they get their data and how trustworthy it is, which you have no way of knowing), this thing is completely illiterate, and probably always will be.

> Highschool homework might need a bit of reform because of this thing.

Yeah, I think that's about its speed.
posted by languagehat at 6:23 AM on May 16, 2009


For something aimed at scientists it seems to have a peculiar attachment to imperial measurements. I gave it a weight in kilos and a height in centimetres. It gave me back BMI calculations but all the figures were in pounds and feet. It is possible to toggle to metric values but only for one section of the page at a time.

If this catches on expect more space probes to miss their intended planets.
posted by rongorongo at 6:28 AM on May 16, 2009


Interesting. I use natural language questions ("How many x's in a y?") all the time in Google and it (Google) returns a surprising number of fair to good answers. If the questions are technical, then the answers tend to be even better. Of course, Wikipedia pages have most of the answers I'm looking for, and if they don't, then I can go to Google books or even JSTOR. It's all a matter of Googlefu and knowing which database to search. IMHO.

Still, the more the merrier, so I say this is good. Wikipedia sucked at first. For that matter, Google was just a bit better then my old favorite Altavista. Give it time. My long term money is on Citizendium.
posted by MarshallPoe at 7:04 AM on May 16, 2009


No wireless. Less space than a nomad. Lame.
posted by gwint at 7:27 AM on May 16, 2009 [3 favorites]


Google has the obvious advantage of already having the web's aggregate best guess for any query and can (and does) add "pre-processed" query types at their leisure, so that as certain queries reach a critical mass deemed worthy of a special case result, Google can just add it, along with the web's best guess (calculator, stock quotes, maps for addresses, etc.)

With that said, competition is always good and WA right out of the gate does a lot of little things well, more layout/design-wise than natural language processing-wise (I like how the above mentioned mortgage calculator not only comes with the requisite charts and graphs but also comes with an option to export as a PDF.)

Related: The browser address bar is the new command line.
posted by gwint at 7:43 AM on May 16, 2009


I did a search for "self aggrandizement" and it told me how many copies of "A new kind of science" have been sold, and how long Stephen Wolfram's dick is. It also plotted this information versus time.
posted by metastability at 8:18 AM on May 16, 2009


how old is the creator of mathematica
Wolfram|Alpha isn't sure what to do with your input.

Weak.

who created mathematica?
Input interpretation:Mathematica 1.0 Released | people involved
result: Stephen Wolfram

Which means the last one was actually a parsing failure, rather then a lack of knowledge.

how old is stephen wolfram?
49 years 8 months 16 days

So it knew, but it couldn't figure out the text. Weak.


I think you've got your expectations backwards. Storing information is easy. Interpreting natural language is very, very hard. If this thing is going to have problems, it makes perfect sense that they'll be on the parsing end, rather than on the knowledge or calculation end.
posted by nebulawindphone at 8:46 AM on May 16, 2009


I think you've got your expectations backwards. Storing information is easy. Interpreting natural language is very, very hard. If this thing is going to have problems, it makes perfect sense that they'll be on the parsing end, rather than on the knowledge or calculation end.

That makes sense, but Wolfram seems to be pitching Alpha as some kind of magic bullet that can do everything, including natural language parsing.
posted by madcaptenor at 9:09 AM on May 16, 2009


It's an almanac seems to be about the best way of looking at it. The public internet has been missing that kind of information, forcing the armchair researcher into the frustrating process of combining Wikipedia, the CIA World Fact Book, Yahoo Finance, and random Google results to end up with middling results at best. Still, who knows how effective it can really be, and they need to make money too..
posted by Chuckles at 9:42 AM on May 16, 2009


We were looking for is 'What is tungsten, or wolfram'.
posted by Zambrano at 9:49 AM on May 16, 2009 [2 favorites]


I'm looking forward to seeing what people can do with this tool. I can't tell on a quick test whether it's deep and amazing, or shallow and very limited. Right now I feel like I'm talking to a reference librarian with the mental comprehension of a beagle.

But what the fuck is with serving results as images? That's just outrageous. There do seem to be useful alt tags on mouseover, at least.
posted by Nelson at 9:50 AM on May 16, 2009 [2 favorites]


Try asking it about US greenhouse gas emissions.

That only works because it knows to turn "greenhouse gas" into "greenhouse gas emissions".

But if I want to look up: Toyota US Sales...

It concludes Toyota and Sales are both locations (in Japan and France respectively), and then tells me how far apart everything is. Great.
posted by smackfu at 10:02 AM on May 16, 2009


All I can say is this is my new telephone answering reply.

*ring, ring!*

"Wolfram Alpha! (said cheerily)"

"..."
posted by Lipstick Thespian at 10:36 AM on May 16, 2009 [1 favorite]


I see much more going on here, though not available yet. The interface is right for a major expansion of discoveries to be made from all of the data stored in different places. If, say, a product X caused disease Y, it could potentially discover this by someone simply asking the question, and then telling it to find out if it didn't know. It could do this by finding a list of people known to order product X, cross checking them with medical records, or an email survey of everyone involved, or both. Obviously those resources would be reserved for someone with trusted status, but the computational interface for this type of ethereal being is emerging in W|A.
posted by Brian B. at 10:56 AM on May 16, 2009


What do you dislike?
posted by naju at 11:12 AM on May 16, 2009


But what the fuck is with serving results as images? That's just outrageous. There do seem to be useful alt tags on mouseover, at least.

There's also some copyable plaintext if you click.
posted by naju at 11:25 AM on May 16, 2009


Wolfram|Alpha isn't sure what to do with your input.

I think we're fucked.
posted by TheOnlyCoolTim at 1:23 PM on May 16, 2009 [3 favorites]


Even as an almanac, its in its infancy. I entered 'Giacobini-Zinner', a comet I contributed observations on when I was a teenager, hoping I would get back not only stats on the comet but also when it would appear to us in its orbit.

It didn't even recognize the term.
posted by vacapinta at 2:00 PM on May 16, 2009


I like how they're going into how the weather is bad in the video, and how absolutely nothing will go wrong. It's just like the beginning of Jurassic Park. According to my understanding of chaos theory, this means the servers are going to rebel. They even choose to give the site the evilest sounding name they could think of...
posted by mccarty.tim at 5:24 PM on May 16, 2009


I think this system needs to get a wider database. It's nice to know everything is correct and checked by an expert, but I'd like it to fall back on Wikipedia or third-party sources if it isn't in the database. It could include a confidence rating with the result(s), having the highest score for internal data, and judge the chances of info being right on Wikipedia based on the number of times the article has been changed or what tags qualify the post. Perhaps it could have a button for when you know an answer is false, and it could use a Bayesian system to rate sources.

I really want to see the ability for this code to trust the wisdom of crowds, as these types of things really become useful when they learn to accommodate the niche and the obscure without human intervention. That's why directory search engines went obsolete when Google came around. In this case, we have a really intelligent system to crunch data, but it has a somewhat stale cache of data. It needs to find some way to learn. If they can find a way to judge if data is parsimonious, then this could become a real killer app. As of now, it's just a neat way to get back-of-the-envelope statistics quickly.

Also, I'd like it to support queries via cell phone like ChaCha.
posted by mccarty.tim at 5:38 PM on May 16, 2009


"It's an almanac seems to be about the best way of looking at it. The public internet has been missing that kind of information, forcing the armchair researcher into the frustrating process of combining Wikipedia, the CIA World Fact Book, Yahoo Finance, and random Google results to end up with middling results at best. Still, who knows how effective it can really be, and they need to make money too..
posted by Chuckles at 12:42 PM"

Query: "CO2 Emissions and Monthly Global Temperatures over the past 200 years."
Result: "There is no correlation, and the data is completely boring and irrelevant. Not posted to save monitor space and your valuable time. Sponsored by Exxon Mobil."
posted by mccarty.tim at 5:49 PM on May 16, 2009


WolframAlpha interprets poo as Embaixador Walther Moreira Salles. Why do you hate Walter Salles, WolframAlpha?

Yes, this is how I test new search engines
posted by Paragon at 5:58 PM on May 16, 2009


"Juliet isn't sure what to do with your input" is my new life motto.
posted by jokeefe at 10:51 PM on May 16, 2009


Well it can certainly handle the important stuff.
posted by sammyo at 7:53 AM on May 17, 2009


Fuckin' thing sucks
posted by tehloki at 12:53 PM on May 17, 2009


Add me to the "Kills Firefox 3 Every Time" group. Even with plugins/addons turned off, javascript obviously enabled. Damn.
posted by sidereal at 3:37 PM on May 17, 2009


I was wondering what the polyhedron in the logo was, and it was kind enough to tell me.

The computations it's willing to perform with results seem limited. It will tell me sphere diameter 1 cm volume and tungsten density but can't multiply the two to give me a mass.

On the other hand, if you want the density of mercury, it'll assume you want the density of the element but also offers to tell you the density of the planet.
posted by NMcCoy at 4:10 PM on May 17, 2009


Too bad they can't hook this up to a good financial feed. If I type in MSFT I get:

"dividend yield 2.37%"

Oh great, it understands so properties of stocks. So what if I said "dividend yield MSFT"

I get .02374% and a nice little graph. Oh that's pretty cool, I mean it is wrong but I can carry the decimal place myself.

What about non-common things, like the ETF CYV? No it doesn't know that, or really anything else. I also can't do something like say "dividend yield MSFT to dividend yield of small cap companies" which is kind of hard for it to figure out, but what about comparing dividend yield to another company? Like Apple? Fails there too.

What if I type in iron condor? I get a nice little excel type spreadsheet and some graphs explaining an iron condor, which is nice, but nothing a first search in Google will give me. I have no live data.

But, when I explicitly give it $150,000 5/1 arm 5.2% 30 year I do get a really nice output with low rate, medium rate and high rate along with graphs showing the difference in interest paid and principal over time and possible scenarios with a rather complex instrument. Oh and I can include things like "maximum year adjustment."

Yeah this is really cool, I could really see the power of this if you hook it into a structured database. Let's hope they don't just kind of stagnate, I can definitely see using this instead of trying to search for things like mortgage calculators on bankrate or whatever.
posted by geoff. at 4:56 PM on May 17, 2009


So, um, it's a bit crap isn't it? or am I just using it wrong?
posted by Artw at 12:44 PM on May 18, 2009


Meh
posted by lalochezia at 10:17 AM on May 20, 2009


lalochezia: "Meh"

Ahem
posted by defenestration at 9:06 AM on May 21, 2009


How many plays did Shakespeare write?
Wolfram|Alpha isn't sure what to do with your input.

Which plays did Richard Burbage act in?
Wolfram|Alpha isn't sure what to do with your input.

Who was Hamlet's uncle?
Wolfram|Alpha isn't sure what to do with your input.

This thing sucks so hard.
posted by Greg Nog at 9:11 AM on May 21, 2009


I'm trying very hard to come up with questions that I would want to ask that it should have an answer for, and it's failing every time. At best it picks up a single keyword within the question, which links to something like a wikipedia sidebar.
posted by Artw at 9:19 AM on May 21, 2009


Wolfram|Alpha claims copyright over some of their results but they will be happy to grant you a "low-cost" license for publication.
posted by grouse at 10:29 AM on May 21, 2009


Charming.

"Our incredible self importance will overcome basic uselessness and user hostility!" should be their slogan, or possibly just "Don't be useful".
posted by Artw at 10:39 AM on May 21, 2009 [2 favorites]


Wolfram|Alpha claims copyright over some of their results but they will be happy to grant you a "low-cost" license for publication.

To be clear, they're claiming copyright over the presentations of some of the results, not the results themselves (you can't copyright facts). This is no different from any of a number of websites, e.g. Google copyrighting their image results page but not the images themselves. And it's a low- or no-cost license for commercial, non-academic use. Presumably, your typical small publication will get a free license, it just needs to ask.
posted by naju at 10:53 AM on May 21, 2009


To be clear, they're claiming copyright over the presentations of some of the results, not the results themselves (you can't copyright facts).

I think that is clear as mud. They certainly seem to claim copyright over the "collections... of data" among other things. And I personally think that their terms of use were intentionally written to be vague so that they can claim copyright or plagiarism later on very broad categories of results while outlining only the most defensible examples, which is all you picked up on.

As for "just needing to ask," why would I want to do that when there are other sources where I don't need to ask. Or have them deem my result "commercial" later on. It's not like Wolfram has a reputation for litigiousness or anything.
posted by grouse at 11:23 AM on May 21, 2009


They certainly seem to claim copyright over the "collections... of data" among other things.

Sure, I guess I meant that as well. You can certainly claim copyright to a sufficiently creative and unique collection or compilation of data. And companies have been doing that for decades. For example, companies that compile financial data typically have restrictive copyright policies requiring permission.
posted by naju at 12:00 PM on May 21, 2009


« Older Meet Stephen Strasburg. He throws 103 miles per ho...  |  Today was a troubling day for ... Newer »


This thread has been archived and is closed to new comments