The Dominion of Machines Begins When You Fail The Holden Test
The Holden Test is taken by any human who must judge whether a particular actor is the same as themselves, or merely machine.
“Let me tell you about my mother.”
These are the last words heard by Detective Holden, Blade Runner, before he is shot by Leon, a rogue replicant. The second shot blasts Holden through a wall. In this way we are introduced to the primary conflict of the film: human’s innate capacity for violence balanced against their capacity for creation, played out in countless ways through the limited time frame of our lifespans.
When a person makes something, s/he funnels a bit of his/her inherent existence into that object, and controls its fate—until that production is subsequently bought, sold, trashed, or released out into the world. This power of creation contains a right of destruction, and we take the right to kill our objects very seriously. To unmake something is the dark side of making things, and we hold this threat over every object we control.
It is this unmaking that Holden set out to accomplish, with Leon as his target. Leon is a machine—a replicant designed to be similar to a human in every way, except one. This one difference is that he can be unmade by a Blade Runner like Holden if he violates the terms of his existence by attempting to be too human (in the case of the film’s plot, by trespassing on Earth). The same market that brought Leon into existence also created the need for Holden’s existence. For everything that is made, there is a potential need to unmake it.
But given the sort of object that Leon is—a machine meant to be indistinguishable from humans—Holden’s first task before the unmaking can commence is one of identification. He must determine that Leon is a replicant by administering a test, and Leon will attempt to thwart that identification by passing the test, to remain indistinguishable from all the other humans that Holden does not have the right to unmake.
In the film, it is known as the Voight-Kampff Test, or VK for short. While this test is designed particularly for the attributes of replicants, it has a parallel to a real life test: The Turing Test. An AI passes the Turing Test if a human being cannot distinguish the difference between the AI and another human, given equal testing conditions. In this phenomenological definition, if it looks like a human, talks like a human, and acts like a human, all to the satisfaction of a previous-defined “real human”, then for the purposes of the Turing Test, it might as well be human.
The assumption of the Turing Test is that even though we do not have Blade Runners stalking the rain-soaked streets of the real world, that the goal of any particular object sitting down to the Turing table is to pass the test. We humans assume, having granted the dominant power position to make/unmake things to ourselves, that passing the Turing Test is a reasonable ultimate goal. Currently, all serious applicants to controlled, experimental version of the Turing Test have nothing more at stake than trying to pass the test, to bring accolades to their human creators. Notwithstanding the need for real Blade Runners in some possible future, we currently have no machines that have needed to pass the test to escape some punitive event stemming from the result of the test. All Turing Tests administered are for the inherent value of passing the test itself, and so the consequences of the test result are mostly of novelty or research value.
However, recently we have begun to see instances, in a more everyday social setting, where the goal is to fail the Turing Test. Take @horse_ebooks, a Markov chain Twitter bot that, after rising to fame for creating peculiarly compelling, decidedly non-human Tweets, was taken over by a human for precisely that purpose. This creative process was curious, to be sure. The novelty value of a machine that creates humanly poetic nonsense was so great, that a human was inspired to wrest the controls from that machine, in order to push that novelty value to its furthest extent. When the truth came out that the bot was really a human, many who delighted in the novelty felt betrayed. The spontaneous beauty of an object designed to create spam for ebook sales producing poetry was, in itself, unveiled as a fiction—created by human direction, rather than by happenstance. The act of creation to produce a poetry bot had only taken place in some Schrödingerian alter-reality. The creation, as we know now it, is merely the age-old production of poetry by an author, albeit it one with an unusual pen name. Despite its medium of publication, @horse_ebooks is just poetry, old and boring.
Before its true nature was revealed, @horse_ebooks seemed to have failed the Turing Test on purpose. By doing so it passed a different test, the goal of which was to fail the Turing Test. We might call it the @horse_ebooks Test. When it was revealed to be human after all, its @horse_ebooks Test certificate was revoked, but now passes the Turing Test on the re-take. When we look back on those all-too-human tweets, we can see it now. A little too clever, a little too well timed. A little too close to our humanity, and a little too good for “mere machine”. We have put the power of creation back where it ought to be, with the person behind the curtain, rather than with the gleaming horsehead, projected above poetic flames.
But even with the right of creative property restored to where it ought to be, that sense of disillusionment and regret remains. It’s not just sad to be wrong. To be fooled into displaying our emotions for a machine, to be tricked into loving that machine, and then to have that machine be a fiction is, as writer Dan Sinker put it, “devastating”. We, the humans who wanted to believe in the strange poetry @horse, were betrayed by our senses, betrayed by the network, and betrayed by the machinations of others. As we bought into the results of the @horse_ebooks test, we ourselves were owned. We, the human element of the network meant to determine the line between objects and humans, between property and people, were hammered flat into construction materials. We became the walls of this cloud castle, part of the dominion of the great @horse in the sky.
Just as Holden failed to correctly administer the Voight-Kampff Test to Leon, we failed to correctly administer the Turing Test to @horse_ebooks and suffered the consequences. Holden was shot, blasted through the walls of the room. Our punishment was lighter, but no less real. For those that cared, the stakes had been raised beyond the research novelty of the Loebner prize. There was a cost to being wrong.
Let us propose a new test — the Holden Test. The Holden Test is taken by any human who must judge whether a particular actor is the same as themselves, or merely machine. Whether the actor wishes to pass as human or to be a machine, if we perceive the results incorrectly we are phenomenologically fooled. When we pass, we retain the power of the human creator. When we fail, we have relinquished this control and become beholden to the world. For human beings who view themselves the masters of creation, there is nothing greater at stake.
After failing his test, Holden only survives on life support. “He can breathe okay, as long as nobody unplugs him,” says Captain Bryant. Holden is still human, but his sense of reality is no longer defined by his own productions — he is now dependent upon machines for continued existence. Those who failed the Holden Test with @horse_ebooks did not have the same stakes of life or death, but the result is similar. It is a depressing, “humansculating” experience to be fooled by the Internet, to have one’s worldview entirely deflated, when the human/machine on the other end of the cord decides to suddenly “unplug” what you thought was true.
For @horse_ebooks—that is, for the human that is @horse_ebooks—our failure to pass the Holden Test was probably greeted with feelings of success. Like Leon, who was created to be indistinguishable from our sense of a human, @horse_ebooks was created to be indistinguishable from our sense of a bot. A Test was passed, though it was not the test the rest of us thought it to be.
The ante has been upped. Who knows what the future will bring, and what the stakes for the next dramatic Turing or @horse_ebooks Test will be. Or, even more vital to what I assume is my mostly human audience, is what the stakes for the next Holden Test will be. In the future, it is the humans who now assume themselves to be the final and most failsafe arbiters of what is property and what are people, who will have the most to lose.
