Stephen Wolfram’s Introduction to Wolfram Alpha (screencast) | Hacker News Reader

Stephen Wolfram’s Introduction to Wolfram Alpha (screencast)

blog.wolframalpha.com

126 pointscallahad41 comments

Couple of comments.

First, it's an awesomely cool idea. Congrats to the Wolfram team for pulling it off so far.

Second, it's a freaking huge project. As much as they have, I can't help but think they're only about a 100th of the way towards something that might be truly comprehensive. Maybe a 1000th of the way.

Still -- kudos for setting it up. I sincerely hope it grows and becomes all that it can be. It will definitely take web research to a new level.

Mathematica itself has a rich plugin (packages) model that extends it to particular domains. It's a very important part of the system, sort of like extensions make Firefox.

Perhaps they'll open Alpha up in a similar way to help make it comprehensive. Can you imagine how powerful this thing could become if third parties are also developing for it?

There's a Q&A with press and someone asks this questions and Wolfram essentially says that yes they very much want third party participation, the only caveat being that there would be some manual review process of the data for accuracy.

So ultimately I think thats what this is like a Google Base + computation where they have added (hopefully) enough of the core to push it over the hill.

A place to put structured data where it can be fully leveraged minimal effort from the data provider.

Impressive technology, BUT... I've just tried Wolfram Alpha with a preview account, and I have to express my complete disappointment. It seems to me that Wolfram Alpha is a classic example of geeks building application that is useful for them but that is irrelevant to the other 99% of the world. And that 99% are the buying customers. I tried some usual web search phrases, and for each of them "Wolfram wasn't sure what to do with my data". It even suggested some alternative searches, but when I clicked, it still wasn't sure. Of course, it was happy to analyse some sinus function for me, but imagine how many people would like that? In my opinion, Wolfram will be an excellent niche search engine for mathematicians, statisticians, and the geeks alike, but nothing near Google in any respect.

This and Google are serving two different purposes. Google is an interface to search the web. Wolfram Alpha is an interface to search their own database of facts and relationships. When you use Wolfram Alpha you are doing a fact search, not a web search.

The economic and financial data alone will give Wolfram more paying customers than he can support. Just think about Bloomberg.

I am puzzled that people think this is a Wikipedia or Google killer. I think it is much more disruptive to general websites. This project makes many websites obsolete for many purposes. If people now want to know how many calories are in a product, they go to a website that specializes in that (directly or they find it through Google), but why do that if Wolfram Alpha gives you that information quickly, and presented in a more interesting way? The same goes for many of the topics he presented. Baby names are also popular website topics, but none of the specialist babyname websites are as comprehensive as a Wolfram Alpha search on different names.

I predict it doesn't obsolete anything except for power searchers and data researchers.

People are still going to type "how many calories in a big mac" into Google and get specialized sites that answer those questions fine. I don't think they'll go to WA when they have a habit and good success rate with getting those casual answers on Google. I think professionals and power users may use WA in cases they know it will give them great responses, but I don't see it changing many casual user habits.

What would be interesting is if WA actually caches search results and makes them indexable by other search engines like we do with bug.gd (which is similar in the sense that bug.gd is a specialized search service). This proves to create a tremendous amount of traffic and could make the provided/refined content much more likely to obsolete those niche sites.

I don't know about that. I think users will catch on. Think about Weight Watchers, for example. I can get a whole nutrition label for the lunch I just ate in one place at Wolfram alpha. With google, I have to add it up...

I also think users will catch on. Maybe not immediately, but google and wikipedia were also just for power users at first. The benefit is so great, and it is so easy to use, that people will talk about it on forums and explain how it works. They may not at first use it as their general knowledge tool, but they will use it for the information they want most.

if Wolfram Alpha gives you that information quickly

Scaling will be a problem. It's doing more work than google, but without their datacentres, parallelization techniques or incremental experience. Being fast was at least half of google's success. (I agree it's not competing with google; it's just an example of scaling well.)

But, on Google, did you know they acquired http://www.gapminder.com ? And, if they get mobilized, why couldn't they follow where Wolfram has led - but with faster results?

If nothing else this is a gorgeous application -- it looks great. And in fact it looks like something I might use as often as Wikipedia (though not as often as Google).

I can't wait to play around with it; some of these examples (nutritional labels, the actuarial stuff, ISS location, etc.) are just dead cool.

Wow. Don't know about you, but it seems to me that Wolfram to Google is what Google was to Altavista. A principally different search paradigm that may yield dramatic increase in a search quality. It's no longer a web search, but rather an information search.

It seems like it fills holes that Google has rather than acts as a complete superset of Google functionality. I use Google for a lot of the queries they demonstrate, but I have to click-through and dig further to find my answer. With Wolfram|Alpha, there is no need to dig. Yet, it doesn't return "regular" ranked results, which don't require computation -- hence, it's not a superset, and not a killer.

I can see Google responding with a similar offering, especially with their impending Squares release. It'll be interesting competition flaring up in this space.

I was tempted to agree there -- but that's the thing, whatever comes along and replaces Google won't look like Google, and we probably won't even notice it happening.

I don't think Wolfram Alpha is that. I think its goal is honestly too ... computational, logical, factual ... to have the broad appeal of Google. But I don't think that because something is fundamentally different than Google means that it's not the fabled Google killer.

Come now. The average user will continue using google. This is interesting, and maybe scientists will use it, but it's not going to be of any use to the average googler.

Would love a few folk who know how this works to clue me and anyone like me in on what this could be doing in five years?

Take a DVD player as an example, if he was inclined to load the data I could put in a model number and it could show me the technical specification, picture, manufacturer etc.

If I put in two model numbers he could compare them. Could he include in the contents things like product reviews? (is the only problem a trusted source for the data?)

Or, he seemed to have a link for movies - presumably if I type the name of a movie he can list actors, year of release, sales gross etc. If I put in a band can he tell me whether they're touring, back catalogue?

Its obvious from the demo he hasn't been building something to help me pick a DVD player to buy, but does anyone know if this system is likely in the future to support this kind of thing that would be useful to the layman? because that's what might scare Google (or not if I'm barking up the wrong tree here).

I've just tried Wolfram and I would say that Wolfram is to Google as a Statistics section in a library is to Google. In other words, nobody except geeks will care.

You know what's more awesome than Mathematica? A free Mathematica with toppings (curated, structured data.)

I hope they release a good API or just a formal query language, so I don't have second guess the NLP crap.

HOLY COW! WOW!!!!

I WOULD PAY $100 A MONTH FOR THAT.

I can think of at least 10 different scenarios where I would have (tried to)used that and saved at least 2 hours per scenario.

My head just exploded - seriously.

Google will probably copy from Wolfram Alpha those things that are likely to be of interest to many users. And it will do so using automated scalable methods.

I think they are out to become like Google for a very technically and numerically inclined subset of the populace.

If this is not all just rigged demoware, then he will succeed, and he's going to make a killing! He will benefit from the same incumbency factors that Google enjoys now, and he'll be a leader in the industry he is about to create.

If it's rigged demoware, he might fall flat on his face.

Or Google buys Wolfram Research? I hope not, but how much would it be if it happened?

Stephen has said many times that he will not be bought out.

The important thing to understand is that Stephen Wolfram built Alpha as his personal "publication platform." He is going to use it to in effect "publish" his ideas about NKS and computation in general into immediately useful form, to the entire world. Rather than write papers.

it would be very odd if he went around saying he's hoping that google and msft will start a bidding war for his company.

Hehe, true. However you need to take into account sw's career. Its pretty obvious if you look at his record that he is in this for the long term legacy.

Alpha isn't just some web startup looking for an exit. Its the culmination of a life's work in computation, and the offspring of a 20+ year old company.

the real question now becomes where is the information WA is getting its results from and how is that chosen as the premier source for a particular query?

Google has been doing simple things like unit or currency conversion, ticker symbols, basic geographic facts, arithmetic expressions, etc for a long time now. WA is nothing more than just a very, very advanced version of that, plus charts. Well, it is certainly impressive at least in terms of human work that made this amazing service possible. But from the engineering perspective... don't know, not impressed that much.

I'm not sure who started the hype about WA as A.I. or Google killer or iPhone smasher or an immortality pill, it's certainly less than all these things. Just a pedantically well executed service. We'll see.

The engineering used to simply parse the queries is amazing. It's lightyears beyond Google's NLP.

What exactly did amaze you in this demo?

One thing that amazed me was when he searched for that DNA sequence, and then traversed up and down a chromosome. Mapping out the human genome has been done for some time, but I've never actually seen any sort of interface or app that hooks into this sort of data, and this does.

I can't tell you what revolutionary applications or mashups will make use of this data, but I can definitely say I was 'amazed' when he did demo that part of alpha. It's like I'm searching my body!!!

warfangle said "It's lightyears beyond Google's NLP". I'm afraid there's nothing NLP about WA, neither Google. It's just very basic parsing in both cases, if you know what I mean.

Not only chromosome traversing, there were a few other impressive things in the demo, but that's some level of structurization and formalization of certain knowledge domains done by respective professionals. Good job. No, very impressive, stunningly scrupulous job. But I see nothing revolutionary (edit:) technology-wise.

I'm afraid there's nothing NLP about WA, neither Google

You might be mistaken there, Google very much uses NLP , though at the backend and not at the user interface.

Norvig himself said so ... I cant find to the exact talk where he said this( It was one moderated by Nova Spivak,if I can remember correctly) but here's a link to a TC article that discusses this ..http://www.techcrunch.com/2007/12/18/googles-norvig-is-down-...

every single query.

Do people actually use natural language in search queries though? I don't think they do.

Technically impressive? yes. Useful for most people? nah.

You may not, since you understand the concept of natural language in the context of search queries and that it's a hard problem.

I would wager that a non-trival subset of users in fact do use natural language in searches, especially in the context of the 5 w's; "what is...", "where is...", "when is...", etc. These aren't technically more difficult to understand than your average search query and I would imagine with the volume of google's searches, they do a good job of dealing with these types.

I have a weight loss website. While most people arrive at my site with queries like "calories in apple pie", a significant number of people use queries like "how many calories do i burn by doing the dishes?" I never see double quotes or +/- signs in the queries.

I think this is the search system Wikipedia has been needing to sort its data. This is not a tool for discovering knowledge like Google is.

This is the most persuasive argument for structured data I've seen so far. Google's biggest limitation is that it can only help you find things that someone has put on a webpage.

As interesting as the web application looks the APIs would take this to the next level.

Absolutely, completely amazing! I hope they launch a solid API!! preferably in RDF...

Reply on news.ycombinator.com