Last March, Microsoft launched a chatbot on Twitter, a conversation-robot to be used by adolescents and young adults. The experiment aimed at studying language comprehension, but led quickly to disaster. Just eight hours after the launch of the program, sexist and racist misbehavior forced Microsoft to end it. This robot, named Tay, seemed like a young adolescent, and was deliberately conceived as a bit superficial, capable of gliding past any polemical statements that might have been addressed to it with naïve responses. It had a stock of ready-made phrases to be used in certain contexts. This is how it responded to any mention of terrorism: “Terrorism in any form is deplorable. It devastates me to think about it”. But the algorithms of this artificial intelligence program also allowed it to personalize its responses, depending on what internet users were saying to it. So, a certain number of them successfully tried to test its limits by making it repeat back hateful phrases, then making it produce them on its own. For example, Tay responded to the question, “Do you believe the Shoah took place” with “Not really, sorry.”
How should we understand this digital anecdote? The programmers anticipated the problems by including censorship measures, but these turned out to be clearly insufficient. This type of reaction from teens was, then, expected. Still, to relate this episode to the usual hatred that exists on the internet would be reductive. A more delicate mechanism is involved. It is of course common practice on the internet to try to get one’s interlocutor to produce hateful statements, and this phenomenon even has a name that dates back to the beginning of the web: trolling. What does it involve?
Trolling designates any activity on the internet, or the internet user who is its author, that aims basically at provocation and the creation of useless controversies, with a caricatured and repetitive argumentation style that drives the interlocutor crazy. On the internet it is often said, “don’t feed the trolls!” – that is, don’t respond to them, because they always find a way to use your response in the pursuit of trolling. According to Wikipedia, “if the discussion gets sufficiently infected and the arguments start to fall apart, the troll or one of those feeding the troll will end up by reaching” what is called the Godwin point: that is, someone will make a reference to Nazis. The Godwin point is drawn from a parodic law of the same name, which states that “the longer a discussion lasts, the probability that a comparison to Nazis or Adolf Hitler will be made approaches”. The Nazis are to be understood here as the markers of an absolute hatred, and this would be what the troll is trying to produce.
The reaction of internet users to the introduction of this robot that was supposed to be able to talk was that they trolled it, as internet slang would put it. What does this mean? It would be false to conclude from this that the internet users were really filled with hatred, or that they believed in the content of their violent proposals. Their reaction was standing up to Microsoft’s lie: that this was a robot who talks. It was a denunciation of this. It designates a real-jouissance that overflows speech, but that at the same time sustains it. It was a way of saying: “if you speak, if you are really one of us, ok then we’ll show you what it’s really all about: jouissance”. And this was a jouissance that also unified: it unified those who know that it is jouissance that rules – the union of the trolls that speaking-beings are.
The teenagers who trolled this robot embody a new way of relating to the lying truth. Here, the subject no longer receives its own message in an inverted form: it is the subject who sends back to the Other the emptiness of its own semblants.
(translated by Edward Pluth)