The latest buzz in knowledge systems is Wolfram Alpha. Let’s make this clear: it is not a Google-killer and doesn’t claim to be. It is billed as an “answer engine”, instead of a search engine. Search engines give you links to websites that answer your query; answer engines just try to give you an answer, full stop. Just to clear that up.
I’ll try here to find the border of its knowledge, just to have a good idea of what it can handle and what it can’t. So, let’s calculate the odds of a “royal flush” in draw poker:
All it understood was “royal flush”, and gave me odds for five and seven cards. No note as to whether this is natural deal, or discard and draw.
“Which country had the most wars?”
Here’s a more typical Wolfram Alpha experience: it’s stumped. No idea what I wanted. Neither of the provided links lead me anywhere near the answer I wanted. This is an example of a major problem with knowledge systems – they have to be ready for every wacky thing that users will throw at it. And this query is just the kind of thing that would be practical and handy to know.
Other queries which have stumped Wolfram Alpha so far:
- “calories in bat guano” – stumped.
- “world’s oldest piano” – stumped.
- “alternatives to photoshop” – stumped.
- “map of Dictionopolis” (It interpreted it as “Divinopolis”) Dictionopolis is the first city in the book “The Phantom Tollbooth” by Norton Juster.
- “map of Mordor” (did you mean “Gordon, Nebraska”?) Look, if I wanted MapQuest, I know where to find it.
- “number of teeth in adult”
- “who is charlie brown?” It gave me the United States musician Charles Brown, whoever the heck he is.
- “lifespan of a cockroach” – totally stumped!
- “microchip inventor” – stumped. Here’s the exact same string in Wikipedia’s search, with the correct answer as the second result. And just clicking ‘microchip’ gives you only the company, “Microchip Technology”!
- “who is winston smith” – It gave me statistics for both surnames, unware that he’s the protagonist of the novel “1984“.
Now before everybody goes, “That’s not fair! You’re trying to trick it!” Well, the PR has gone out of its way to document these silly Easter Egg examples. Somebody spent far too much time trying to rig this demo and too little time making a useful tool. And before anybody cries foul on my searching for fictional characters, it sure as heck knew “who is john galt”:
You know, because if you want the social web to hype your engine, you’d better cater to the Randroids.
I had better luck with straightforward calculations. But how far could I push that? How about converting between two measures of length, using the most arcane units I can think of…
Ta-daaaaa! I was almost accidentally impressed for just a minute there.
Bottom line: This thing has a long way to go on all fronts before it even comes close to what we want an answer engine to do.
- Playing with the Google calculator
- DBPedia – a New Way to Play with Wikipedia
- Google and the Concept of a Landing Page
- Search Engine Study – part 2: History of Early Search Engines
- Why Search Engines Work Like They Work