12.00pm - 12.30pm CECCHETTI & CUNNINGHAM FUSION
In this session, ChatGPT provided three dance movements that combine the Merce Cunningham technique and the Cecchetti technique. The proposed movements included Port de Bras Fusion, Turn and Release, and Floorwork Dialogue, all of which incorporate elements from both dance techniques. Following that, the audience requested music suggestions from ChatGPT and were encouraged to use Ludovico Einaudi's Elements album, particularly the song titled Ephemeral Harmony, as it blends classical and modern elements to match a fusion of two dance techniques. As the workshop facilitator, I was also in charge of the music. Unfortunately, I couldn't find the music recommended by ChatGPT on YouTube, Spotify, or Apple Music. So I played another song from the same album, Elements. After the session, I conducted some research on the music but couldn't find it. I believe ChatGPT gave the wrong answer. ChatGPT made it look like the answer was correct, but in reality, it made up the song. That was an error.
8.45pm - 9.15pm MODERN DANCE MOVES
As the facilitator and initiator of the conversation, I asked ChatGPT to provide five modern dance-based movements. ChatGPT offered a combination of general and fluid ideas. The dancer quickly elaborated on these written cues and began performing the moves. At the same time , the audience actively engaged in the chat, offering additional ideas and feedback. However, despite their dance backgrounds, audience members initially were shy and hesitated to interact. For example, one person was initially shy about making a suggestion but soon gained confidence and suggested adding a song that fits those movements. Another, encouraged by the collaborative atmosphere, proposed three more movements to complete the sequence. These inputs influenced the dancers' steps and transformed the basic sequence into a collaborative exploration.
During the 30-minute session, ChatGPT's responses evolved as the conversation and actions progressed. Being the facilitator of this process, I occasionally guided the discussion towards specific topics and encouraged the group to try new things. The dancer was able to turn these written tips into movements of her body, showing her ability to think and act independently.
The final dance, lasting 20 seconds, represents a combination of ChatGPT's initial ideas, suggestions from the audience, and the dancers' capabilities. This collaborative and unique approach showed the impact of AI, as well as the power of collective creativity among all participants.
8.15pm - 8.45pm SET UP
8.45pm - 9.15pm TRIAL 1.1
9.15pm - 9.45pm TRIAL 1.2
9.45pm - 10.15pm FEEDBACK
11.45am - 12.00pm SET UP
12.00pm - 12.30pm TRIAL 2.1
12.30pm - 1.00pm TRIAL 2.2
1.00pm - 1.30pm FEEDBACK
9.15pm - 9.45pm MCGREGOR STYLE CONTEMPORARY MOVEMENTS
In my role as the initiator and guide of the creative process, I once again prompted ChatGPT to provide three movements inspired by renowned British choreographer Wayne McGregor. This decision aimed to create a more pointed and focused tone for the creative session. Alongside the suggested movements, ChatGPT briefly explains of McGregor's dance style. This time, the chat was more active than it was in the previous session. Notably, the second exploration was smoother. Participants who gained familiarity with the process found it easier to execute their roles, contributing to smoother collaborative and dynamic trials. The dancer, driven by a commitment to precision, invested substantial effort aiming at perfecting the quality of her performance.
12.30pm - 1.00pm DANCER'S MOVEMENTS INSPIRED BY BOURNE
This time ChatGPT provided four contemporary-modern dance movements inspired by Matthew Bourne's style. Unlike the previous attempt at the first workshop where ChatGPT gave a short description of Wayne McGregor style, a brief explanation of Matthew Bourne was missing this time. The conversation delved into the potential inclusion of storytelling elements in the third movement, emphasising the freedom to experiment. The emotions that could be expressed in this dance style were then explored, followed by a request for music recommendations for the choreography, which were provided with detailed explanations. The last response suggested using "Rez" by Underworld for sharp and dynamic movements, highlighting its electronic and energetic elements.
The last session was the most interesting research-wise because of several difficulties. In the beginning, one audience member had Wi-Fi problems that resulted in her being unable to access ChatGPT. Noticing a potential bug in the system, I let her use my computer to input prompts and stay within the workshop. However, shortly after, another audience member had trouble connecting to the chat, which brought up questions about problems with ChatGPT itself. In another notable event, one user tried to talk with ChatGPT in Maltese, but the AI model did not recognise this language and answered in Turkish. To address this, I requested ChatGPT translate its response.
Results of the research:
I applied Lygia Clark's idea of participatory art by including audience members in my practice and asking them to contribute to the chat and provide input. Moreover, I used a real-time information system, ChatGPT, to develop spectators' ideas and inspire dancers' movements.
The following results are a combination of qualitative data, feedback received after each workshop, and quantitative data. Moreover, I created a survey to collect quantitative and qualitative responses.
As a result, I obtained four original short dance sequences and understood how to use ChatGPT for dance composition. When interacting with ChatGPT, both the audience and I had to be very specific. I recognised the importance of how the audience and I phrased our requests.
I understood that when someone asked ChatGPT for music that suits movements provided before, it not only suggested a song but also briefly explained why the song suits the material. Contrastingly, when asking about specific choreographers or techniques, it didn't always provide brief explanations of the choreographer and their choreographic movements.
In Workshop 1, with two dancers and three active audiences, a 20-minute setup preceded a one-hour movement investigation that revealed two problems, resolved in an average of 10 minutes. The final sequences lasted 20 seconds each, and participants provided feedback for 30 minutes, totaling 2 hours. Workshop 2 involved two dancers and three active audiences, a 15-minute setup, and a one-hour movement investigation addressing three problems in an average of 5 minutes. The final sequences varied: T 2.1 was 1 minute, and T 2.2 was 30 seconds. Participants shared opinions for 30 minutes, for 1 hour and 45 minutes.
The intention during the first attempt was for both the dancer and the active audience to generate a significant amount of movement material. This led to requests to ChatGPT additional moves. However, I had to encourage participants to utilize their creativity and contribute to the ChatGPT during the initial attempt.
Contrastingly, the second exploration demonstrated higher levels of participant curiosity, resulting in increased engagement. The attention shifted towards a more detailed analysis of a select few movement steps.
Although I did not establish a time limit for choreographic development, both trial resulted in a 20-second final sequence.
Due to the limits of the AI model, it was possible to understand that ChatGPT won't substitute the human mind, but it can assist it. Indeed, ChatGPT can be a useful tool for initiating ideas or generating movement concepts. Dancers can then interpret the chat and bring those suggestions to life.
Sometimes, when you don't know what to choreograph, ChatGPT can help you. It can also remind you to include elements of choreography, such as dynamics, that you might forget. The main problem we faced in the second workshop was the amount of information at once. There was a lot to process. Perhaps, in the future, as suggested by the dancers, I should ask less or stop ChatGPT when it generates too much content. Furthermore, both dancers said that they would have never thought about choreographing with ChatGPT.
In the second workshop, I noticed that I was more confident. Despite facing more problems than in prior session, I was able to solve them quickly. Both dancers tried both roles - dancer and active audience - and enjoyed both. After the conclusion of the workshop and collection of feedback, I began questioning how to interpret the results. Is it more about the experience, analyzing the movements? Or is it more about the process, considering this method as a fast way to choreograph?