…let's put it this way: comparing Google and WolframAlpha is like comparing apples and oranges. Because contrary to what we could read on some newspapers these days, WolframAlpha is not a web search engine. It's a knowledge engine, which means that where Google will return (indexed) content, WolframAlpha will return facts. Let's take an example with this query: “What is the first satellite?”

What Google does when asked what is the first satellite is to split the question in (key)words, discard the irrelevant ones (such as “is” or “the”) and look in its big index to find the best page corresponding to the remaining keywords. Then, what Google returns is a link to a web page, or to be exact many links (in the order of relevance). In this case, it returns (almost as usual) the Wikipedia page of Sputnik 1, which is already a very good result.

The first satellite is http://en.wikipedia.org/wiki/Sputnik_1. 
The first satellite is http://en.wikipedia.org/wiki/Sputnik_1.

The challenge is slightly different for a knowledge engine. What you want is a fact, not links. You cannot return many results with the hope that the answer will be among the so-called “10 blue links”. There is only one fact. And that’s exactly what WolframAlpha returns. It does not return a link to a page about Sputnik 1, it returns Sputnik 1 itself.

The first satellite is Sputnik 1.
The first satellite is Sputnik 1.

Ok it’s nice, but the customer value seems to be identical in the end: I entered the question in a box, I pressed Enter and I got my answer directly on the screen. So it’s not better than Google after all. But remember what I said at the beginning, WolframAlpha is not a Google-killer (yet).

So if it’s not a Google-killer, it’s not interesting, right?”

Hum… actually it is still very interesting.”

But if it’s interesting, then it means it’s a Google-killer?”

…”

I’ll go even further than “not a Google-killer” (yet). Currently, the results are still quite incomplete (just try “what is the second satellite?” if you want to convince yourself ;-). And the technology, although state-of-the-art, does not look so impressive; after all, the Wolfram folks are not the only ones doing natural language or semantic search.

So why is it so interesting? Well, WolframAlpha cannot answer all the questions, but it gives damn good answers for the ones it can! Even better, it shows exactly how it interpreted the question and the many relevant pieces of information related to the computed answer. And when it does not know the answer, it states it very clearly, which sounds better than returning irrelevant links (although it may look like a bad experience for some people).

And guess what? It’s nothing compared to what is coming. The potential of this new kind of search is really unlimited and WolframAlpha is just a next step towards a seamless access to information and content. Natural language (along with speech technologies, not featured here) allows great, almost “magic” user experiences. Semantic search allows queries with several levels of abstraction, interpreted directly by the engine. Combined, you could ask for much more complex queries such as “show me the pictures of the engineers who built the first satellite ever”.

Finally, it’s no surprise to see the digital giants moving also very quickly in this space. Steve Ballmer put more than 100 million dollars on the table last year to acquire PowerSet, a semantic search engine running over Wikipedia data. As for Google, there are already rumors about a pure semantic search feature (in addition to the first semantic tweaks released last March in the main engine, and the many instant answers that already exist).

But there is something that Wolfram folks achieved before both companies: they built a great proof-of-concept of what mainstream semantic search may look like in the next years. And that’s already something.

image 
WolframAlpha’s Holy Grail