Consider this.  You have a small group of loyal participants willing to protest against any particular event, on cue from the PPA, Paid Protest Administrators.  What to do?

Yesterday.  Write letters to everyone and every news agency to protest or complain.  Usually, small numbers are ignored implying the protest is ignored.

Today. You have programs that generate unique letters based on keywords and then send using thousands of correct email addresses (because you have access to data banks of addresses, on sale just anywhere).  Ditto for Twitter addresses.  This multiplies the small group by hundreds, thousands or even millions. 

For example, such a program could generate and send a protest letter from me (or you) on some issue I don’t care about or even disagree with. I would never know. Note, news agencies have not the resources to check legitimacy. They just count!  Recommend: News agencies create a form requesting confirmation of emails sent.  Easy to do, yes, but also easy to counter. In Congress, we suggest all members share emails with central units that scan for multiplicity and duplicity.  Alternate services such as Turnitin could be used toward this end.

There is morality. The new leftist Revolution USA may be orchestrated by a remarkably small number of people.  First, it is dishonest, but feeling you are right and just often overrides morals.  Second, using the machine-driven methodologies, it is devious. Third, it creates distractions from substantial issues that are lost in blizzards of messaging.  Forget the neo-Nazis in this as they are generally stupid and love physical activity. Letter writing in any form is not their game.

You talk about ISIS using the Internet in sophisticated ways?  They are mere amateurs compared to really serious players, who not only recruit but also mass message and try to control the agenda.

There is more.  Large service providers can create millions of false email accounts, and each of these can create accounts on FaceBook, Twitter, Snap, etc. (Remember Wells Fargo and their fake bank accounts?)  Hence, they are empowered to send thousands or millions of messages supporting or rejecting this situation or that.  Large organizations receiving millions of messages or multiple thousands will use deep learning to read the message for favorability or not, maybe keywords and themes, and then produce a count.  In this scheme, we have machines creating the messages and machines reading them.

The organizations could be CNN, FOX, or some random Senate leader.  The key point is with the volume.  It must be too large for the messages to be read in total, but not too large that the sheer count looks suspicious.  The volume must not be too small as then the message theme is ignored.  Completely fake websites can be created to recruit cadres of useful participants to attend rallies or protests of any sort. As well, these programs could also respond to blog entries in any desired fashion.

In this way, just a handful of people can shape the news or the Congress – or foment a revolution.

Indeed, the so-called Fake News may not be fake per se, but the organizations may be naively faked-out, reacting to a blast of fake messages. These blasts shape what to do and what they do. It even explains how they can suddenly move from one topic to a completely different one something like a school of fish simultaneously and almost instantly changes directions.

Turing Test Redux: The Turing test for a machine is set to task a person to decide by asking a series of questions if he/she is talking to a machine or a person.  The new test is this:  Can any organization by sending a single return message decide if the particular message was sent by via a machine-driven network or an actual person?  Probably not.

The great unknown in all this relates to knowledge, and that knowledge pertains to the sheer power of machine-driven actions, for which language is not a problem, for which numbers are not a problem, for which verifiability is difficult, and for which a presumed ignorance of how it is done is paramount.  The massive IT firms all know this well.

What is discussed above is not only technically possible but has likely happened. As framed here, all is conjecture, but possible. Verifiability is nearly impossible.  Who’s going to admit they did it, much less publish it?