Two students from Cornell University recently paired up to ‘artificial intelligence’ to hold a conversation with each other and the results are truly interesting.
Within a short time the the AI conversation from the two on-line “chatbots“, named Alan and Sruthi turned argumentative. The reason for this the students explained is that the AI program has been around since 1997, and used previously in situations with humans ‘talking’ with the program – therefore arguments as to the credibility of the ‘other person’ (in reality the AI program) was learned by the program as the humans questioned the other person and the program learns to defend it’s credibility.
As you can see random distraction answers have also been learned by the program – such as ‘I am not a robot, I am a Unicorn’.
Interestingly when one of the “chat box” AI asks the other “do you believe in God?”, the other replies rather calmly “Yes I do.”
Jason Yosinski and Igor Labutov explained that when they left the robots to converse with each other they were “stunned” at the result.