Wednesday, March 11, 2015

Cracked Article+

As many of you know, an article that I wrote for Cracked was just published this Tuesday.

There were a few things to add.

Firstly, the editors threw in a reference to the data broker Acxiom, which they said had personal data on 190 million people. This is actually a very old statistic. As of 2013, Acxiom claimed to have "3,000 propensities for nearly every U.S. consumer", and multi-sourced data on about 700 million people worldwide

Since the article has been published, a few people have told me that they have not been particularly impressed with the demo of IBM's Watson, which IBM claims can detect personalities from just a few hundred words of text.
(By the way, I'm curious if IBM will notice the sudden influx of visitors to the demo from Cracked and panic, shutting down the page. Your move IBM PR department.)
Anyway, a number of people have told me that they had sent writings to Watson and found the analysis less than impressive.
Firstly, I wouldn't be surprised if IBM is exaggerating Watson's capabilities in this regard. They have to sell the idea to investors and clients, so a little stretching of their capabilities is par for the course.
On the other hand, a single writing sample from someone is like a single data point. It's just from a single slice in time on a single subject with a single purpose, so it's not surprising that submitting that alone to the demo might throw off Watson's algorithms.
On the contrary, social media like Twitter, provide data points about users from many different time periods on many different topics, so analysis of them might be far more predictive. The partnership between IBM and Twitter announced last October should certainly give them more access to that.

1 comment:

  1. New technologies are being created in response to this threat, both to negate data security risk and improve the digital customer experience.
    iDeals data secure solutions

    ReplyDelete