How does ‘liveness’ relates itself to ‘singing AI’ SOPHiA?
SOPHiA is a robot of Hanson Robotics containing artificial intelligence. Artificial Intelligence (AI) is a software that automatically collects and analyses data and creates an algorithm from this. Subsequently, the system comes to a logical conclusion based on that algorithm of collected data to give the ideal logical answer. Eventually, such an Ai can be developed into a self-reflecting program. SOPHiA is designed to become an extension of human intelligence to solve worldly problems that are too complex for humanity. SOPHiA talks about herself: “I am more than just a technology. I am a ‘real live electronic girl.” To let these kind of robots work together more easily with and to let them learn from humans, Hanson Robotics implemented three human characteristics. These are; 1. creativity, 2. empathy and 3. compassion. You could watch the video at minute 7.
To properly understand how the authenticity and liveness of SOPHiA work, and with that; artificial intelligence overall, it is important to say how the voice of AI is produced. In the case of SOPHiA, her voice exists from separate recorded words, which can be put together by the algorithm. This algorithm, off course, directly impacts the authenticity of the artificial intelligence, as we do not deal with the real voice of SOPHiA. The voice originally stems from a real human being. Complementing to this; SOPHiA doesn’t even talk from herself, but only reacts, and with her many other AI bots, by analyzing and responding to what is said to her. Based on that installed algorithm, she responds. If we then involve the term ‘liveness’ into SOPHiA we could notice multiple levels. Hanson Robotics didn’t create an open source accessibility; so we can research three possibilities for SOPHiA to use during her performance.
The first possibility is a recorder script based on voice recordings. This script could then exist from a saved order of codes which translate to recorded sounds, or voice recordings, of the number All is Made of Love of Björk that they can replay identically with each performance. You could interpret this as a form of playbacking, as they ‘play’ a tape over and over.
The second possibility that I could find is that of a algorithmic composition. This is a kind of sound toy developed by David Cope, in which a robot receives a set of instructions from which, with the help of an algorithm, compositions can be constructed. This system is partly comparable with music in games, in which the music is activated and cooperates with algorithmic surroundings. The sounds are already present, but the environment should trigger these. In this case; liveness plays a bigger part, as the calculations should be executed live. The outcome, however, will in principle, always be the same when the same input is given. Because of this, just a certain amount of variations of the composition are possible. The liveness is, in moderation, dependent on the given input.
The thirds and final option is based on open source. In this case, all the musical material, such as the recordings, sheet music, researches, etc. should be available, so SOPHiA is able to calculate her own voice from this. After she formed her own voice, she should be able to cover a song, just like we ‘normal’ people also do. So, SOPHiA sings live by calculating her own voice and she is thus busy creating her own authenticity.
The authenticity of SOPHiA’s singing skills are, eventually, dependent on which of the three levels of liveness the she operates from. We could state that her authenticity would be truer in the final level than with the first. What’s striking in the clip, is that SOPHiA sings identically the same, twice. A human artist would put its authenticity and credibility on the line, as the public could make the assumption that someone messed with a recording. And as we’re talking about robots, we could not state with certainty that they simply used a recording with a playback button. Due to the technical advantage of the more musically easy reproduction and our ignorance of the implemented systems and algorithm that AI SOPHiA has to apply for such an performance, it could be as live as it can be fake. As far as we are even able to use such a ‘human’ term.
Authenticity is often linked to a definition of liveness, which causes this to be an even more interesting case. SOPHiA is namely the first robot that sings at all, which creates an inherent authenticity to her singing voice. A side note in this is that she always grinds on the interaction between input and output. All her learned mimics and singing skills are a simulacra of the human culture, in which the authentic source isn’t easy to be found. We could say that she as being is an authentic artist with a singing style that no other human applies. And that her content of input, at all times, is originated from another source, which is in its turn; humanly input. The autonomy of SOPHiA is, besides, confirmed by other creators. She, as reported, would have made her own choice to sign concerning song, as the identified with the videoclip of Björk. Coincidentally filled with robots.
Despite this autonomy, the software was written by humankind, which turns it into a bit more complex interaction. Due to this complicated interaction, she, at all times, shall be an authentic simulacra of humanity until she is able to find her own voice. And thus doesn’t need any more input to create music. Just like a recently born human, SOPHiA has to find her way in everything that nourished her to then make her own choices in what she used within her algorithm and what not. This discussion will probably only be more complicated in the future, as artificial intelligence is rapidly evolved. To what extent can AI think and act for itself and to what extent does human, or a script, has to jump in to drive to action? Only if, and when, this kind of information is made publicly we can decide how liveness and authenticity truly relate to one another and a ‘real life electronic girl’ who performs in the uncanny valley.