Google Prediction APIs: My Prediction

I’m waiting for the assigned Google scribe to start generating notes for the session on BigQuery and Prediction APIs. Am I in the wrong virtual room?

In any case, a quick read of the documentation for Google’s machine-learning-in-the-cloud APIs makes me realize again the power of Nick Carr’s IT Doesn’t Matter proposition.

I have had some quibbles with parts of Carr’s big theory. But if Google is going to be the general store for all kinds of sophisticated analytical software then what does it mean for startups like Hunch and others that are developing their own proprietary recommendation code?

There’s really nothing stopping a few programmers in a dorm room from launching their own prediction system using Google’s web service-based supervised learning platform.

Obviously there are still ways that these types of companies can differentiate themselves, but I think they will have to say more than the deliciously vague  “we have a small gaggle of MIT computer scientists with backgrounds in machine learning.”

It appears that Google is, in a way, taking the excitement out of these AI-ML solutions, making them boring (as Carr might say), and giving more folks than ever the opportunity to claim that “my recommendation system has been worked out by Google’s quants.”

We’ll see what happens with this latest Google initiative, but my prediction is that
quite good recommendation plugins may soon be a standard part of typical business analytic software.

Reblog this post [with Zemanta]