Before you read this there’s two things worth knowing about me – I’m very hard to impress with new technology, and at DARPA (where I was a Program Manager for 3 years), I was known for hating demos unless I could get hands on.
Therefore, when I was invited to attend today’s webinar demonstration of Wolfram|Alpha, the new Web “computational knowledge engine,” I didn’t expect to be very impressed. On the other hand, since the demoer was going to be Stephen Wolfram himself, and since there’s been a lot of buzz going on about the system, I figured I should watch. I’m really glad I did.
I was an early user of Web now-giants like Google, Facebook, Wikipedia and others, yet I cannot remember a time when I was so excited. Not because the product is so obviously ready for prime time (hard to tell that without putting one’s own hands on it), but because of the potential it opens up.
The idea behind W|A is easy – the company collects, curates and manages huge amounts of data about all sorts of things – including scientific domains (I noticed Physics, chemistry, mathematics, earth science) – as well as about general stuff (people, places, culture and media) — lots of areas where there is a lot of data. The system then functions like a search engine over these things, but instead of bringing back results of documents, it finds appropriate data, and processes it (using Mathematica as a back end apparently) and provides relevant answers and related facts, in various math, graphics and text form.
So if you type “GDP France” you get not just the number you might find elsewhere, but a graph over time, different currencies, etc. If you ask “integrate x^2 sin^3 x dx” you get the function, and the graph (and if you want, you can see the steps used in computing it). One of my favorites was “Height of Everest / length of Golden Gate Bridge” (the answer is 3.233).
Impressive also was the query "ATTGTTAATTGGTTAA’ which returned a bunch of biological information about this particular sequence (and changing a character or two clearly changed the result – we’re not talking about a fixed set of inputs best I could tell)
I could go on for a long time – “2.5 molar H2SO4”, “uncle’s uncle’s grandson”, “3,7,15,31,63” (which showed the next entries and a for the series) and much more. In fact, what made it impressive was how many different things it was able to handle, and the quality of the return. (oh, and “ISS” was pretty cool as it showed the current position based on getting the data from an appropriate web site and then computing the known orbit of the space station).
OK, is Google going out of business? Is Yahoo! done for? – No, W|A has a lot of work to do in improving its handling of queries; increasing its data sources; and working on its integration of sources across various fields. At the moment it looks like it would be a useful tool for some fields, and mainly a play toy beyond that — at least for now.
But the potential is incredible. I really feel like it ushers in a new generation of Web applications and opens the door for getting people to realize that search is only the very beginning of what the Web is about.
So keep your eye out for the public beta, and play with it — it may not be what you need tomorrow, but it will likely be something you use a lot someday…
-Jim Hendler
ADDED NEXT DAY:
I just found out that last week ReadWriteWeb got permission to post some screen shots – they will give you a better idea of what this is all about.
I agree it sounds exciting, but presumably the types of query that are returned are hard-wired, so how does Wolfram|Alpha deal with queries that are not in the area it knows about?
I’ve been very impressed with Wolfram ever since I read A New Kind of Science a few years back. I’m also curious about what Bob asks. Is the system able to learn and innovate based on earlier queries?
From what I have read and seen I am liking some of the hype, but really want to kick the tires. Especially in-line with comments like Bob’s.
Just adding my name to the list so that when you respond to my question (same as above), NN will let me know…
Any idea what the timeframe for the public beta will be?
Soon after I started blogging (in 2005) I came across the secret of the internet, OWL (Omnipresent Wisconsin Librarian), a person who essentially powers the blogosphere. One of his interests is Wolfram/Alpha, so do feel free to sign up to his Friend feed group, where you will find links to articles about this innovation. Some of them may attempt to provide answers to these questions. (via the link I provide, you can also find OWL’s old “blog”, Librarian’s Place, which also contains links to articles analysing Wolfram/Alpha.)
Certainly looks intriguing – any idea when it’ll be let loose?
Answering the questions so far:
1 -right now it tries its best to guess what function/dataset best responds to the question you asked. Stephen got several “nothing in the dataset” responses to queries we asked him for during the demo. However, there is a page that says what the coverage is, and it looks pretty good – unfortunately, we aren’t allowed to show any screen dumps of what we saw as part of the conditions of the demo, so I could not get the full list – but it cover a lot of science.
2 – Stephen said at this point the functionality is there, but they are launching with about 10,000 computers on the back end, and they are still getting everything coordinated. He thought within two weeks, end of month at outside (But I’d add that these things do slip most of the time, so we’ll see)
3 – Craig, I couldn’t agree more – I too want to try it – my suspicion is a small set of users will find it phenomenally useful (but on the Web “a small set” can stil be a very big number), a medium set will find it useful, and most of us will think it is cool, but only go back from time to time when we need. Of course, that was also what I said (and it turned out to be more or less the case) when Google first came out… now it is my home page…
-JH
(and Maxine, thanks for the pointer)
Ahh, I just found out that last week ReadWriteWeb got permission to post some screen shots – they will give you a better idea of what this is all about.
On one of the screen shots it says that only 93 people in Vatican City use the internet. The lowest in Europe. Why am I not surprised?
OK, it’s crap. On the screenshots they query “Bob”, and neither myself or Twin Peaks comes up. :-(
I still want to play with it when it comes out, though.
Hmm, neither of you comes up when I Google for Bob either, wonder if that’s a good or bad sign :-)
What impressed me most is the data center – although I think it should be a reasonable thing for the system of this scale. To get system scalable on the web, a data center is also a must (of course not all) – yes, if you are familiar with what I’m doing lately, I’m thinking about semantic wikis.
Another thing I’m thinking about when I saw this article is the Google’s recent post The Unreasonable Effectiveness of Data . Algorithm may rock – if it can handle real and real amount of data. That’s something so much semantic web efforts are not (yet) successful about, and W|A is about.
I tried it today, and discovered that the phase of the moon on the day that I was born was: “waning crescent moon”. Apparently, Woking is 9743 km from my place of birth.
I like that they have made it so that people can contribute/participate:
Participate in the Wolfram|Alpha Project.