Here's the bit about all of this data tracking that bothers me the most.
The data is, in large part, a one-way street. The companies which are most aggressive in collecting information are also the most aggressive in not disclosing how they use, share, distribute, aggregate, correlate, etc., this information. Look at the NDAs and "no talk" policies of Google, Facebook, any of the credit bureaus, to say nothing of the vastly less visible enterprises which service the B2B markets of data mining and information. It's an area in which I had some experience and washed my hands in disgust years ago -- it's also an area in which I'd very much like to leverage open channels and tools to provide the public with the ability to fight back against the problem.
I've had a very clear view for over a decade now that this will not end well.
The truth is that the public has little understanding of what data are being used, how, or by whom. Much more pernicious is how data can be aggregated. Your insurance company gets information about your car and license. Your smog check station runs a standard battery of tests against your vehicle and reports this to the state in a large electronic record. Your state turns around and sells this database, at a very, very low charge, to companies providing services to the insurance industry, so that based on VIN and license information, a huge dump of data form your car's onboard data collection systems is now available to your insurer. They're mostly interested in total mileage, but as car data collection systems advance, there could be a great deal more information there -- accelleration/decelleration, speed, conceivably in the future, GPS waymarks.
And you'd never know about it.
The mileage stuff? That's for realz and has been for a decade or more.
The truth? That few even within the datamining field know what others within the field are doing with data (see above WRT one-way information), unless that information is being directly marketed/sold. Where derived products (e.g.: risk/profitability profiles based on models in which individual inputs are not disclosed) are sold, even uses which are directly sold may be using information in way undisclosed to the data suppliers / users, let alone the members of the public who are being profiled.
> he predicts
He's avoiding it because he thinks it's the natural progression of the product these companies offer, not because they do, it might never happen.