A new experiment showed that just like with humans a conversation between chatbots can rapidly turn sour, especially when God is mentioned.

Researchers from Cornell University rigged up a chatbot system which allowed chatbots to talk to each other, witnessing a conversation that rapidly went from childish taunts to pseudo-metaphysical problems.

The conversation took place after two graduate PhD students at Cornell University gave voices and 2D avatars to a pair of online "chatbots", which they named Alan and Sruthi.

In previous experiments, humans who had conversed with chatbots complained about the chatbox's lack of attention, focus and intellectual capabilities, but as the new experiment shows, chatbots also rapidly become frustrated and annoyed with each other, making the interaction all the more funny.

The chatbot-vs-chatbot avatars are a British man, Alan and a South Asian woman, Sruthi, and are both instances of Cleverbot which were developed by artificial-intelligence programmer Rollo Carpenter, who used a software which learnt phrases from millions of conversations it has had with humans on the Internet.

During the conversation, the British man avatar declares itself to be a unicorn and later on tells the women avatar she is unhelpful, which he says makes her a "meanie."

Sruthi however proves more philosophical, focusing her attention on metaphysical questions such as the existence of God. Asked if he believes in God, Alan answered "it's not everything," to which its counterpart said "not everything could also be something, for example not everything could be half of something, which is still something and therefore nothing."

Proving he still had some wit the male avatar also told his female counter part, "You were mistaken. Which is odd, since memory shouldn't be a problem for you."

Cleverbot won the 2010 British Computer Society Machine Intelligence Competition.