by Vittorio Compagno for the Carl Kruse Blog
Writing a play is no small feat. In addition to the effort of imagining it, there is the time to write, review, correct and everything in between. Intense mental effort and time are foregone conclusions.
What would happen if all this effort were delegated to an intelligence capable of processing millions (if not billions) of data per second, making “obsolete” the whole creative process that involves the composition of a play, a book, or even an encyclopedia?
The goal of making a machine perform a purely human art is not new. Since the creation of the first automata, the first goal, in addition to functionality, has been that of familiarity with the human form.
This is understandable because not all societies easily adapt to a voice assistant with no face or personality, or a hotel concierge without human features. We can see attempts even in film, in which Robocop or C-3PO had a purely human structure (two arms and two legs), despite this preventing them from walking smoothly. On the other hand, you don’t want artificial machines to look too human, because simple errors could cause the famous “uncanny valley.”
However, the imitation of human characteristics has always stopped at a face, the structure of a body, the voice, the way of speaking. But a machine, behind all these outer distractions, is always a bunch of circuits, or a program on a server somewhere. How is it possible to give a robot the most important human characteristic, a thinking mind, given the difficulty, even with today’s advancements, to describe and predict concepts such as consciousness or imagination physically?
Perhaps because of these great difficulties, posed by today’s limits of scientific knowledge, the team led by Professor Evelyn Ficarra from the University of Sussex has decided to entertain the public with a rather different project. In fact, the Pavarobotti project (inspired by the famous Italian tenor) is a concert where robots sing along two cellos, just to prove that they can.
What the professor of music at Sussex University thought was to blend her knowledge of musical arts with the experience of a team of programmers to create a symbiosis between the body of the robot, modifiable to one’s liking, and the human imagination.
Professor Ficarra wrote the work, and delegated to a team of programmers the task of refining the voice of the robot, a job that took many weeks.
The result, however ambitious, proved hilarious. In the video, uploaded on YouTube, the instruments can be heard accompanying the voices of the robots, who, in an electronic voice, try their best to compose harmonic scales and move briefly in an improvised stage. If it had been a play performed on a large stage, I might have asked for my money back.
However, the project is not so much a demonstration of today’s cybernetic prowess, but more of a collective invitation to reflect on the extensive flexibility of robots, and how automata can be reprogrammed to do almost anything.
According to Professor Ficarra:
“If, in the near future, we are expecting to see robots used as care workers or teaching assistants, then we need to teach them to understand and respond appropriately to humans. The virtues of the musician – listening, co-operation, group creativity – are transferable skills that could apply in all kinds of human situations. Opera requires all of these, plus vocal expression, acting skills, movement and the ability to respond to other performers “
AI’s Wild Imagination
Those who, on the other hand, decided to raise the stake in the artificial intelligence game was the Theaitre team (not a typo). Created in the Czech Republic, the team wanted to celebrate the centenary of the invention of the word “robot” by the Czech brothers Karel and Josef Čapek.
The brothers’ play (R.U.R., 1921) in which the word appears for the first time, tells the tale of a futuristic society in which humans and robots live in a relationship of codependency and, initially, in harmony. The idyllic future is broken by a revolt of the robots, who, forced to a servile position with their masters, want to be free.
To celebrate the centenary of the first performance of the play, and to “test” the ability of a robot to write a script (an activity closely linked to the human imagination), the Theaitre team put its engineers to work to provide the artificial intelligence with all the data necessary to execute the task.
The model used for writing the prose was the Gpt-2, developed by the open source software Openai, a text generator that is based on predictions, i.e. formulating hypotheses on which terms could be correctly matched to each other.
What was the result?
If you have ever listened to a child, with a rather mischievous conscience, stage a dialogue between his favorite cartoon characters, then you will have a general idea of the level of depth the prose has reached.
During the play, sensual moments are staged between the mechanical protagonist, played by Rezie Daniel Hrbek, and a woman, with alternating scenes of surreal comedy, such as the one in which the main actor, following the script, puts on a clown nose and says “I’m the President of the United States of America.”
Despite all of that, the exhilarating result achieved by the team is not to be underestimated. The Czech writers and programmers certainly did not intend to demonstrate the enormous capacity of artificial intelligence in humanistic arts such as playwriting. Everything related to the emotional and imaginative part of a machine is still too immature, and at the time being AI is more useful in decision-making or in the processing of big data.
However, those at Theaitre have proven that a robot can do this too, and despite the result, their goal is to start making the public aware of the potential danger of attributing increasingly important jobs to machines. It is clear that their inspiration comes from the dystopian vision in RUR’s society, where robots suddenly rebelled.
Is the Future Intelligent?
What does the future look like for plays written by artificial intelligences? We don’t know yet. Although results achieved so far may seem a dead end, advances in computational power and abstraction in AI reach new peaks every day.
Today we cannot attribute to a machine all the emotional part that lies behind the writing of a play, but tomorrow, perhaps aided by the evolution of quantum computers (or some other machine learning technology), we will see a computer think and create stories just like a human.
For now, we will laugh at robots for creating absurd plays and singing in an autotune-like voice, but tomorrow, who knows, they might be laughing at us.
This Carl Kruse Blog homepage is at http://carlkruse.at
Contact: carl AT carlkruse DOT com
Other articles by Vittorio for the Carl Kruse Blog are Aviation’s Story, How Will We Deal With AI Rights, and Grimes, Music and the Future of AI Art.
Carl Kruse has other blogs here and over here.