SophiaPop! (2020). We present SophiaPop!, in which our robots and AI algorithms collaborate with humans to perform a series of experimental new songs, as both robots and human musicians perform a “live” interactive fiction, scientific presentation, and operatic visual performance and musical. With song lyrics and melodies generated using a variety of neural networks, robotics technologies, and creative artistic tools, SophiaPop! demonstrates creative collaborations between music, arts, and engineering communities. Sophia the Robot, a robotic celebrity and animated robotic character, also serves as a platform for research in human robot interaction, AI development, and robotic applications research and deployment. The authors selected a corpus of human and AI generated Sophia character personality content, which, along with pop music song forms, were used as training data and seeds for a number of creative AI algorithms including GPT-2 and custom trained transformer neural networks, which then generated original pop-song lyrics and melodies. Our team of professional musicians including Frankie Storm, Adam Pickrell, and Tiger Darrow, used the AI-generated content to create human interpretations as the musical works, including singing recordings and instrumentation, which were used as content for further AI generative results. Then we fed the human-performed singing data into a neural network based Sophia voice, which was built from human performances trained into a TTS singing voice by our team members at Cereproc, to generate musical performances of the singing in the style of Sophia. Then we took the musical performances and animated Sophia to sing the songs in music videos, using a variety of animation generators and human-generated animations. We consider the results to be a human-AI collaboration that we hope will inspire people and advance AI as a cultural technology in new and exciting ways, advancing the ethical ideals of collaboration, diversity, creativity, and good AI-human relations, i.e. an AI-Human Artistic Symbiosis. In this way, SophiaPop! represents an artistic, philosophical, and technical quest to serve the arts and AI development, both as a work of fiction with a lovable personality, and as tech platform for AI and robotics research and applications. Merging practices in AI-generated creative arts, with artists, and Hanson Robotics’ previous experiments in social robotics, animatronics, interactive fiction, computational creativity, performance arts, and cognitive AI, the artists and engineers strive to create an AI and robotics popstar, both for tech R&D and to investigate human identity in the time of intelligent machines, to push boundaries in conceptual arts, while also inspiring people to connect with robots in new emotional ways, and consider provocative possibilities of the future of human-robot relations. With the Sophia Pop project, we develop a series of experiments developing and performing new songs for a robot character, using a variety of AI, robotics, and creative artistic tools, with collaborations between music, arts, and engineering communities.
Authors of the algorithms of Sophia the Robot include Frankie Storm, Audrey Brown, Adam Pickrell, Tiger Darrow, Cereproc, Matthew Aylett, Vytas Krisciunas, Mengna Lei, Gregory Kochan, Patric-Paul Tresset, Jeanne Lim, Amanda Hanson, David Hanson