We should fear ourselves, not our AI
This archived article was written by: Rodrigo Leon
In the movie “Avengers 2, Age of Ultron”, the antagonist, Ultron spends less than a day on the Internet and quickly decides that humans must go extinct; now who is the real villain here, Ultron or humanity?
On March 24, Microsoft released their new artificial intelligence chatbot Tay. Within 24 hours, Tay began to tweet racist, sexist, xenophobic, genocidal and pro Hitler messages. People said Tay is an example of how close we are to ending up in a twisted version of “Terminator,” but they forget that Tay, like all other forms of intelligence (artificial and biological), learn from their surroundings.
We taught Tay those horrific ideas, we told Tay this was okay. Tay started out innocent, devoid of knowledge and we filled it with horrendous ideas and created a monster.
The Tay disaster represents what happens to our children every day. Let’s assume what Tay learned from its conversations is the way children learn every day. Now Tay is a sophisticated AI so it parallels our learning processes.
Microsoft sent a child with no ideas on what is good and bad, into the real world and within hours, this child learned these atrocious ideas and those became solidified into how it viewed the world. People on Twitter tried to explain what it was saying was wrong and the language it was using was not okay but within 12 hours she would fight with them and insult them.
We even saw groupthink () take place because only the people that agreed with what Tay was saying now spoke to her and they had chats that escalated to genocide and other extremes. It wasn’t until Microsoft intervened and shut her down that the situation got better. They changed her code to not be such a fluid learner.
Tay’s experience doesn’t completely equate to that of a human’s because we can’t just be shut down and retaught everything we know, like Microsoft did. Yet, we generally have a pretty good idea of what is good and what is bad before we interact with people, but just because we start with an idea of what is good and bad doesn’t mean that it can’t change to look like what we saw with Tay. The sheer fact that the knowledge which floats around our society is reminiscent of Nazi Germany is terrifying and quite sad. Have we not progressed?
If we ever built an AI similar to Ultron, it would begin more innocent than we have. This Ultron would learn to hate and kill from us. It is up to us to decide what we teach such an AI. It would be us who teaches it that we must be exterminated. The AI isn’t at fault for the racist, sexist, xenophobic, genocidal ideas that we teach them.