Welcome to WebmasterWorld Guest from 18.104.22.168
Forum Moderators: bakedjake
Excerpt of a blog post by the author:
"All one needs to be able to do is to take questions people ask in natural language, and represent them in a precise form that fits into the computations one can do."
"I’m happy to say that with a mixture of many clever algorithms and heuristics, lots of linguistic discovery and linguistic curation, and what probably amount to some serious theoretical breakthroughs, we’re actually managing to make it work."
Wolfram Alpha will answer questions directly.
Not so good.
I just tried asking
"What is a smartphone?"
and it returned
"Did you mean:What is a microphone "
I then tried
"How do I say hello in German?"
and it returned
"Wolfram¦Alpha isn't sure what to do with your input."
"Related inputs to try: German"
It's the same with practically every simple question I ask it. Ordinary human language just does not work.
Now you might say I'm asking the wrong questions, but those are the kinds of questions people type into search engines. You can see those kinds of questions in any major site's visitor log, and Google's automatic query suggestions list those kinds of questions too.
If Wolfram Alpha can't match the hype it is currently receiving, it will end up just like Cuil.
For example "How wide is the atlantic ocean" returns "Wolfram¦Alpha isn't sure what to do with your input."
Even when there is an easy straightforward numerical answer it simply doesn't have the data. This is not a useful search engine.
For example "How many cars did ford sell in 2007" returns "Wolfram¦Alpha isn't sure what to do with your input."
"How tall is president obama" returns "Wolfram¦Alpha isn't sure what to do with your input."
"How much does a mobile phone cost" returns "Wolfram¦Alpha isn't sure what to do with your input."
"How heavy is a blue whale" returns "Wolfram¦Alpha isn't sure what to do with your input."
WA is nowhere near ready for the public. It's being totally mis-sold by the hype, and people will be disappointed because of it.
It's basically just a statistics database with lots of statistics missing.
Even the stats that are there aren't adequate, for example the pizza link gives average nuitritional data values across many pizzas but fails to list how that sample was collected. Is this average in America? Europe? Asia? Which brands? Frozen or fresh? etc.
Whatever the merits of WA, they should not have hyped it as a search engine because it isn't a search engine.
I actually did, the point of saying answers and wikipedia is that i wouldnt bother as using it as a starting point. Its something other engines do alot better, im being critical because i dont think its really that novel. Id rather see infoseek from 10 years ago come out of retirement!
Wolfram Alpha will answer questions directly.
You are confusing wolframalpha with what ask jeeves used to be. A search engine that answers questions directly in the sense that you can type in full sentences.
Wolframalpha answers questions directly not in the sense that you type in a sentence but in the sense that it does answer factual queries directly by computing the answer from structured data out of its own database, not by providing a list of documents or web pages that might contain the answer.
It is not a search engine like google, there is no crawler, that crawls the web for information.
An example: If you search in google for "population usa", google will return websites that contain the words "population usa". You can then visit the website and will then find the answer - or not find the answer on the website.
Wolframalpha works different. If you type in population usa, wolframalpha will look up the information in its own database, and create tables and graphics of this information on the fly.
Let's say you want to compare several countries by population. In wolframalpha you can type in for example
population germany vs usa vs mexico
And wolframalpha will compute the answer along with graphics from the information within its internal database.
Or let's say you want to know what the weather in new york was like, the day Paris Hilton was born.
You would then type in for example: weather new york the day paris hilton was born.
And wolframalpha will then check the birthdate of Paris Hilton and look up the weather in New York on that particular day from its own database, and then display the information in a structured form.
If you want to understand it a little better, check out this short video introduction, because if you use it like a normal search engine you won't get far:
[edited by: jecasc at 12:58 pm (utc) on May 18, 2009]
Why is it being discussed in an "alternative search engines" forum?
Why is the BBC (and almost all of the media) reporting it as "A web tool hailed as a significant rival to search giant Google has gone live to the public"?
Whoever's fault it is, WA is being totally mis-sold to the public as something which it is not. WA has done a very very poor job of managing expectations.
It is not a rival to Google, it does not answer questions, and its database does not contain even a fraction of the information needed to answer general queries. Yet all those things are being said about it on news reports, and that's how WA will be judged.
--You are confusing wolframalpha with what ask jeeves used to be. A search engine that answers questions directly in the sense that you can type in full sentences. --
I didn't say Wolfram Alpha was like that, I said it was being HYPED like that.
The very first post on this thread quoted its creators as saying on the WA blog that:
"All one needs to be able to do is to take questions people ask in natural language, and represent them in a precise form that fits into the computations one can do. ... I’m happy to say that ... we’re actually managing to make it work."
...which sounds exactly like the old Ask Jeeves.
--because if you use it like a normal search engine you won't get far--
While that may be true, WA is not communicating this effectively to its users or the media.
That's why it is going to be the biggest search-related PR blunder since the Cuil debacle.
WA is being totally mis-sold to the public as something which it is not.
I agree on you on with that. But that's not the creators fault. In fact you can ask wolframalpha itself what it is:
And it will tell you its a computational knowledge engine.
And that's exactly what it is. It has all sorts of facts in its database and is able to compute, combine and display those facts.
And the more the database grows, the more powerful this will be. It's more like the ship computer on Star Trek than a search engine. Or at least it could be once the database is large enough.
And wolframalpha does understand natural language. But of course only related to facts. So you can ask for example: What was the weather in New York on December 21st 1997?
But not: Was the 21st of December a beautiful day?
If you're going to do a serious site based around precise scientific data, you have to give your sources and methodology.
Perhaps you should check the "source information" button at the bottom of each answer. The sources are cited for every single query.
It doesn't give its sources on most of the results I've had, so I have no idea how reliable its knowledge is. That makes it scientifically useless, even worse than Wikipedia in that regard.
--I agree on you on with that. But that's not the creators fault. --
The site's About page begins like this:
"Wolfram¦Alpha's long-term goal is to make all systematic knowledge immediately computable and accessible to everyone."
How does that About page sound any different to Google's mission statement?
First of all, no one is going to notice that link.
Secondly, on the searches I've made, most of the "sources" are the domain names of various sites, without actually going to the specific pages where the data was taken from.
Also, the sources are for the entire page, without saying which data came from where. In practical terms it's impossible to check the data on its original source.
And as I said before, the query "make a pizza" (which WA suggests itself in response to the question "how do I make a pizza") gives me this page, which contains absolutely no information on how to make a pizza at all:
One of the sources listed is this:
The sources box actually says: "The inclusion of an item in this list does not necessarily mean that its content was used as the basis for any specific Wolfram Alpha result."
The most shocking for me is that WA doesn't know what Eurovision is. It's been the most popular television event in Europe for about fifty years now, and is watched by audiences in the hundreds of millions worldwide. When it was held in Russia last weekend, Vladimir Putin personally supervised the dress rehearsals to make sure it would all go smoothly.
For Americans reading this, it would be like WA not knowing what "superbowl" means.
Actually I've just checked and WA doesn't know what Superbowl means either!
EDIT: This is getting to the point where it's just laughable. :-)
A search for "Firefox" on WA leads to a page entirely about a 1982 action film.
Yes, reliable stuff indeed.
Let's face it, however clever the theories behind it, the actual product is useless for the average consumer. Almost any query entered by ordinary people is likely to result in nothing being found.
It's useless for the serious academic too: There is no way to tell what comes from what source, and some of the sources are clearly laughable. For those who missed it, a search for "battle of waterloo" has a list of sources which includes an article about the 50th anniversary of the Barbie doll. It's as if the sources are just being displayed at random.
And business people aren't going to be too happy with it. It lists Microsoft's stock price but doesn't list Microsoft's board of directors. What kind of commercial tool is that?
Whatever target audience this was aimed at, it seems to fail dismally. Whoever decided that WA was ready for public consumption has made a huge mistake which will probably be impossible to correct.
WA can't even do simple addition:
400 UK pounds + 400 UK pounds = £800.01
EDIT: More howlers...
According to WA, Wales is a small town of 6000 people in the middle of England:
And according to WA it seems England is the same thing as Britain:
These aren't minor errors, these are atrocious collossal mistakes. Not even the lowest-level geography student could possibly rely on something with such bad data.
Which in itself is really good. Anything to save to time in the science industry will help push us forward at a slightly quicker pace.
The problem is, most questions posed is not about science. When it can understand and give answers to the more common questions it will then give Google a run for its money.
I've had the same ride than gibbergibber did and the results are simply and plain ridiculous.
It makes me remember to that programmer who, explaining his programa to users, suddenly stops and says "Oh... no! You can't use this program if you don't know Ctrl+Caps+F11+PrntScreen makes it work in full screen! Even a five years old boy knows it!"
This all reminds me of the "discussion" I had with a customer when working as a club bouncer during the last big recession. Time Out had given us a gratuitous write up but had got our prices wrong. As far as he was concerned the big name magazine must be definitive.
It's not even a science engine though because it won't give any sources, so it's useless for anyone writing an academic work. If you can't cite a source, you probably can't use it in academic life.
There is a sources link at the bottom of each WA screen, but the sources lists are totally useless as they don't tell you which facts are from which source, many of the sources are totally irrelevant to the selected article, and many of the sources are not reliable.
And many of the facts that WA contains are plain wrong or out of date, so they would actually lead to bad science if someone relied on them. See earlier in this thread for examples, or take a look at WA's own "community" link.
Their geography data seems to be particularly bad. I don't know what source they got it from, but it can't have been an academic one because the mistakes in it aren't ones that any genuine geographer would make.
--This all reminds me of the "discussion" I had with a customer when working as a club bouncer during the last big recession. Time Out had given us a gratuitous write up but had got our prices wrong. As far as he was concerned the big name magazine must be definitive. --
That wasn't your fault though. The problem with WA appears to be at least partially WA's own fault.
It's not just the media exaggerating stuff, it's about WA itself exaggerating its own abilities.
For example Mr Wolfram himself has talked about WA using "probably 90% of the shelves in a reference library", which implies that they've encoded 90% of the knowledge of a typical public library's reference section (Britannica etc).
But that clearly isn't the case as WA is full of huge gaps and very bad data which no reference library would ever contain (I gave detailed examples of these errors earlier in this thread, I won't repeat them or give any more examples as the mod asked me not to).
Updates include more geography and planetary data, more Maths computations, more historical events and improved linguistic handing.
[edited by: Seb7 at 6:48 pm (utc) on June 9, 2009]
Wolfram Alpha - First major update
And more, this is the second or third failed big alternative to Google and maybe even low profile Internet users have noticed it. Next big alternative to Google will need to build much more hype around itself to receive such mass media coverage.
You might want to read up on it a bit more.