Paul Buchheit
Like Google, Only Much, Much Worse. Wolfram Alpha is yet another pretender to the search giant's throne. -
Like Google, Only Much, Much Worse. Wolfram Alpha is yet another pretender to the search giant's throne.
"Wolfram has discouraged comparisons between his new site and Google, insisting each serves different purposes. But that doesn't make any sense; you'll use both sites to look up stuff you don't know, so it's irresistible to compare the two. On a side-by-side test, Google wipes the floor with Wolfram Alpha. Using the search engine, I found life expectancy information for California and Kansas, the murder rate in South Africa and Baltimore, M.I.A.'s album sales, economic data for San Francisco, and the top speed of a Veyron: 253 mph. When you ask Google how many calories you'd burn on various sports, you find this page on, which lists caloric measures for all those activities and dozens more, including playing drums, horse grooming, and raking the lawn. That example illustrates the difficulty Wolfram faces in trying to match Google. Much of the data that we look for online aren't found in formal, structured tables like the CIA Factbook. Even Wikipedia misses mountains of facts. But there are countless one-off pages, like the one at NutriStrategy, that collect information we don't know we need until we search for it in a moment of whimsy. Wolfram misses all of that stuff; Google catches it." - Paul Buchheit from Bookmarklet
When I met with Microsoft's search team a couple weeks ago, they said most of tech media covers search engines "wrong". So I have held off on giving this a full spin, but I've been quite underwhelmed with the tests so far. Why should it be so difficult to determine what is the right kind of query and the wrong kind? Why should I feel like it's user error when the product doesn't easily define what it is supposed to do? - Louis Gray
I disagree, I think this is a great tool for research-based searches but not so great for trying to find a specific item - mjc
It still doesn't seem like a search engine to me. And it's actually hard to define what is and, specially, how to use it. - arnaldostream
@Michael, inspecific research is almost an oxymoron - Starepolsky smok
Yes but can google divide the amount of calories burnt playing drums by the weight of a slinky and provide the output in joules per slug-fortnight? - Mitch
"That example illustrates the difficulty Wolfram faces in trying to match Google" --- a complete straw man, disingenuous and foreshadowed by the opening commentary. Who said Wolfram was trying to match, or even play in the same game as, Google? - Isaac Hepworth
His test queries were very similar to the Wolfram demo queries though. I ran into the same thing -- it seemed like a cool idea, but I couldn't get it to actually give me any interesting answers even though I tried to construct problems that it should be good at. - Paul Buchheit
The most succinct response to this question that I've heard comes from Tim O'Reilly: "Why are you asking Wolfram Alpha questions that would be answered by a search engine. Kind of like using excel as a word processor" ( - DeWitt Clinton
@Alex, I didn't say nonspecific research. what I meant was if you are looking for a specific piece of content, wolfram alpha doesn't have a big enough index yet. if you are however looking for data on a subject that doesn't have to be from a specific piece of content (eg, plotting/solving equations, data about persons or trends...) then you will likely do very well with alpha. it's of course not as mature as google yet though. don't knock it until you've used it for a math course! - mjc
@DeWitt++ - Isaac Hepworth
DeWitt, I agree with the idea of it being something else, and like the concept, but it just doesn't work very well for me. Have you actually gotten anything useful out of it? (not using a suggested search, but one you thought of yourself) - Paul Buchheit
Criticism is easy. Innovation is hard. - Gregg Scott
I wouldn't think of Alpha as a search engine any more than I would consider Wikipedia to be one. - Gabe
wolfram alpha is much closer to public semantic data and querying platforms such as freebase than to search engines like google - Mike Chelen
Gabe: further innovation exists in mathematic queries and comparison to wikipedia is apt because of integrated data set - Mike Chelen
Paul, try searches that take disparate pieces of data and compute an answer, e.g. (or try population of USA/population of India) - Deepak Singh
You need to use your imagination a little bit to grasp that the comparison of life expectancies in all US states, or all countries in the world, is really within the reach of Wolfram|Alpha, even if "it doesn't know, *just now*, what to do with your input", when you ask it about these things. Same remark about homicide rates. Sure, those may not be easy numbers to get at for some countries, but homicide data is very easy to find, and trivial to incorporate into Alpha for lots of countries, regions, cities, towns or neighbourhoods. I agree WA will not be used by lots of people: very few people ever go explore the main shelves of the nearby reference library, and they wouldn't even if they only had to cross the street to get there. But hopefully, more and more people will progressively find a computation engine useful to things that matter to them. I think the distinction between search and computation resists Manjoo's analysis. - François Dongier
Paul: practical usage similar to online spreadsheets, for calculation of frequent tasks such as finding equations by regression - Mike Chelen
I get the concept Deepak, but was that an actual query for something you wanted to know, or a demo of Wolfram? My question remains, has anyone actually used it to answer an actual question they had? When I tried (something about the mass of oxygen in a 20cm sphere), it didn't answer. I go to go to Google for answers. I go to Wikipedia for answers. I could go to Wolfram for answers. In that sense, they all compete. - Paul Buchheit
@Paul -- in response to have I gotten something useful out of it? -- no, nothing other than a contrived query for weather data ( that I heard would work well (it did). But I don't see that as a failing of Alpha. I don't use Excel much either, and never Mathematica. But I do know what they're good for, and if I ever needed a spreadsheet or a computational software I'd know where to turn. I'm fine with Alpha being designed for a specific need. - DeWitt Clinton
Google (and Wikipedia of course) were in a much less advanced state than Alpha when they launched. - François Dongier
I agree that it could be very useful as "Mathematica on the web", but the demo video and other hype made it out to be much more. Hopefully it will get better though -- I do like the concept of what they are trying to do. - Paul Buchheit
Forget the hype, think for yourself. Alpha is Mathematica 7, 8, 9 on the web. - François Dongier
I think the underlying thread in this thread, as it were, is that we want to believe in something innovative and new, but we are frustrated as to how we are expected to modify our behavior to work around its limitations (intended or otherwise). - Louis Gray
It's qlpha, right? Let it go through beta and then full release--and then if it doesn't execute mathematical computations go ahead and dismiss it. - Gregg Scott
So you're waiting for WolframBeta Gregg? :) - Paul Buchheit
Yes! Gmail is still in beta after 5 years. I don't think it was intended for widespread use by non-academics anyway. It's machinery is for researchers, mathematicians and scientists. We should always support the innovators. Remember those black and white Apple ads? - Gregg Scott
Raise your hand if you use Google or Yahoo to search the web. Keep your hand up if you think Alpha is overrated. Now put your hand down if you *don't* use Mathematica or NumPy or Maple or Matlab on a regular basis. How many hands are still in the air? - DeWitt Clinton
Gregg, do you think the media coverage was an accident, and that they only intended for academics to use it? - Paul Buchheit
When Alpha (or Beta) will incorporate some simple logical reasoning ability, its applicability and overall usefulness will explode. - François Dongier
Paul. I don't know about their media strategy. I do know that tech journalism loves a new story so they may have inflated expectations by misunderstanding it's application. Or that could've been WA's marketing mistake. I heard Leo's interview with Paul Wolfram and I didn't get that he believes this is an app for the masses at all. - Gregg Scott
Leo's interview was excellent: go 2/3 towards the end of - François Dongier
Paul, WA may be guilty of inhabiting that bubble that all of us who love new technology inhabit by thinking that all the world is just like ourselves. It's a niche but potentially very powerful application. I think over time it will become more non-academic friendly. I could be totally wrong. Of course. - Gregg Scott
It didn't help that Doug Lenat gave a ringing endorsement. :) - Ray Cromwell
Comparing Google to WA is like comparing a cell phone to a graphing calculator. - Gregg Scott
Search engines always have this problem where you don't know their domain coverage exactly. If you don't like the results, are you asking the question wrong, or does it not have the answer? Even Google has this problem, but when you can't tell who has what, you end up always using the search engine with the broadest coverage -- which is Google right now. - ⓞnor
What I would like from WA is a nice hierarchical directory of all the data they have broken down by subject area. Their directory of examples is close, but isn't quite authoritative enough. That way I could really get my head around what they have. A well-curated set of general data banks with a good calculation engine on top seems like it could be really useful, especially if they make it possible for the community to participate in the data curation process. But it's not a search engine. - ⓞnor
The thing which seems to be missed by a lot of people is that Google has a pretty decent "computation engine" right now. See, for a useless example vs Obviously Google isn't exposing many entities ("facts") to it now, but the possibilities are pretty clear (See Google Squared...) - Nick Lothian
Wolfram|Alpha reviews seem to be a very interesting litmus test... there are those that grok the concept who may be disappointed by what it delivers - in this very early incarnation - but they don't write off the ideal. Those that don't grok it end up comparing the service directly to the Google in one way or another (often while proclaiming that they aren't) and write off the company without recognizing what a huge task has been undertaken. Believe it or not, there were many people that didn't Grok spreadsheets in the early days of VisiCalc! - David HC Soul
I think the underlying thread in this thread, as it were, is that we want to believe in something innovative and new, but we are frustrated as to how we are expected to modify our behavior to work around its limitations (intended or otherwise). - Louis Gray -- Think so too. I tried lots of things for kicks and was unable to get to anything really interesting, I had to look elsewhere for successful queries (like the life, universe, everything query), maths are fun, but they're mostly for practical uses (homework etc.), not much for full fun there. Let's wait for some upgraded index. And ++nor on the hierarchical directory. - Zu from AOD
Pretty good for crosswords though ;p - Zu from AOD
I agree with François. It is just "Mathematica on the web". And only this "mathimatica" part of the system works well. Other parts are data integration/inference and NLP they both are unsolved problems now. Wolfram|Alpha does not propose any new solution to these problems. It uses state-of-the-art methods and as result it does not work. - Maxim Grinev
Nor: gives a good idea of the currently available datasets. Clearly incomplete. The interesting question is whether making it complete enough to be useful is or not realistic. - François Dongier
From what's new in Mathematica-7: , the fun exercise is to imagine what new computable datasets will be included in future versions. Things could go very fast if the community contributes. - François Dongier
Most of my searches,questions and answers are on Twitter and FriendFeed based on my trusted friends(experts).Main trend. - Igor Poltavskiy
The data representation is great but the problem is that Wolfram expects the user to learn about it instead of it trying to learn about the user - Kiran Patchigolla
I see Wolfram Alpha more as a competitor to Wikipedia than Google. The media are the ones who are pitching Wolfram Alpha as a Google competitor, yet they are the same ones that are saying that it can't compete... crazy. - Stuart Maxwell
I hope the arrival of Wolfram Alpha will make Google start asking themselves the question "are we infallible." And I wouldn't be surprised if, in the near future, Google would begin paying more attention to how it presents what it emits. WA's one big and DISTINCTIVE advantage over typical search engine's output, is that it attempts to enhance and embellish the flow in graphically-palatable chunks (even though they've gotten few things wrong, and abuse small type in graphics-tables far too much). - ianf ⌘
Has anyone tried searching for "swine flu" - Peter Stuifzand
It is not "a great tool for research-based searches" even because it does not provide references to the sources where the answer was taken - Maria Grineva
Wolfram Alpha is a good first attempt at building a computational knowledge engine, but it needs to massively expand its database to become useful and appealing for most people. It is still a demo, a tantalizing taste of things to come, not a fully realized product. At a minimum, it needs to convert all the facts in Wikipedia into a computational format. - Sean McBride
"Maria - it does not provide references to the sources", yes it does, clik "source information" - Anders
Sean: Wikipedia data could easily be imported through DBpedia right? Haven't heard much about how to bring data from the RDF web (LOD) into Alpha, but that should be relatively easy. The hard part is what to do with it, as long as Alpha doesn't do logical reasoning. Maybe the best way of merging the two would be to extract RDF from the web of data through a SPARQL query, send the data to Alpha's API and have it return some useful interpretation of that data. - François Dongier
Compare Google and Wolfram Alpha for searches on the population of Pakistan: 1. search[Google; pakistan population] 2. search[Wolfram Alpha; pakistan population] The Wolfram result is more elegant and directly informative -- but takes quite a long time to produce. - Sean McBride
Compare searches on GOOG YHOO: 1. search[Wolfram Alpha; goog yhoo] 2. search[Google; goog yhoo] The WA search results are clearly superior. - Sean McBride
I like the concept, but I see two major flaws that need to be addressed. The first is probably easy: You can't, afaict, compare two disparate data sets over time (can you get it to give you a graph of russian gdp vs crude oil price?). The second is that they're asking for "curators" to help shepherd new data sets, but they don't seem to make them publicly available in raw form. I'm not going to volunteer to maintain their private database. - Joel Webber
Sean: Actually I don't really mind Alpha taking a long time to produce its answer, I even seem to enjoy it... Feels like watching the machine thinking, trying to guess what sort of calculation and graphs it will come up with :-) - François Dongier
I'm curious how long it would take google to approximate pi to a million decimal places... Mathematica 6 did it on my x2 3800+ w/ 2gb ram in about 8 seconds. Alas, 2gb of ram was not enough to approximate it to a billion decimal places :( - LarchOye
Pablo summed it up pretty well. Besides WA just launched a couple of days ago and as the users themselves its still in a learning phase. I just wish i had an iphone + WA back in school during all my horrible maths exams lol - Pretty Monkey Studio
I'm not listening to anyone's take on Alpha who isn't a Mathematica user. The rest of you tech journalists can go find something else to hype. - Mr. Gunn
FD: the problem with WA often taking so long to fully display its results: I'll wager that many users move on to another web page without realizing that they've seen only partial results. - Sean McBride
If WA still processes results "behind the window" after first showing them to the user without indicating it explicitly, then this is clearly a design error. - ianf ⌘
hmm... do you always stop thinking after answering a question? - François Dongier
cuil, anyone? - dario
No, François, but I look extra pensive to indicate more may be coming. Anyway, comparing lifeless computer output to a human answer is taking anthropomorphizing one step too far. - ianf ⌘
ianf: was joking :) - François Dongier
Thx, reading this article an thread give me some insights of the usefulness and reach of this tool - Luis Enrique León
Stephen Wolfram himself put it best, I think, in his interview with Leo the other day. He said basically that it's going to take some time for people to get used to the kinds of questions that WolframAlpha is going to be good at answering. And I have to add to that, over time the people on that project will discover and add new data sources to the engine, and come up with new techniques of comparative analysis. As they do the engine will become more powerful, and more of us will be able to correctly ask it questions over time. It may be that over time we can start expecting more something more useful than a bunch of indexes web links, but we aren't there yet. That's what we want, and since WolframAlpha doesn't provide us what we expect we don't like it. And for the few of us who are smart enough to give it questions that are of the kind it could be able to answer (Like you probably did, Paul), it's likely to disappoint until it has a significant enough store of data to work from. - guruvan (Rob Nelson)
That said, I thought this was a poor article, written by someone who really didn't try to understand the project. "Wolfram has discouraged comparisons between his new site and Google, insisting each serves different purposes. But that doesn't make any sense; you'll use both sites to look up stuff you don't know, so it's irresistible to compare the two" Wrong. It's just irresistible to write a blog article about it.And the comparison to the chatbot was even more inane. Just as inane as calling it a "wannabe Google competitor" - guruvan (Rob Nelson)
Wrong. - Chris Spizzirri
We noted your cute one-liner, Chris. - ianf ⌘
Rob: I agree with your conclusion (poor article). The structure of the argument is this: 1. Alpha is an answering engine, not a search engine. 2. Alpha does not answer most of our questions (although answers are on the web). Therefore: Alpha is not a good search engine. - François Dongier
I agree with Pablo, besides its good to see inventive new services, I will try to get benefit of it instead of comparing to Google. Like I was doing to Google when using Altavista. - Jacque
I like the new service. I can compute my motgage very easily with it. - Derek Wei
There is one thing wrong with WA: I dont want to be trained to use a search engine. - Burcu Dogan
But you've already been trained by your experience to know how to ask Google a question, and get the answers you seek. Why should a different tool be any different? For that matter, don't you need training to use any kind of tool? (even self-training) - guruvan (Rob Nelson)