Wolfram|Alpha Is In Production!
109 Comments 
Wolfram|Alpha is officially launched!
Wolfram|Alpha went live in test mode at 8:48pm CST on Friday. Our teams worked intensely through the weekend to complete load testing, fix bugs, and begin to address the feedback you have provided—over 22,000 feedback messages. During testing, Wolfram|Alpha processed nearly 23 million queries; by our estimates, approximately 3 out of 4 gave satisfactory results.
By late Sunday night, we were able to test all compute clusters at full capacity.
Thank you again for all of your enthusiastic support and feedback during this period. It’s been exciting to share your interest and discoveries. We’re really pleased to see communities growing up around Wolfram|Alpha.
Stephen Wolfram reminded us of his vision:
Fifty years ago, when computers were young, people assumed that they’d be able to ask a computer any factual question, and have it compute the answer. I’m happy to say that we’ve successfully built a system that delivers knowledge from a simple input field, giving access to a huge system, with trillions of pieces of curated data and millions of lines of algorithms. Wolfram|Alpha signals a new paradigm for using computers and the Web.
Today marks the dawn of a new era on the internet. Over the next days, weeks, months, and years we’ll keep building out the capabilities of Wolfram|Alpha.
We look forward to seeing where you take it!
when looking on the map of Israel, where I live, i realised that according to the site the Golan heights (north-eastern border of Israel) are part of Syria, instead being part of Israel, as it is these days.
This matter is being negotiated for more than three decades as the main conflict between Syria and Israel, but I wish you not to take any stand about it. I think that the leaders of Israel and Syria solely will decide whether this region will be part of this country or another, and if your service wish to remain objective and useful as it is, I ask you to fix this problem.
Best Regards,
Adi
It seems to me that the area you’re talking about isn’t part of either country.
http://www.wolframalpha.com/input/?i=israel%2C+syria
I assume you’re talking about the region that falls to the west of the middle of the map on that page.
Very odd htat you would interact with a wonderful resource like Wolfram Alpha, and narrow the whole experiences down to a geo political drivel. No, the Golan Heights are not part of Israel. Officially, they are part of Syria, but they are occupied by Israel since 1967.
Let’s leave geo politics to other sites, and let;s enjoy this great resource.
Tony
This is going to be fantastic if I can figure out how to do input that Wolfram/Alpha
‘understands,’ instead of getting the message ‘doesn’t understand your input.’
I’m trying to find out how many new automobile drivers there are each day in China,
and given current average gallons/day usage, how much their aggregate demand
between now and 2012 will add to world demand for oil as expressed in barrels
per month.
My reason for asking that question is to get a sense of how much upward pressure
there will be on the price of oil. How does one break that down for your input?
With VERY LIMITED public transport, (especially interstate), similar study is required more for USA. They consume (waste) the maximum and PROBABLY would be the hightest contrubutor for non-productive oil/fuel consumption.
please .. you have got to be kidding me .. Stop arguing the question and just answer it ..
This is the same type of trouble I am having; I can get various datasets out of the engine but I cannot make it understand that I want to experiment with correlating them or otherwise comparing them. For instance, you can compare two stocks with each other but you cannot compare a stock to other historical trends. I guess I am supposed to buy Mathematica to do this kind of voodoo?
Outside of the pure math functionality here, I am struggling to find a use for this other than as a big fact book that has a viral citation clause, though thankfully it is at least somewhat legitimate.
There isn’t good data on average miles driven per car in China. The US DOT publishes very good data on that for the US, but it is considerably harder to find elsewhere. In the US, average vehicle-miles per vehicle-year run 12-15k, while in the UK it is more than 8-10k and in Japan much lower, as low as half the UK figure. It makes a considerable difference, therefore, where the average Chinese auto falls on that spectrum. Probably toward the Japan end, given the expense and state of rural roads, but maybe headed toward the UK area gradually.
New vehicles are easier, that is running about 10 million per year in China these days. Average mpg can be ballparked at 25. Assuming Japan levels of miles per year that means about 5 barrels of oil per car-year. 35 million cars added in 3.5 years would means 175 million barrels of oil demand per year, or half of one million bpd (over 3.5 years). World output is on the over of 80-85 million bpd, so new Chinese cars are adding half a percent in the 3.5 year period.
For comparison, the recession in the US on top of high gas prices last year led to the largest recorded fall in vehicle-miles driven in the US on record, a 4.4% decline. The absolute level is 216 billion miles per month. Basically that is 2 years worth of new car additional demand in China, from the recession effect in the US.
I hope this helps.
I looked this huge project only 15 min., but it enought to resume that this is very nice, usefull project with brilliant ideas. Super…
I wish wolfram alpha could improve more and helps best for wolfram users .
I do not think they prioritize a spanish version, go learn english instead.
Looks interesting. Lets see how will be its performance in searching in future.
Please be innovative. Try to find out our needs by taking feedback or observing us and make the product for that purpose.
Very impressive. Good to see such a innovative site coming from someone else than google.
Congratulations on a great effort! Just a minor quibble could you please switch to something more non intrusive like Flash instead of Apple quick time?
wmv or avi would also be great. Can WA convert from one format to another. That would be great as I keep finding stuff in a format I can’t read.
Simply put Wolfram|Alpha is a juggernaut. As a senior in high school I can only wish that this tool was developed years earlier as it would definitely streamline the process of homework as well as research projects. Nevertheless it will surely prove to be useful in college next year. I set W|/\ as my homepage and have found myself immersed for hours at a time looking things up! I managed to convince my teachers at school yesterday to experiment with W|A, and they were blown away.
W|A is the beginning of something amazing, and I cannot wait to see what is in store. KEEP UP THE GREAT WORK,
Thanks,
Zachary L.
I totally agree .. I however am 30 years older than you and it is truly amazing the amount of time and resources that this engine will save ..
Spread the word!
Would be nice if we could click on the picture to get a it in a better size because how it is, it is a litlle bit to small. Congratulations to all the Project Members. Wolfram Alpha is a revolution. Does anyone knows in which programming language the Site and Cluster Software is written?
You can right-click on the picture . Do ‘Save Picture as’ to your desktop or whatever. Then open with Paint or similar and expand it.
Creo que el programa Wolfram Alpha será una revolución si todos
habitantes del planeta pudieran enterarse , en el idioma que hablan,
de todas las posibilidades que ofrece o pueda desarrollar en un futuro
próximo. Viva la traducción automática y todos los desarrolladores
de este fabuloso programa.
The system does not seem to be able to understand requests that are names of towns or regions
that have a direction as part of the name. For example,
South Range,Michigan Upper Peninsula of Michigan
Wow..!! I have been fan of mathmatica since my graduation years…Now,this search engine would fuel my RF optimizations works….If they can provide any spread shhet calculation…it would be more helpful…as people like us….who deals with telecommunication optimization works….
we have to deal with a million of data calculations in spread sheets…i hope…in future , wolfran would try to add this feature.
I left a suggestion on the Forum that WA stops analysing a question as soon as it comes to a word it does not understand and tells the questioner what it expects at that point.
The item went on to the forum but then disappeared.
This is truly an excellent resource. The mathematical aspect to it in particular has got to be my favourite at the moment, given that I’ve got two exams involving lots of differentiation and integrations this week. I’m sure that time will make it into an even more powerful tool.
very good effort. Highly polished and streamline output.
Thank you for putting this together.
Despite my criticism, I still have hope for this going forward.
Re: by our estimates, approximately 3 out of 4 gave satisfactory results
My personal experience was about 1 in 3 being ‘Sorry Dave’ (a bug you mentioned) when first tested.
As of now, it appears to satisfy about one in five queries in any meaningful way for me and only about one in ten (after I have learned what types of things it *can* do) yield answers that are worth the effort.
Ironically, one of the examples that I used to show how the system failed (’who is fred’) was subsequently used to good advantage later. In that instance, I was mystified by the fact that someone named ‘Gail’ with whom I had corresponded was the first person named ‘Gail’ that I could remember actually meeting. I remembered how this system behaved when I asked ‘who is fred’ and in this case, the answer to ‘who is Gail’ revealed what I was looking for. Although I was well aware of the name, it is rather rare — only about one in 12,000 people have the name.
This example — ‘who is xxx’ is an excellent example of both the strength and the weakness of the engine. When questions follow a pathway that makes sense to the engine, it comes up with an unambiguous answer that *can be* wonderfully complete. When (as in most general queries) it does not follow an antiicipated course, its answers are very poor or non-existant.
Someone said that this is more a result of misapprehension and/or ‘user error’. However, it is a false dichotomy to think that the engine can only either offer good answers to good questions or no answers at all. Like it or not, Google does stuff that is specific to a domain such as mapping and language translation when asked within that domain. However, it also does stuff that is specific to a domain intelligently from the general interface (example: brockville weather) AND it gives a ‘best’ answer as well as it can from nearly any input.
To say that comparison to Google is Apples/Oranges is correct in one sense, but not in the sense that is most meaningful. If a person can generally go to Google to get just about any answer, it will always be the first choice destination and currently it is. I used to use Yahoo many years ago. When Google came on the scene, I started using it more often as it became more capable. Now, if I do not find it on Google, I do not even bother going to Yahoo.
If you can do *almost* as good as Google at a general case and *better* at the intelligent queries, you will start to draw traffic from them. However, if they improve on the specific faster than you improve on the General, you have a problem. If you do not aggressively and rapidly expand your database and create a series of intelligent ‘fall-backs’ from queries you think are poorly formed, you will find that Google will remain the engine of choice as they learn from what you do well and improve their own engine on specifics.
Google has a *very* gentle learning curve. Even so, some people don’t use it that well. WolframAlpha has a learning curve steeper than most will be able to climb.
It is possible that this will become firmly entrenched as a niche site offering what it does. However, even if it does, over time, Google will continue to press its advantages and WolframAlpha will be hard pressed to compete.
Currently Google advantages include:
MUCH larger database
MUCH better response to general queries (no matter how formed)
MUCH better (likely insuperable) response to specific mapping questions (I live in a small town and I can actually see my house on Google)
MUCH better (likely insuperable) response to specific language translation questions
MUCH better response to questions about images
MUCH better response to questions about News (even about Wolfram)
MUCH better response to various other domains including even files on your own desktop system, finance, site specific, etc.
Wolfram offers something unique that Google does not. However, Google has proven very fleet of foot in responding to any information retrieval that looks promising.
As an aside, I would like to mention that the various states a Boolean can *actually take* in practice are:
Yes — set to true
No — set to false
I Know I don’t know– attempted to set but failed because answer was not known.
I don’t know if I know– never set.
Somewhere I was reading something to the effect that this engine can only return answers to things it knows as ‘facts’ or things it can derive from them. In the real world. These apparently come from some ‘curated’ sources.It is not just possible you will knot know everything in a list, it is probable. You should still return what you know or a best guess from the facts at hand. That is what things giving real answers do all the time. In fact, even this engine is generally returning things it ‘mostly knows’. Some constraints need to be relaxed. The front end needs to juggle inputs (and correct spellings, etc) intelligently in order to give as high a probability of a useful answer as possible.
In instances such as the ‘who is fred’ / ‘who is gail’ questions, the engine needs to be able to return a reasonable *list* of answers that cover a good portion of the ‘answer space’ Part of what constitutes a probable best answer is a learning exercise and I think the engine will improve iin that regard. However, in some instances, there really are more than one excellent answer to the same question, depending upon additional parameters that Google will often supply for you. I have yet to try this on WolframAlpha, but on Google it gives a great set of answers:
Who was the greatest president
My formal education is in Science and Software development. My research area is ‘Data Packaging’. However, good research is a creative pursuit more than an analytic one. The analytic is generally a sort of ‘post facto’ rationalization or ‘clean up’ of the original creative idea and investigation. In the messy world of real research, more is thrown away than kept. Pablo Picasso is quoted as saying something to the effect that ‘Comptuters are useless. They can only give answers’. Even though I have been a dedicated computer geek for decades, I side with Picasso when it comes to the narrow area of simple processing. The great power of the Internet and things like Google is that it gives you not only answers but by supplying lots of organized data to even poor queries, it gives you access to questions as well. As any veteran of surfing the net can attest, these lines of inquiry will often lead you to the answer to the question you had, but did not know how to ask.
Bob, regarding your comment about the various states of a Boolean, George Boole would beg to differ.
He said: “Let us conceive then, of an Algebra in which the symbols x, y, z, &c. admit indifferently of the values 0 and 1, and of these values alone.” — from “The Laws of Thought” by George Boole, page 37.
I have just one quick question. Instead of understanding what I want, this app requires me to understand what it can understand. Isnt the primary goal of knowledge computing to eliminate artificial metalanguage and semantics from human-machine comunication? Because if so I don’t see any focus on it. It’s apparent to me that all knowledge I can piece from this app are more readily available through other frameworks, with considerably more relaxed semantics and at lower computational and human cost. In fact, Google is more of a knowledge computer than this, although it has no explicit representation of knowledge. Is this really feasible, given this?
I just now saw that Bob beat me to asking the same question, in a more ellaborate manner. :)
It’s amazing to see the quibbles and nits this amazing product has produced. This is just the beginning for a tool that will aid researchers and ordinary folks in myriad ways..
It’s not perfect yet, obviously but it’s a giant leap for mankind already and I eagerly await future developments.
The true value of this site will be unleashed when the system is opened and the community can provide their own computations from their own data sources. Imagine … orders of magnitude more than what you already have
Incredible effort that has really made me think about new ways of analyzing and viewing different data sets to answer complex problems. How is qualitative and quantitative data used together? I am studying emerging efforts to define meaningful metrics (frequently hard to do) and then collect reliable data from many different sources to conduct dynamic strategic assessment of international security force efforts in Afghanistan and Pakistan. What does the actual data coming out of Afghanistan tell us about international efforts to support Afghanistan security, governance and development? What are some new ways we can better assess where we get “more bang for the buck” when committing precious resources (people at the top of the list of many) to eliminate safe havens for extremists and terrorists, and retard the spread of extremist ideology and violence in the region? I know it may take generations to help Afghanistan recovery from many years of violence and war but what can the data tells us when we ask, “what are the most beneficial ways the international security forces can both address Afghans critical short term needs while also addressing Afghanistan’s long term capacity?” The asymmetric, irregular, and insurgent based conflict observed in Afghanistan today illustrates the types of conflict we are most likely to see in the future. Your important efforts to use expanding computational capability on ever expanding data sets may really help us better understand the dynamics of these types of conflicts and to view our actions with a more critical eye. Lots to digest but I believe your product may have revolutionary impact on the way modern computing and programming power is used to sort through and use the most meaningful and relevant data to determine the optimal measures of effectiveness to assess attainment of strategic goals. It certainly provides much “food for thought” for our national security analysts. I hope you will consider looking at some case studies in modern conflict to expand the body of knowledge in what I have described in this post. Thank you!
I’m completely blown away by Alpha. Thank you for making me feel like I’m in the 21st century!
Please do mankind a favor and ignore the barrage of trite suggestions you’ve been getting in the blog comments. Don’t change the name, fonts, colors, or anything else. Don’t ruin a great thing! And don’t worry: at least some people do understand that Alpha isn’t a search engine.
hello, please trduccion al castellano, en argentina queremos participar
At first I have seen about wolfirm alpha in a blog. It is realy very helpful for everyone.
Hello,
when I try the following search
http://www.wolframalpha.com/input/?i=from+26.11.1999+to+26.11.2000
and press on “more forms”, i see the entry “11 months and 30 days”.
;-)
But without any irony: You did a great job with this site and I try to use it as much as I can!
Chris.
president 1845 US
Wolfram|Alpha isn’t sure what to do with your input.
US president 1845
gives correct answer.
formatting restrictions pretty tight…
great job, guys! i think this could be the alternative for google, google has met its match finally, u think this is going to make learning by going to school and stuff, like how it is now, redundant? hmmm….
Very impressive beginning, in my view. But here are my questions.
What is the ontological framework behind this effort to organize knowledge? If there isn’t a clear ontology, one may be getting started OK but have perplexing gaps. But CS folks are now cited in the very definition of ‘ontology’, so it isn’t being unfair to ask, I don’t believe. It seems like most scientists view the “world” as a cummulative hierarchy of real systems: particles, atoms, molecules geoid systems, plant ecosystems, animal ecosystems, and human cultures (E.F.Haskell construct). As humans — who have ‘invented’ a construct they named “knowledge” — we ask questions about parts of the hierarchy and have in our knowledge banks ‘answers’ with very broad to very limited generality. Questions range in difficulty to answer from description (What is the character of?) to explanation (Why?). QD vs AG forms a little grid with singular description in upper left and universal explanation in lower right. Wolfram|Alpha appears to do very well with UE at the lower reaches of the hierarchy of real systems, but ‘retreats’ to SD (singular description) as you move towards the top of the hierarchy. That is OK, ’cause scientists studying ‘human cultures’ don’t have ‘laws of nature’ that can be mathematized for a computation engine to evaluate. Same in animal predator-prey subsystems, although there are some fruitful suggestions out there for PP that I could not find in Alpha. The ‘take home’ is that as you go up the hierarchy of natural systems, Alpha will no doubt ‘retreat’ toward singular descriptions. I.E. don’t expect a partial differential equation solution to the Darfur situation.
Keep trucking!! Lots more fun ahead!!
Hugh potential!!
Please internationalize the quiered databases.
For example
my son’s name Kushal, or wife’s Ritu (Bengali origin) comes up empty.
My boss Joan (man) from Catalonya (Spain) says is a woman’s name (not 100%).
Hi,
Well done…Your production is perfect! Really, i like wolframalpha. I want to working your developer teams…
Thank you.
[...] marks one month since the official launch of Wolfram|Alpha. This is a perfect time for us to say, “Thank [...]
[...] we approach the anniversary of the launch of Wolfram|Alpha, we’ll be moving into Wolfram|Alpha’s next phase, centered on [...]
