Item — HN Reader

"exponential" might be an attention-grabbing stretch, but it is clear that we will discover new uses for this.

I am wondering if anyone knows the relative size of the data behind wolfram alpha compared to google. I know that wolfram uses "trusted" sources for its data.

I am recalling a recently quoted paper by Norvig that seems to imply that having multiple of orders of magnitude of data available changes the knowledge engine game.

1 comments

"Wolfram Alpha, the so-called "computational knowledge engine" that launched this week, claims to have access to a vast repository of information from trusted sources around the world: 10tn pieces of data filtered through 50,000 models and algorithms."

http://www.guardian.co.uk/technology/2009/may/21/1

> having multiple of orders of magnitude of data available changes the knowledge engine game.

Maybe machine learning as well...

"The cloud also makes possible our approach to machine translation in which thousands of computers process billions of words of monolingual and bilingual text to build statistical language and translation models."

http://googleenterprise.blogspot.com/2009/05/breaking-down-l...

I had previously missed the part at the end of the article where wolfram does suggest crowdsourcing as a way to tune, which is what google is all about, in some sense.

I am also intrigued with "50,000 models and algorithms". What would those be? Is this just something to impress the journalists?

Most of those are probably "models."

So I am wondering what is meant by a model by them--a row in a table listing constraints, a relation, a prolog rule?