The ability to tell a good joke is considered a special human ability. It seems that artificial intelligence has already successfully tackled the oral challenge to a good level. Open AI’s GPT3 is a leader in the written expression capabilities of an artificial intelligence system and is part of a trend in natural language processing (NLP) systems of pre-trained language representations.
The quality of the text generated by GPT-3 is so high that it can be difficult to determine whether it was written by a human or a machine. An April 2022 review in The New York Times describes GPT-3’s abilities as being able to write original prose with human-like fluency. Other artificial intelligence systems successfully deal with writing poetry and texts of various types. But what about jokes? humor?
Humor is considered one of the holy grails of artificial intelligence because it contains distinctly human abilities: use of a language, understanding of context, timing, and, of course, a sense of humor. AI only has access to information humans choose to give it so if we want AI to make us laugh, we need to be clear about the kind of humor we want to teach it.
Beyond the technological challenge per se, commercial and social motivations push forward the development of robots’ humor capabilities. A chatbot is perceived as more human and likable if it includes some jokes in its output. But most of the existing joke generators were not designed to be integrated into chatbots because they are built on pre-written texts and cannot respond in a good way to changing situations. Robots intended for social care are also required to be able to respond to nuances and incorporate humor as part of creating a pleasant experience.
Finding a way to teach machines to laugh on their own would be a major breakthrough—one that could reshape the way we relate to the devices around us. To understand people’s humor is to know what they like, how they think, and how they perceive the world. An AI that succeeds in creating humor will be able to do much more than just crack jokes.
In October 2019, just before the outbreak of the Covid-19 crisis, there were several performances of a robot named John, developed by a researcher named Naomi Peter from the University of Oregon. According to the reviews, the robot gave a stand-up performance that was not a bad performance at all. John the Robot, is the first robotic comedian of his kind. In this stand-up show, it seemed like a “simple” robot that plays recorded jokes while an operator holds the microphone for it.
In January 2022 at the International Conference on Computational CreativityTuplin presented a research paper describing Witscript, a joke-generating system trained on a dataset of television joke monologues that identifies keywords in the input text and generates a relevant punch line. Unlike other forms of robotic comedy, the system — which Tuplin has patented — can generate contextually relevant jokes instead of just a response to the user’s text. A chatbot or software-enabled voice assistant can humorously respond to user queries (when appropriate) without ruining the interaction.
Another approach is that the better use of AI in comedy and the arts would be a sort of infinite idea generator free of the blinders and biases of human thinking, one that can bounce off an infinite number of potential themes and associations that human writers and performers can. This concept connects to the ideas of Maurice Conti, the futuristic designer, about combining human forces – with artificial intelligence.