..snip..
The problem, ultimately, is features. And it will always be features.
That is correct as an observation. I don't like it, but the world does not care whether I like it or not; I am 1/7000000000 regardless of my skill, taste, or persuasiveness. And so it has always been (cf. "bread and circuses"). However, what I object to is the tendency of features to destroy functionality by way of collateral damage, viz., for platforms to be constructed to deliver features and only to deliver features. That is what "freedom to tinker" fears. That is what risk is all about, risk being solely a consequence of that upon which you depend. That is why I've all but stopped buying new things (computers, cars, appliances, etc.) -- their orientation around features reduces my ability to configure, to repair, nay even to understand what is going on inside, much less that it is legally questionable as to whether I even own them despite having paid my money for them.[*] (Even were I willing to run Javascript, my old computers can no longer handle the burgeoning demands -- Javascript has clearly become the technologic embodiment of "When rape is inevitable, relax and enjoy it.") Big data, especially of the so-called deep learning kind, is of a parallel sort. Where data science spreads, a massive increase in tailorability to conditions follows. Even if Moore's Law remains forever valid, there will never be enough computing hence data driven algorithms must favor efficiency above all else, yet the more efficient the algorithm, the less interrogatable it is, that is to say that the more optimized the algorithm is, the harder it is to know what the algorithm is really doing. The more desirable some particular automation is judged to be, the more data (which is to say foodstuffs) it is given. The more data it is given, the more its data utilization efficiency matters. The more its data utilization efficiency matters, the more its algorithms will evolve to opaque operation. Above some threshold of dependence on such an algorithm in practice, there can be no going back. As such, if data science wishes to be genuinely useful, preserving algorithm interrogatability despite efficiency-seeking, self-driven evolution is the research grade problem now on the table. If science does not pick this up, then Lessig's characterization of code as law is fulfilled. In short, features drive. They drive because of democratic principles evidenced by immensely rapid uptake. They rely upon a user base that is forever "barefoot and pregnant." And it is increasingly difficult to opt out of features without opting out of society altogether. As there is zero difference between "personalization" and "targeting" beyond the intent of the algorithm, those who don't accept features will be adjudged anomalous, and we already treat anomaly detection as the sine qua non of public safety. --dan [*] http://www.wired.com/2015/04/dmca-ownership-john-deere/