Wolfram Alpha Live | Hacker News Reader

Wolfram Alpha Live

wolframalpha.com

91 pointsgrinich117 comments

Many news sites (e.g., TechCrunch) have been hyping WolframAlpha as a threat to Google, and it's natural that people are trying Google-style search queries. However, looking at the examples at the site (http://www.wolframalpha.com/examples ), it's clearly not intended as a direct competitor to Google, and it's not at all surprising that it doesn't perform so well on the Google-style queries people are trying. My first take is that it's more like learning Mathematica: you need to learn how to ask it the right sorts of questions, and the examples look like a useful way of learning to do that.

No doubt many of the news outlets who've been hyping WolframAlpha as a competitor to Google will now denounce it as having failed, when it wasn't meant as a competitor at all. It's the news sites which have failed.

As mathematical queries give by far the most impressive results, this is essentially just a very elaborate advertisement for Mathematica -- just as NKS was.

NKS? Just curious.

A New Kind of Science

A New Kind of Science--it's a new scientific philosophy championed by Wolfram, and also the title of his book on the subject. A couple informational URLs:

http://www.wolframscience.com/

http://en.wikipedia.org/wiki/New_kind_of_science

Aaronson rips NKS a new one in this review: http://scottaaronson.com/papers/nks.pdf

Mathematical queries aren't the most impressive results because those have already been around in things like mathematic/maple/matlab for years. Symbolic integration is cool, but typing in ISS, and getting the current position of the international space station is cooler.

This was also my first impression here. Try to compare things, it works beautifully! -> http://tinyurl.com/qevs58 yes, it works also with "mars earth" but commands are way cooler :)

I just followed your link for the Mars to Earth comparison, and noticed a cool detail: it reported the current sky position for Mars for MY location, which I assume means that I.P. detection is being used to geographically customize answers. That is useful.

>which I assume means that I.P. detection is being used to geographically customize answers.

Yeah, the screencast a couple days ago mentioned this. They use it for more mundane stuff, too - like guessing which Springfield you mean.

I'm not so sure the blame falls entirely on the news sites. As I've watched a seemingly unnatural amount of attention fall upon an unreleased product over the past few weeks I've come to the conclusion that we're being led around by Wolfram's PR firm. If that's the case and they were using 'google-killer' to pick up buzz, I think Wolfram deserves to fall on its face.

(That's not to say it's not an interesting product).

I herd it was a Google-killer, but it turns out to be a Google Calculator Killer

It really does kill Google calculator! It does symbolic stuff, like Mathematica. http://www.wolframalpha.com/input/?i=y%3D+y^2+*+x^2+sin%28x%...

Wow, college math homework beware.

That's true, it's one hell of a calculator.

i can never get my php shell to work right so having simple math commands like floor() handy will be awesome. I think i'll tokenize them in Launchy.

W|A is good at dealing with huge numbers, unlike Google Calculator:

http://www.google.ca/search?hl=en&q=9999999999999999-100...

http://www.wolframalpha.com/input/?i=9999999999999999-100000...

From my newspaper today. Direct quote from Wolfram - "It will change how people search online".

I don't think it's the news outlets fault if the founder is making outlandish claims.

Because newspapers never misquote people or take things out of context. :)

Did it quote Stephen Wolfram as having said that, or just Wolfram? In fact, another source quotes Conrad Wolfram, who is apparently Stephen's brother, as having said that:

http://current.com/items/90053264_googles-search-engine-riva...

WolframAlpha is Not threat to Google it just have different target of people. It's more like online scientific encyclopedia + can preform complex math calculation.

I think wolfram alpha is a better deal than Cuil, but might get as bad a name because of their flawed positioning.

The method does not work without domain specificity. Its not a general purpose tool. Marketing it as one is the biggest mistake wolfram has done and is the reason most people will be dissapointed and inevitably try to slot it as another cuil.

However, if you are willing to disregard that largely cosmetic flaw, the computational engine is very impressive stuff. Assuming, in the future they are able to market/position it properly for specific domains for which they ensure that data is enough, this is going to be an awesome tool and may well compete with Google for the domain specific queries.

I think the majority of the populace will not have the wherewithal to appreciate it. The part that does will see its potential, and I think that's all Wolfram cares about.

I am very surprised with how it can intelligently generate forms based on a question. For example, the search "Am I too drunk to drive?" (http://www.wolframalpha.com/input/?i=Am+I+too+drunk+to+drive...) generates a form where you can input various variables to answer the question.

Really, if you manage to formulate that query, think about the answers, and type them in, you are most likely not that drunk.

Edit: Still, interesting.

Data can be interesting and smart without being useful.

Do I win hacker points for being the type of person who would think to do exactly this while drunk?

You lose hacker points for posting a question staring with "Do I win hacker points...?" But if you are only joking, you only lose joke points.

So bookmark the query. Filling out a form is a lot easier to do while drunk than driving safely. Your fingers may have trouble hitting the keys correctly, but surely your brain remains usable?

hmm, I filled out this form and it says I have a BAC of .06% Then it said time to .08% = 1hr

How does it know I'm going to keep drinking? Probably because it's only 9pm...

depending on how recently you drank, it takes time for your body to absorb the alcohol from your stomach...

Need more than one example to call it intelligent. I tried "Am I mentally retarded?" (IQ test?), "Am I at risk for cancer?" (risk factor questionaire?), "Am I at risk for lung cancer?", "Do cigarettes cause cancer?" all without ANY results. Help me out here, what kinds of questions is it smart about?

From using it, it seems it's not so much smart as it is a whole bunch of hard coded solutions.

http://www.wolframalpha.com/input/?i=How+much+wood+would+a+w...

You guys need to go easy on the negativity. I think HP laughed at Woz's first computer prototype..

Fortunately for Woz, he probably didn't have all the hype in the world surrounding his product. Wolfrom might be an interesting experiment, but all of the Google-killer talk might make it into more of a Cuil.

But did we really expect this to be a "Google killer"? Nahh... My problem is that even when compared to the pitch made in the demo, even as a tool for simply data retrieval, it seems to fail.

Using it, you can tell there is some huge potential.

For example, these nifty queries:

us debt / us population / us life expectancy in days

http://www.wolframalpha.com/input/?i=us+debt+%2F+us+populati...

and

(MSFT net income / MSFT employees)

http://www.wolframalpha.com/input/?i=%28MSFT+net+income+%2F+...

us debt / us population / us life expectancy in days

What does that mean, to you, in everyday language? I see what WolframAlpha did with that, but what is the implication of that answer for a real-world concern?

It is the amount of debt each American has to pay off each day.

Except, you know, it isn't.

Assuming zero interest :-)

Now if we could tell it to consider interest then I would be impressed.

You'd need an interest function. Use historical data? Specify a regression? Have it try to calculate last average?

Even then, you'd need to have some objective value of worth to understand what $0.92 means. Right now, it only measures the value of the dollar. :-/

Some kind of debt slavery calculation? :)

It definitely still has a long way to go. For example, it cannot understand area Pittsburgh / area Kanazawa or population density Pittsburgh / population density Kanazawa which seems less complex than the above queries. Alpha is, however, aware of the populations of each city.

It still needs work for more complex queries:

http://www.wolframalpha.com/input/?i=debt+%2F+pop++japan%2C+...

So there's a small bug when you look at France. Also, it took trial and error to get a query that actually worked. (ex: comparing debt / pop / LE between multiple countries instead of debt / pop leads to more bugs)

I entered over a dozen queries, and only got a meaningful result for a single one. All others were completely unrelated stuff.

Based on the bad results, I can see that this is probably only useful to someone doing math stuff, anyone else its more or less completely useless

If this were an easy thing to do, don't you think Google would have done it a long time ago?

If Wolfram claims to have done it, don't you think he should have done it?

I wouldn't underestimate the intelligence of computer scientists working at Google.

I wouldn't overestimate them either. Intelligence isn't as much a factor in the fruition of this as ambition. Wolfram is more on the level of Sergey Brin and Larry Page than the people that work for them, ambition-wise.

There are lots of easy things that Google hasn't done. Some easy things aren't all that useful. I'm sure Google could beef up its calculator, but would that help them in any significant way? Probably not.

this is probably only useful to someone doing math stuff, anyone else its more or less completely useless

Some of the other examples posted as comments here look useful in disciplines outside math. I just put in a hairy system of two equations in two unknowns[1] made up randomly by my students, so that none of the coefficients or constants were preselected for easy computation. Alpha calculated the exact rational number solution set for x and y right away. Alas, the export results to .PDF link wasn't working just now, but that would be a cool feature for teaching my math class.

[1] 7672x+1357y=3, 63/8659x+1862y=5672

x = -9514160366/17670859355, y = 53828777381/17670859355

Interesting that the results are given as an image: http://ff.im/2W5UQ You think they did that to stop screen-scraping?

I think it might be because the output comes from Mathematica's engine, and the best way to render various formulas, tables, charts, etc. is to just use the same graphical output.

That's the first thing I noticed too after wondering why the fonts were rendering so poorly.

This is bad for both accessibility and readability.

Search for "wolfram alpha" -> http://imgur.com/mbda5.gif

At the same time "Armenian Genocide" is not a historical event, it's just a movie according to WA.

Yeah, it doesn't know what happened on April 12, 1961 and can't answer some other trivia questions. That's not a big deal though, feeding information into the system is far easier than building the system in the first place.

That's an explanation of how it is interpreting the input, not a claim to importance.

Everyone should stop being impressed that it can answer complex math questions. It obviously has instances of Mathematica available to it, that's probably the first thing Wolfram hooked it up to.

This isn't a search engine. It's a web-based expert system with strong data-manipulation and presentation capabilities.

If I understand correctly, most of it is actually written in mathematica.

True, it's an expert system and doesn't appear to be anything new in terms of AI, but it's still pretty impressive.

(edit: according to Wikipedia, "It is written in 5 million lines of Mathematica [...]")

I'm completely underwhelmed. Chalk it up to hype I guess.

I couldn't get anything for "Most popular names in 2008," and when I clicked the "Examples" tab and clicked Socioeconomic Data->Countries and used its default of France, it gave me three pieces of info (country code, full name and something else) and apparently tried to load some kind of information underneath, but it never happened.

Maybe it's because it's launch day.

Edit: Here's another thing that bothers me. It doesn't give you the source of its info. So when I click "Names" and have it give me info on a specific name, when it tells me it's the 8th most popular (in 2007, nicely outdated), I'd like to know how it deduces that rather than take it as fact.

Interestingly, there's a "Source information »" link at the bottom that doesn't work.

Think I'll stick with well-crafted Google queries and more primary sources.

Obviously it isn't a competitor for Google search, or any other web search, but it might be a competitor for similar products being developed by those companies.

For that reason a search company may want to acquire them either before they get too big and expensive, or to prevent a competitor from performing an instant catch-up in this market by buying instead of developing.

Therefore, I was wondering what the market cap was of Wolfram Research, since they aren't exactly a small, pre-profit startup.

This isn't a complaint about a lack of data, since that can be updated after the preview feedback allows them to adjust the more important technology; I just enjoyed the irony of it having no information about Wolfram Alpha LLC, and only providing some useless web traffic stats for wolfram.com when asked about Wolfram Research.

http://www.wolframalpha.com/input/?i=can+entropy+be+reversed...

It knows: http://www.wolframalpha.com/input/?i=what+is+the+answer+to+t...

http://www.wolframalpha.com/input/?i=what+is+the+airspeed+ve...

But it doesn't know "How do you get to Carnegie Hall?"

Perhaps a silly question: is it's longterm name Wolfram Alpha, or is the second word merely an indication that it's in alpha phase?

Wolfram is the company, Alpha the product name.

Maybe a good thing:

http://www.wolframalpha.com/input/?i=is+wolfram+alpha+self+a...

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

Creepy: http://www76.wolframalpha.com/input/?i=Are+you+self-aware%3F

I gave it "unemployment rate in germany in 1928".

It correctly parsed my query, but didn't have the data.

Google gives an answer (8.4%) but it doesn't come from trustworthy sources.

EDIT: "first prime greater than 1000000" gives 1,000,003.

I've been playing with this for a while, and Hans Rosling's TED Talk keeps popping into my mind (http://www.youtube.com/watch?v=YpKbO6O3O3M). It seems like the potential of this isn't from the search, but the much easier access to data. Once we start to see some fully developed applications using the API, then we'll begin to see a much better picture of W|A's true potential.

What a completely useless site. Unless you are careless enough to run javascript, the site will not return any information. They say "To see full output you need to enable Javascript in your browser", but really return NO information without javascript. I don't know what they return with javascript, because I don't use it, like any sensible person.

Searched for "india" -> Firefox + CPU got hanged. Bad omen for india I guess especially after todays election results.

My first search was "google" which sent the CPU to the skies. Had to kill Firefox.

killed firefox for me too, using 3.5

Grinds my Firefox to a halt no matter the query. Tried with Seamonkey and it looks OK.

maybe it is firefox-killer and not google-killer

I'm sorry but I see no connection between my results and the ones in the demo. I'm tempted to go back and mimic the queries from the demo.

It fails to retrieve simple data, let alone conduct computations with it. Could not find the minimum wage in my state. Could not retrieve average income. Basic CIA Factbook stuff...

I'm afraid I can't not stump this. And yet I feel as though I was the one in error...

I was expecting it to be a little more encyclopedic than it seems to be. I mean, math stuff if awesome and although I can do it all (i think) in Mathematica it's very good for the world to have a free tool that does those kinds of things.

But when it comes to facts... well, they're facts... You only get as a return the uninteresting information.

I think Wolfram Alpha can succeed if they point their strategy into something which is rare on the web: certified knowledge. Being able to type a name of a country and instead of getting facts, getting real information about its history. Information that the users can trust as a bibliographic source.

Because facts are facts... they exist but they're so uninteresting.

It gave me a load capacity error with a quote from 2001:A Space Odyssey, then directed me to a live webcast of the control center:

http://www.wolfram.com/broadcast/wolframalpha/

Random observation: the Wolfram|Alpha Twitter account follows the Hacker News Twitter account: http://twitter.com/Wolfram_Alpha/friends

It's a marketing technique. You follow an account with high number of followers and some of the will find you, like you just did, and maybe follow you.

Using poorly-rendered (but 'designery') images to represent textual results is infuriatingly stupid.

Especially when there's a FOUC -- perfectly nice plain text is replaced with gross slowly-loading images.

It's not necessarily stupid. Right now, Wolfram Alpha is being hit from everybody that follows tech news, and so it's taking a huge load of hits. The delayed loading doesn't come from the images: look at the page as it loads and you'll see that the longest delays are coming not from the process of loading the image but from performing all the necessary calculations in the background. That's the slow part.

"Infuriatingly" is a bit of a harsh descriptor, don't you think? There are a few things that make text better than images online, and loading time is rarely one of the factors. The first issue is indexing - but Wolfram Alpha's results are so specific to the question asked that there's not much being lost if they're worrying only about providing a useful service and not one that Google hits on. The second issue is interactivity - but they provide easily copied text, so they're taking care of that one. (That may help indexing as well.)

Meanwhile, providing information as images means no time fiddling with style on their part, and it means I can very easily transfer their generated image for use online or in files and have it look nice.

'Designery'? Do you mean, 'Styling that accurately and attractively shows the relationship between various pieces of text and their various importances'?

I was able to copy-and-paste from the result I got (as related in another reply in this thread), so I have no complaints about the format of the answers. If the answers hid the textual part of the answer so much that it couldn't be copied-and-pasted (as I frequently find in addresses on websites of businesses), I would be annoyed, but I can deal with Alpha as it now is.

My Alma Mater:

Input: Virginia Tech

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

What is this?! Amateur Hour!

Oh well.

It did manage to find Georgia Tech http://www.wolframalpha.com/input/?i=Georgia+Tech&asynch... (Even where I'm going next year, Emory University, is there...)

But no impressive details or analysis for any schools. 'Tis all rather disappointing.

Nice :P You can have wolfram and I'll take ACC titles and we'll call it even :o)

try "Virginia Polytechnic Institute"

I guess it needs more Google-style knowledge to accurately parse more queries.

After playing with WA for a while it seems to me that it is not so much a better search engine as it is a better encyclopedia. Many search results read like condensed Wikipedia articles.

Out of curiosity, where did you get the idea that it was intended to be a search engine?

From thousands of articles saying it's a Google killer. And I guess I'm not the only one who got that idea, because Google search for "wolfram alpha search engine" returns over 600.000 hits.

Also, it looks just like a search engine.

It could not answer my simple question: "Does P = NP".

Until then, it's useless :-P

Been reading Abtruse Goose? http://abstrusegoose.com/149

That's the first thing he though of too, apparently.

I tried P vs. NP, NP-complete, NP-completeness, and NP complete. IIRC, It only realized that the last one was in the field of computational complexity.

Of course it can:

http://www.wolframalpha.com/input/?i=p+%3D+n+p

Both "NYSE 1950-2008" and "Nasdaq 1960-2008" gave me incorrect values for the stock exchange volumes. Instead, the results were NYSE Euronext and Nasdaq OCX, respectively.

You mean http://www.wolframalpha.com/input/?i=NYSE+composite+index+19... ?

http://www.wolframalpha.com/input/?i=hi+there

http://www.wolframalpha.com/input/?i=how+are+you%3F

http://www.wolframalpha.com/input/?i=what+can+you+tell+me+ab...

Well, it tried out the famous (and probably apocryphal) correlation between hemlines and the Dow Jones index. It doesn't have skirt length or hemline data. Then I tried correlations between the Baltic Dry index and the S&P 500. It didn't know about the former.

I guess financial isn't their top priority. Maybe there should be a new group of experts focused on that alone?

No low-level API yet! Maybe it's just for paying customers? I saw no indications of supported syntax (how do you group?) or data schema (what is the smallest fact I can extract and how?). If the natural language processing isn't good enough to read my mind, at least tell me how to speak to it effectively!

Not really I got the http://www.wolframalpha.com/input/ once and when I entered a query it redirected me to the normal launching may 2009 site.

Maybe they are transitioning servers or something but its not up yet.

I was able to get the results of a single query, "our solar system" in extensive detail. Trying another query redirected to a "launching May 2009", after which pressing back landed me on a working search field, but mostly with not found results.

Brothers and sisters have I none...

http://www.wolframalpha.com/input/?i=Brothers+and+sisters+ha...

Hm. I've been avoiding the hype about this, but it fails on queries like "Chancellor of Germany" which seems like something this should do. Am I missing something, or is there just very little data in it?

Who else started with "air-speed velocity of an unladen swallow"?

I did. Also "how much wood would a woodchuck chuck if a woodchuck could chuck wood?"

Anyone else get an interesting error http://i44.tinypic.com/359es9e.png

yeah i got that once (out of 40 or so queries i messed with)

I love how they provide a link to the launch webcast along with the overload error message. Too bad it's not live anymore, though.

It's a good start.

Anyone else gettting unescaped \n in the results sometimes?

Yeah. It seems like they're sending the results back (with tons of \n's), then replacing that text with images of the same text. I wonder why they're displaying images instead of simply styling the text? Strange...

i guess that means i need to buy haswolframalphabecomesentientandtakenovertheworldyet.com

reminds me of hasthelhcdestroyedtheearth.com

the results are not searchable using the 'control + F' functionality of the browser.

Reply on news.ycombinator.com