Computer algorithms help us choose which movie to watch, which kind of music to listen to, and which kind of literature to read.

But what would it be if the algorithm itself transcends the role of human cultural medium and begins to create culture on its own?

In 1950, British mathematician and computer scientist Allen Turing published a paper "Computer Machines and Intelligence", and first proposed a thought experiment called "imitation game". In one room, there is a human "questioner" and a male or female "interviewer" in another room. The goal of the game is to let the "questioner" determine whether the "interviewer" hidden in another room is a man or a woman. The imitation game must be completed through a series of questions and answers, and the sending and receiving information must be carried out through third parties or typing. "Win" imitation game means that the first step in identifying the game is successful.

Afterwards, Turing modified the imitation game, replacing one of the "interviewers" with a computer, observing whether the computer can complete the conversation smoothly, and making the "questioner" unable to distinguish the difference between the computer and the human "interviewer". This version of the imitation game is called the "Turing Test".

This simple but powerful thought experiment proposed by Turing provides a general artificial intelligence testing framework that can study all aspects of the boundaries between humans and machines, and conversations are just one example.

Comparing poetry with artificial intelligence

On May 18, at Dartmouth College, one of the Ivy League universities in the United States, professors from the Department of Computer and Music will work with students to discuss different areas of artificial intelligence and focus on the issue of machine-creating art. Specifically, in the "Turing Test of Art Creation", we will observe whether participants can distinguish between sonnets, short stories, and music created by humans and machines. Of course, the art created by machines is far less than Shakespeare, O'Henry and Daft Punk.

The dance music competition (“algorhythms”) requires participants to select the music that is most suitable for use on the dance floor from a preset library to build a pleasant dance music collection. In this case, the computer software will randomly select a track from the dance music database as the initial "inspiration seed" to start the creation. The software will select, modify and mix from the library based on this initial track to create a 15-minute dance music. It includes standard notes of 20 features, such as genre, rhythm (bpm), rhythm points, saturation (pitch) and brightness (tooth).

The sonnet competition (“poeti”) and the short story competition (“digilit”) are even more challenging for computers. The competition requires participants to submit independent software packages based on specific “inspiration seeds”, or enter a common noun phrase (such as “dog” or “cheese grater”), and the software then creates the required literary works based on it. In addition, the participating software algorithm requires an unlimited number of different works from a given prompt.

In order to conduct the test, we will first browse the computer's "creation" and eliminate the obvious product of the machine. We mix human creation and machine creation together, and then invite a group of "referees" to tell whether these artistic creations come from humans or machines. In the competition for dance music creation, we handed over to a group of students to dance with the music created by humans and machines. Works that are not different from human works in statistics will become the ultimate "winners".

This competition is open to everyone. So far, participants include scholars and non-academic practitioners. However, as of now, no company has officially announced its participation. This is a bit beyond our expectations. After all, "machine writing" companies in the literature field have sprung up like mushrooms, and the use of text generation software is becoming more and more common, such as profit reports and sports event summary. Of course, in the field of streaming music playback, many companies are also using artificial intelligence to automatically generate lists, the most famous of which is Pandora.

The judgment of the entries is not simple. Even in the initial imitation game, the gender of the "interlocutor" must be gradually revealed over time and revealed from the text. Similarly, in the Turing test, people cannot judge the lack of humanity of computer discourse from a single interaction implementation, but require a long period of repeated testing.

The background issues of imitation games and Turing tests are also worth considering. Is the possibility of winning imitation games independent of time, culture and social class? It is very likely that in current Western society, due to the increasingly flexible definition of gender, the original imitation games will be harder to win. The same is true for the Turing test, in the 21st century, we increasingly communicate with machines. Whether we like it or not, SMS and instant messaging applications have greatly changed our communication methods and expectations of the communication itself. In people's communication, abbreviations, spelling errors and information omissions are almost the norm. The same issues apply to the art field.

What is the boundary between an artist and a creative auxiliary program? Who is the artist?

Thinking about art forms raises another question: Who is the artist? Are the programmers who compiled sonnet creation software poets? Are the programmers who compiled short story creation software writers? Are the programmers who compiled music mixing software DJs?

Where is the boundary between artists and creative auxiliary programs? What impact does this boundary have on the classification of artistic achievements? The fixed art form of sonnets is a high-level algorithm for creative work, although performed by humans. Today, when Microsoft's Oie software helps you "correct" grammatical errors and makes "suggestions" for your words, and you choose to obey it (whether you are willing to accept or be purely lazy), is your creation still your own? Or has it become a "human-machine cooperation product"?

We look forward to seeing "programming artists" submit more programs. Regardless of the performance of these programs and whether they can pass the test or not, their results will continue to open up the boundaries of "creation" and "human-machine co-evolution".
Chapter completed!