Wolfram|Alpha: Our First Impressions
The hype around Wolfram|Alpha, the next "Google killer" from the makers of Mathematica, has been building over the last few weeks. Today, we were lucky enough to attend a one-hour web demo with Stephen Wolfram, and from what we've seen, it definitely looks like it can live up to the hype - though, because it is so different from traditional search engines, it will definitely not be a "Google killer." According to Stephen Wolfram, the goal of Alpha is to give everyone access to expert knowledge and the data that a specialist would be able to compute from this information.
Note: Wolfram asked us to refrain from posting screenshots, so we will not use any in this post.
Update 4/30: We just posted our screenshots here.
Alpha, which will go live within the next few weeks, is quite different from Google and really doesn't directly compete with it at all. Instead of searching the web for info, Alpha is built around a vast repository of curated data from public and licensed sources. Alpha then organizes and computes this knowledge with the help of sophisticated Natural Language Processing algorithms. Users can ask Alpha any kind of question, which can be constructed just like a Google search (think: "hurricane bob" or "carbon steel strength").
In today's demo, for example, Stephen Wolfram searched for "internet users in europe," or "weather oakland" - two queries that most users would also use in Google or any other search engine.
Where Alpha exceeds, is in the presentation of its "search" results. When asked for how many internet users there are in Europe, for example, Alpha returned not just the total number, but also various plots and data for every country (apparently Vatican City only has 93 Internet users).
Another query with a very sophisticated result was "uncle's uncle's brother's son." Now if you type that into Google, the result will be a useless list of sites that don't even answer this specific question, but Alpha actually returns an interactive genealogic tree with additional information, including data about the 'blood relationship fraction,' for example (3.125% in this case).
Alpha, of course, doesn't hide its relationship with Mathematica. Indeed, according to Stephen Wolfram, Alpha is built on top of 5 million lines of Mathematica code which currently run on top of about 10,000 CPUs (though Wolfram is actively expanding its server farm in preparation for the public launch).
Alpha can handle a lot of the mathematical questions that Mathematica can compute today (think: "integrate x^3 sin^2 x dx"), but every query will only run for a few seconds, so really complex queries will inevitable time out. Mathematica, however, will also be one of the first programs to make use of the Alpha API, so that Mathematica users will be able to access Alpha's repository of data.
Alpha also has a sophisticated knowledge of physics and chemistry data, and during today's demo, we also saw examples for nutritional information, weather, and census data. Most of the data in the system is curated, but real-time financial data or weather information will go through a system that checks the data for validity, so that outliers can be identified as potentially faulty information.
Pro Version
Alpha will come in a free version, but there will also be a paid version, which will allow users to download and upload data to Alpha. Stephen Wolfram did not go into too much detail, including pricing, but pro users will, for example, be able to not just see a graph, but also download the data behind this graph for use on their own machines or in Mathematica.
Embedding and Alerts
Wolfram is clearly taking a page from the modern Internet playbook and will allow users to embed not just a Wolfram|Alpha search box on their own pages, but they will also be able to embed results and a custom Alpha portal on their own sites. Users will also be able to receive email alerts when a result changes.
A Few More Random Notes
- every search results page on Alpha will feature a link to the sources it used to compute the results
- when a fact is disputed, Alpha will note this in a footnote
- Alpha will only be in English for now - Wolfram notes that this was already a very hard task and that the company does not currently have the resources to replicate its natural language processing techniques for other languages
- money: alpha will feature ads in a sidebar, but Wolfram will also partner with other corporate entities. He didn't go into any details, but it sounded like these corporate partnerships might include other search engines.
- Wolfram will release toolbars for FF and IE, as well as an IE8 accelerator
- Alpha will also display results from traditional search engines (Google, Live, Yahoo) and will feature links to relevant Wikipedia articles
Will it Kill Google?
No. Wolfram|Alpha will be an amazing product, but it's quite different from Google and other search engines. Indeed, maybe it is actually wrong to call it a search engine at all (and Wolfram prefers to call it a "computational knowledge engine"). If you wanted to know what sights to see on your next trip to New York City, for example, Alpha, from what we've seen so for, will not be able to help you.
Alpha, however, will probably be a worthy challenger for Wikipedia and many textbooks and reference works. Instead of looking up basic encyclopedic information there, users can just go to Alpha instead, where they will get a direct answer to their question, as well as a nicely presented set of graphs and other info.
Today's demo mostly focused on math and engineering data, so we'll still have to wait and see how Alpha copes with questions about historical events, for example.
Every product, of course, looks good in a controlled demo (though Stephen Wolfram also happily entertained questions from the audience for almost an hour), and we will have to wait and see how well Alpha performs when faces with real questions from real users. Based on what we've seen today though, it seems rather unlikely that Alpha will be the next Cuil.
Share
Facebook
Comments
Subscribe to comments for this post OR Subscribe to comments for all ReadWriteWeb posts
I like the idea of this. No, I love it.
Sounds pretty cool and it has a freemium business model - nice!
Strange name sounds academia/cultish at the moment, but in time it could sound as normal as Google sounds nowadays!
Did you have to mention Cuil at the end? It’s Bad Karma! It seems like a partner such as Mahalo, could help to round out their offering. They both have unusual approaches to answering users questions. It's is however, refreshing to see a startup that already has a monetization plan. I can’t wait to try it.
Seems to be quite interesting. But as any of those search engines, they will have to prove the quality of data.
good stuff frederic
Posted by: MG Siegler
|
April 25, 2009 11:02 PM
This is going to be different then Google but it sounds very interesting! Come check me out at http://endlesssports.blogspot.com/
another cuil.
http://www.google.com/search?hl=en&q=internet+users+in+europe
leads to first result on Google of:
http://www.internetworldstats.com/stats4.htm
which leads to:
http://www.internetworldstats.com/stats4.htm#europe
which lists the same 93 users in the Vatican. So I'm guessing that's the source data?
Google doesn't present charts, of course -- and I imagine what Wolfram's doing sounds much cooler than the best I can imagine.
On "uncle's uncle's brother's son," well, I'm not even sure what that query was supposed to show. Who would search for such a thing? I mean, it's impressive that it generates results, but what was the question seeking. The blood relationship of the son of a brother to an uncle of an uncle? I'm confused just trying to figure it out.
Anyway, still interested to see it. All the early reports have people coming away impressed.
thanks MG - I think I was quite lucky to get an invite to this demo - wish I could show some screenshots - everybody who saw them wanted to start using Alpha immediately
Posted by: Frederic
|
April 26, 2009 12:01 AM
@Danny - good chance that this is indeed the source. I think it's important to remember this this is data curated by 'experts' (they didn't quite explain who these experts are, though) - so Alpha is making at least some claim to providing authoritative information - something Google can't quite provide.
The genealogy example actually came up because somebody asked about how well Alpha could parse this kind of semantic data. The blood relationship was just one of a number of data points Alpha showed about this query - different ways to name this relationship were part of the answer as well, for example.
Btw. I know you're concerned up scripted demos, and this one didn't feel very scripted at all - it lasted for about 1:20 hrs and the last hour was basically just Wolfram asking random questions and trying different queries - he actually seemed surprised be some of the results.
Thanks again, Frederic. I am looking forward to seeing it.
All sound very exciting, but for the normal Joe public user Google will always be king.
Will it have an API so users can program to it, will it have an RSS feed so we can view real time searches being made and what results the user clicked on, and collaborative refining of the search results may be, on the lines of Wikipedia + Stumpedia?
I didn't make the connection until now. Thanks.
Posted by: Michael Fidler
|
April 26, 2009 2:03 AM
Hi frederic, thanks for this interesting post.
I am fascinated by the "google killer" hunt that web news bloggers are doing. I think it makes discussion bubbles about small things. This year we had "cuil", "powerset", "deepdyve"...
I think google is a giant that cannot be compete just with a better algorythm. The brand is to strong, the use very deeply set in every web user life. The presence very well organized, the google bar is just fantastic as to presence in web user life.
So, first to compete with google I see two technics : traditionnal algorithm/semantic search or a search social service. But whether you chose "algorithm/semantic" or "social search service" you won't face the same challenges : marketing for the first and credibility for the second.
Second, Google is now the search engine for everything, websites, piece of news, images, videos... Maybe a challenger could start by selecting a specific field before thinking about killing google. That is what happened to youtube.
Interesting. I like the name.
Small typo: "New York City, for example, Alpha, from what we've seen so *for*," (far)
Looking forward to trying this new search tool out...
Harvard are broadcasting a live webcast with Stephen Wolfram this Tuesday 28th April, 3.00pm ET
https://cyber.law.harvard.edu/interactive/webcast
You can submit your questions here...
http://cyber.law.harvard.edu/questions/wolframalpha
@L. Mohan Arun the top question to be asked at the talk - Will Wolfram have an API interface for the developers?
"normal Joe public user" had no idea that google even existed some 5+ years ago. it all takes some getting used to for the "public joe" to switch the default - inertia is too big. 22% still using yahoo prove that.
I see the future search techniques shifting away to the context-analyzing engines and adding semantics to the web (either by manually curating data or by creating a superior algorithm, or both as is attempted in this case).
Thanks Frederic for sharing your first look at Alpha. I'm looking forward to using it! Who, in 1998, would have predicted the success of Google? AltaVista was king then and people may very well have said "AltaVista will always be king." History has shown that most of those who attempt to predict the future, based on the past, are incorrect and that history is a poor guide to the future. Google represents the past. Wolfram Alpha may very well represent the future.
Follow me on twitter: http://twitter.com/kentbeatty
Does this mean Wolfram has stopped trying to calculate the cellular automata algorithm at the core of the physics? To build a new search engine? sheesh.
Damn, I missed the demo yesterday... Sounds like this will be nice, I can't wait to test it out..
It sounds amazing but it'll never take off with a name that sounds like a German professional wrestler. Even if it isn't after Google's share I'm sure it still wants to be a success, and a lot of folks are going to be put off from the start.
Impressive indeed! Can't wait to try it.
As somebody said above - "Looking forward to trying it", yes maybe trying but not using. Paid version)))))Are you serious???
The technology is amazing of course, but is it for real life. For academics? For some curious rare visitors? Or is it another step to degradation, when you do not even have to know where to search, it brings you things out of nowhere)))
Idea cannot clearly define itself. Is it a search engine or not?
To me it is like super invention, backuped by good finance, otherwise they could get 10K PCs, but lacking something)))) I cannot say what exactly but it looks like an expensive prototype.
Good luck WOLFRAM
I don't think anyone stands a chance at bringing down Google
This does look promising. Though it won't be a google-killer, I'm sure they will take notice as it is in their interests to make Google the only search vehicle in the public conscious. They certainly don't want people thinking Google is only good for certain kinds of searches.
I have been following up on Twitter and other forums about this computational engine. I agree that it will be different from the Google search engine but somewhere down the line for it to be successful, it has to cross swords with Google. otherwise it can't be a successful business model.
This reminds me of Frink|Alpha
This sounds a lot like "True Knowledge" (http://trueknowledge.com/) which has been around for over a year.
It is always very dangerous to judge the effectiveness of this type of software based on queries provided by the software's creator. I'd be much more interested in how it performs on queries created by ordinary people without any guidance from Wolfram.
I like the curated/referenced data aspect.I doubt it will be a negative thing - if you consider the horror that is yahoo answers and the various US politicians manipulation of Wikipedia..uh..facts. Even if Wolfram/Alpha serves only to filter out all the unreliable rubbish floating around on the net - it will be a super useful tool for the average surfer.
Also, I like the possibility that Wolfram may be able to prove some of his new kind of science "too higher grade for me mumbo" with a real life working system, sort of like the first word processor did way back when.....Nice.
Hmmm.. knowledge searching wanna try this out. cuil and kosmix dint dazzle.. may be they will compete with google some day but that will take some time..
Hmmm.. knowledge searching wanna try this out. cuil and kosmix dint dazzle.. may be they will compete with google some day but that will take some time..
A few marketing points:
-It should simply be called a "knowledge engine" - drop the "computational", makes it sound too geeky (even though it is).
-The "freemium" business model makes it seem like regular users are getting a lite version, which is a put-off. Best to just have a different name for the premium version like "Wolfram Alpha Editor"
-Possibly find a new name overall since Wolfram sounds like a wolf with memory chips. Totally retarded, and also too linked to one person's identity. Imagine if Google was called Brin.
Hi Bobby,
this is an excellent write up. I think a lot of people have misread what Wolfram is. I was also at the demo and I thought it was brilliant. Being a science person who could do with additional equation support and algorithm suggestions, it's perfect. As a user looking for stats and informational data - also brilliant.
Do you think that Wikipedia will suffer? I do see Google losing the information query traffic to some extent, at least from me anyway.
It was refreshing to read a complete and what I deem as accurate review. I'm glad you guys get it!
My own impressions were posted here
Sorry frederic - clearly I meant "frederic" and wrote "Bobby" - not enough coffee this morning! - Aplologies.
Wow dude that is WAY cool!
RT
privacy.mx.tc
Where is the demo? Oh wait, it's vapourware that only works with a small subset of user queries that have been hardcoded into it's database.
Sounds like Yahoo 2.0.
Frederic - Great initial insight. Thanks!
Cameron said...
Does this mean Wolfram has stopped trying to calculate the cellular automata algorithm at the core of the physics? To build a new search engine? sheesh.
Dude, the Google PageRank algorithm was derived using the same methods/algorithms that particle physicists had been using for many decades in their physics simulations. This algorithm is called Markov Chain in which Page & Brin used to derive PageRank with. The majority of cutting-edge computing techniques/algorithms of today originated in Physics, way before the software professionals adopted them and this is fact. I am pretty much sure that Dr. Stephen Wolfram will bring the techniques from scientific computing into the development of his search engine product over time.
One of the state-of-the-art emerging techniques that does enhance the retrieval precision & recall of numeric-based search engine algorithm of today is the Tensor calculus. Einstein first used Tensor in the development of his General Theory of Relativity (GTR) in 1916 when he published his paper. It is only recent that data-analysts realized its important in data analysis including search engine. PageRank algorithm has a tensor of 2 only (ie, 2D data only, ie the PageRank uses a matrix of inbound & outbound link frequencies) and I am not aware if Google has tensorized the PageRank yet, but there is no doubt that they must be researching/developing it (either for pure experimentation or for product enhancement). Recent tensor search engine uses 3D tensor which is reported to have improved (ie, recall & precision) over the current 2D based algorithms. Tensor calculus is a very complex topic and this explains why GTR is a complex topic. The hyper-space or multi-dimensional space that one often sees isn scifi as Star-Trek movie originates in Relativity and Tensor, since Einstein's Relativity is all about multi-dimensional space.
Anyway, since one of Dr. Wolfram's Physics research area was Cosmology, it means that he is well versed in Tensor Calculus since Relativity's application is mostly in Cosmology and this is the reason that I believe that Stephen will bring in the cutting edge techniques of science/physics to the development of his search engine product.
To those who compared Cuil to WolframAlpha is misleading. Cuil was a hype and I doubt that they invented anything original. Dr Wolfram is a scientist and he has been doing original research all his life. This means that he can always figured out a way (ie, always inventing) to improve his product.
Could you ask it what the most photographed landmarks in new york are? or for the most popular/best rated shows? What about economic data?
Even if it's just a free web interface to mathematica, It'd be awesome!
Look forward to checking it out!
Im not a geneologist, but the 'blood relationship fraction' seems wrong. Assuming that all relations are full siblings (not half-brothers) and that the 'blood relationship fraction' refers to average amount of genes shared (which is 1/2 for parent-child and 1/2 for siblings) you just use a factor 1/2 for every step from me to my uncle's uncle's brother's son. Uncle is 2 steps here (child - parent - sibling), so I end up with (1/2)^6, which is 0.016 rather than 0.032 ((1/2)^5). Also, this is when my uncle's uncle's brother's son is actually a different person from my uncle; they could be the one and the same. In this case I get 1/4.
Impressive indeed!
oops, its correct after all. Counted the last sibling step twice. The 1/4 option is still right though
Who is your admired figure ? Discover book recommendations from famous people on:
http://famouspick.blogspot.com
just so that you know
http://en.wikipedia.org/wiki/Characteristica_universalis
http://www.autodidactproject.org/other/leibniz1.html
Try http://www.yauba.com/ for a change.
Their search by segment is different.
WolframAlpha looks good. But I'm bothered about the name. Why did the founder want a rather-difficult, typo-happening name? I'm sure no one will remember it the first time they hear about it. Maybe "alpha" is fine. Not "wolframalpha".
Even amidst all the hype, it is relishing to see another search product coming out of Urbana Champaign.
We share the vision that understanding semantics of Web data will enable novel search capabilities. Our take on the problem is to design specialized vertical search engines. Try www.cazoodle.com
FF@39:
Markov Chains are commonplace applied math, used in myriad disciplines for generations. Whether or not Google uses it for page rank calculation has nothing whatsoever to do with Wolfram Alpha, nor with the ludicrous "new physics" exercise that Wolfram spent so much time on in recent years. The reak computational capabilities of those automata are so weak and inefficient that it's almost silly to suggest they are relevant. Universal Turing machine or not, it's a pessimal way to do computation.
Similarly, reference to Einstein's use of tensors is just name dropping. Einstein also used arithmetic, but that doesn't mean my for loop is great stuff. All this is very old hat math, and has nothing whatsoever to do with whether WA will succeed or fail. Once you start going off into Star Trek physics, you have really wandered into hyperspace yourself.
The comparison with Cuil is apt from a hype perspective, but we will only see if WA has more to it than hype after it launches for real. It's certainly the case, though, that A New Kind of Science was hype-filled, leaden, and in the end, not very productive, hence the reference to it.
1 2 Next