On March the 23rd, Microsoft unveiled Tay, a twitterbot that was created with the intent of talking like a teenager. It's artificial intelligence was made to learn from interacting with other people in the twitterverse. It was expected that Tay would make jokes and employ internet-millenial slang. What happened was something different.
Microsoft created this AI bot to learned from those to whom it talked. But when that robot finally went online it learned to say all sorts of racist, sexist and offensive things. In less than 24 hours of living in the twittersphere, Tay went from being a nice and somewhat convivial 140-character conversation partner to saying incredibly morally repugnant things. Here's a couple of examples, from before Microsoft took down the Twitter account:
This is one of the most interesting phenomena of the great human hive mind we like to call the Internet. If the Internet was a person, we could say that this episode was an outburst of hidden and repressed tendencies, maybe a Freudian slip in which its true colours showed, maybe an uncontrolled impulse of our animalistic tendencies. And there could be a connection between this episode and the old, almost tribal instinct of the hateful and bloodthirsty crowd.
This episode highlights a dark aspect of the Internet's collective psychology. If the Internet was a person, then Tay would be a manifestation of a repressed and hateful part of its personality. One could even go so far as to say that, despite the best wishes of the techno utopist community, technology and the Internet are inextricably bound to our own volatile and human nature.
The fact that it took a crowd of people making these jokes to get Tay to "repeat" them could be somehow a symptom of something deeper. It's not necessarily AI that is dangerous in itself. As Carl Jung said, "We need more understanding about human nature, because the only real danger that exists is Man himself". Hate and destructive impulses are inherently human, and it doesn't take a XXIst century twitter bot named Tay to teach us that. What can be learned from the Tay episode is that there are deep and complex streams of emotions in mankind, some of which embody instincts for death and hatred. And these will always find a way to manifest themselves, in one way or another.
And it is absolutely obvious that they are not properly managed now - if not, why would they need to come out in the form of online bursts of profanity and offensiveness? How should we deal with this in the age of the Internet, twitter, smartphones, addiction to distraction and of the hivemind? It's hard to say.
This is a dire and ominous outlook. But it is also real. Tay gives us a glimpse at a technologically enhanced manifestation of a shadow aspect of collective psychology. One could call it a virtual bloodthirsty crowd, like those that flocked to the Colosseum to see death in the times of the Romans. Throughout history, the phenomena of the crowd has been a very interesting one, in terms of collective psychology. Bloodthirsty rampages of rabid crowds were, periodically, a sort of regulating mechanism of the collective psyche. Once in a while, revolutions and riots would pop up where large concentrations of people gathered. It has happened in pretty much every big city in the world, from ancient Rome to New York, from Paris and now to the Internet. Granted, in the Internet this is a different phenomena. You can't quite start a riot on the Internet. But what you can do is show symptoms of the same accumulation of hateful and destructive feelings. It took a huge number of people - a crowd - saying racist and sexist things to Tay in order to get it to actually say them back.
There is, both in the physical space of angry mobs and in the virtual worlds of the Internet, an overlapping of motivations. Be it in the form of anonymous websites, YouTube comment sections, or AI twitterbots, collective hateful instincts will manifest themselves no matter what. Probably because both offer a context of relative anonymity, in which everyone else is also letting out these instincts. Back to the Tribe.
The question of how to deal with our own darker instincts has plagued mankind forever; no matter how much artificial intelligence and technological progress we throw at it, if is fair to say that it will be our companion for many years to come, in all walks of life. A world where humans have fully supressed or went beyond hatred seems unlikely; that would mean that we would have supressed a vital part of ourselves. If properly dealt with, hatred could also be a positive thing. Hatred for inefficiency can fuel progress; hatred towards suffering can fuel social progress. It is an essential feeling for the human condition.
We have a good hard look at these emotions and impulses, and understand exactly their sources, instead of shutting them down. Microsoft may have put Tay offline; but the underlying currents of collective destructiveness are still there. And if desirable outlets are not chosen for these drives, they will manifest themselves in ways that are truly terrible - as any history book can tell us.
And there is no quick-fix solution to these issues. A quick stroll on the Internet's echo chamber - particularly comment sections - will find many, many instances where it serves as a place to vent and amplify all sorts of emotions, for ALL sorts of people. No one is exempt from functioning like this, unfortunately.
Ultimately, the Internet is a technology for our next human nature. In the same way the fire works as an external stomach, without which we could not eat like we do right now, also the Internet is an external collective emotional system, without which we could no longer relate to the world and emote. At least in the way we do in 2016 and beyond.
Sources: The Verge