in programming

Data is more agile than code

Peter Norvig talks about the need for a startup company to go fast – and also in the right direction – at his Startup School 2008 talk.

“Sure you gotta go fast, but if you’re not getting feedback to figure out if you’re going in the right direction it doesn’t matter how fast you go.” (2:47 in the video.)

That advice can apply to both technology and the business sides of a company, but here Norvig focuses on the feedback necessary to make sure the technology you’re developing succeeds.

He suggests you can get this vital feedback by:

“Acquiring lots of data [and] running machine learning over it… The key here is that no matter how agile you are as coders, and I understand that you’re all great, data is going to be more agile than code. Because you’ve got to right the code yourself, but the data you can leverage… there’s an immense multiplying factor that way.”

I guess this Lisp/AI guru from Google knows a thing or two about using lots of data, eh?

He goes on to describe how Google has used machine learning over large data sets for their image search, text segmentation and Google Sets. It’s a great talk, I highly recommend it.

I like the idea of letting the data and algorithms do as much of the heavy lifting as possible – the knowledge I want to share with my users may already be in the data, I’ve just got to dig it out!

Write a Comment