ATUNDA: Of Dance, AI and the need for a decentralised Data Union.
· Tell us briefly about Atunda.
Atunda is developing a software for analyzing dance moves across cultures, using AI-Driven movement recognition and classification methods to solve the problem of appropriation for dancers in the digital space. It is an interdisciplinary collaboration with researchers, scientists and various dance communities contributing significantly to various components of the project. We are deliberately thinking through the very notion of wealth redistribution, and asking very poignant questions on the possibility of wealth pre-distribution.
What if there is an internet of things, based on a truly shared economy, which uses AI to analyze dance moves, and blockchain to keep time stamped records, and get the assets dancers create back under their own control?
What if there is a data union of dancers globally, collectively ensuring that everyone is properly remunerated for their labor, their data and creative output?
We began by identifying a bias in our Copyright Law, dance is copyrightable. The law however, states that the body of work that is protected must be an original creation that has been memorialized in a tangible form, that is, notated or codified in written form. A “dance move” therefore, is not considered a creative body of work, thus not copyrightable. So the question is, if a “dance move” is not considered a creative body of work, then who owns viral Dance Moves? All the snippets of our dance creations that we, or our dancers, or our audience put online, who can lay claims on them, if they happen to go viral and therefore become valuable?
The bias of the law lies in its poor understanding of dance in the age of virality. The law interpreters use an approach that is now heavily contested as archaic, non-inclusive and racist from a certain viewpoint. Because the law that exist today favors mostly a western approach to concert-theatre choreographic tradition, which elevates the status of the choreographer above the dancers, It is so because, the logic that says “dance moves” or those “snippets” of dances cannot be owned, draws their understanding of dance from the basic movements of classical ballet, which are not unlimited in possibilities and must therefore, be kept available to be utilized as the choreographer’s basic materials.
· What innovative solution are you proposing and its potential Impact?
With a team of collaborators at the digital worlds institute of UF, we have developed the first AI-ready dataset for Afrogenic dance analysis. The dataset collected in partnership with The QDance Center Lagos consists of 394 videos that show 13 types of afrogenic dances. The data bank will continue to grow and be available as an open access test bed that can be used by researchers to develop and compare AI methods that analyze human motion from raw videos.
With this AI-Driven movement recognition and classification technique we are developing, we will be able to analyze dance moves to produce intelligence; we are double crossing the law by taking the dance ownership conversation away from Copyright Law to data right, turning dance into data, data into codes, codes into IP and IP as currency, we are emphasizing creativity and sharing as a core value system, and inviting us to consider the next stage of our creative economy beyond production, services, or technological privilege, but to see creativity as currency, no matter from whom or from where.
· Is this solely focused on perfecting copyright or is there another layer here that is charitable in nature?
As we have noticed - with the appropriation of dance creations on games like Fortnite or NBA, and the way social media giants are using manipulative algorithms to sell our data to the highest bidders - new technology can bring great wealth but oftentimes, not shared prosperity. The ethics of intellectual property has been broken by the advertising model of the internet of virality. Our project is deliberately thinking through the very notion of wealth redistribution. What happens if we can properly quantify the value content creators generate and are able to restrict unauthorized appropriation that generates profit for corporations without appropriate compensations?
The charitability in our project is therefore, expressed through our interest to democratize technology, by putting it in the hands of those who need it to arm themselves against abuse. Atunda is deliberately drawing on the Afrogenic epistemologies and cultural practices, to strengthen the technical infrastructure of our community, to fully participate in the digital revolution in our own terms, and by so doing, we are engaging in the possibility of undoing the increasing inequalities already embedded in the development of AI technologies. Where some countries or disciplines lead, appropriate, and dominate, while others have absolutely no stakes, nor a real chance in joining the current AI race.
· How is it envisioned that ownership would be enforced?
When we talk about dance ownership, we are not talking about the capitalist logic of the big C copyright, that undermines the liminal nature of creativity, or stay in the way of its evolution, but we are particularly interested in ethics and the respect for creative traditions, that does not necessarily organize themselves, based on strict notion of authorship and written agreements. The bias of the law lies in its poor understanding of dance in the age of virality. The law interpreters use an approach that is now heavily contested as obsolete, non-inclusive and racist.
ATUNDA is actually a Yoruba word, that means to reproduce, remake, or to recreate, a feminist process by which organisms reproduce themselves. In a general sense, reproduction means making a copy, or a likeness, the ability to replicate and multiply, and thereby providing for the continuation of existence. In Africanist traditions, dance evolves within a complex communal setting that has implicit rules of engagement — that is, through time and mutation, borrowing, remixing and collage, new forms evolve with its own context, but ultimately, considered as growth and a continuation of creativity.
Similarly, when a dance move or dance style emerges within a specific community, through the act of transmission or virality, it reaches the next community, who then adds it to an existing local context or repertoire, and by so doing, evolves a variant of the previous dance move, which can now be considered original in its own right, but all of this is done with an unspoken but strictly respected terms and conditions that makes referencing seamless. Our effort is mostly to automate this practice and to make a statement that we can re-engage our moral codes, and eventually code our morals.
· Would the AI system work retroactively and identify past movements and dances?
Yes, the AI system will operate like a git, imagining dance moves as software, the AI learning process will associate both humans and machine collaboration, using both hashtags and edit buttons, and traceable changes, to accurately tag each move overtime, both in its historiography, characteristics and ownership narrative. In this research, we propose to fully automate the dance annotation process by training different machine learning algorithms, through supervised learning on existing human motion datasets. The trained model will then be tested in annotating existing videos on Instagram, YouTube, TikTok, Facebook, and any other digital archive, and get them validated online through an crowd-sourced process that associates a data union of the global dance community.
· Is there a possibility that this could impact creativity negatively?
Not at all, all we are doing is using available tools to reexamine why dancers are always at the bottom of the value chain of the creative economy. If the technology for recording and protecting rights of music, over a hundred years ago, hasn’t in any greater way, negatively impacted creativity in music, but instead has created a proliferation of musical genres, musical practices, collaborations, remixes, experimentations and above all, have resulted in an entire industry that is strongly thriving economically and in social relevance; creating jobs, opportunities and engagements to young people both in big cities and in small towns, people who will usually have hard times integrating the current job market in formal economies. Exactly the same can be said of the arrival of cameras.
Dance has been primarily deprived of these developmental possibilities, simply because our technological maturity is just arriving at a point in time where deep learning, computer vision and artificial intelligence is bestowing the technical capacity for seeing and learning from what is seen to produce valuable intelligence. So rather than only using such technology to simply create surveillance policing and self-driving cars, our interest is to use the same to begin a revolution in dance and relish over the multiple ripple effects it will engender in the decades that will follow.
Atunda is borrowing from the African creative communal practice mentioned above and working to evolve a unique technology that automates and authenticates this process. For it is paramount that everyone has the knowledge, the content, and the data they generate in the digital universe under their own control, because our current social currency and relational economy favors stealing and appropriation over fair use, and wealth redistribution, and let us face it, it is a thing that has eaten deep into the very fabric of contemporary societies, regardless of good intentions.
· We imagine that the researchers will be making decisions about what qualifies as a dance move. Will there be an effort to engage communities in this decision making?
When we say, “dance moves,” we are not referring to ordinary motor activities or gestures, commonplace movements or dance steps that have no belonging nor proper identifiability. We are making a case for the myriad of signature moves that enter the category of viral dance trends, both in history and in contemporary times. We define a dance move as a unit of groove or symbolic movement that can be repeated in a loop, and forms the base or a variation of an existing dance style that can be further developed through improvisation.
For Atunda as a digital project however, by “dance moves” we mean connected dances that gather the information about their properties, through cameras or sensors, and share this data from a physical body, within a given space, with other bodies over the internet. Any dance move that is captured digitally and linked over public or private neural networks, becomes part of the world of Atunda. And we are saying that such data should be traceable to its original owner, whether they are aware of it or not, whether they are alive or otherwise, so far there is an oral history that can help us make the trace convincingly, and as I have mentioned above, this will be a collaborative effort with a decentralized data union of the global dance community, affording everyone the ability to participate from anywhere in the world in the decision making processes.
· What happens if you do not want your movements to be included?
If the movement can be found in the digital universe, it is already indirectly part of the world of Atunda, because anybody can upload anything and mention the original owner, it is only wise to claim it if you own it, it is like asking what can I do to not make my data be found at all on google, or on Wikipedia? It is an open source database, it is only advisable to get involved and protect one’s own narrative online. The data union however, is based on voluntary participation and engagement, only those registered as members become part of the union.
· What is a Data Union for dance makers?
Data union is an existing web3 framework on the ethereum network for building applications that pay users for crowd-sourcing and crowd-selling solutions, which enables people to earn by sharing valuable data, and liberate our data from centralized monopolies. Some examples include real-time health data, web and app user data, pollution monitoring, smart city data and more. Most applications collect and sell user data secretly on the side, but Atunda will be the opposite, for its sole purpose is to collect human motion data in a fully transparent, fair, and configurable way. We thought the most ethical way to achieve this is to explore the Data Union model for the global dance community, by incentivizing the collection of datasets for the purpose of AI training.
We are building a pipeline that provides a framework for dancers and dance makers to upload their original or reproduced data by choosing from an array of privacy settings like in the creative commons system. This uploaded data on its own, does not hold much value, other than it being an evidence of ownership in times of dispute, but when combined in a data union, it can then aggregate into an attractive marketplace for buyers to extract insights, it could be part of a motion library for gamers, for animation film makers, it could be minted as NFTs or for other commercial purposes including ads etc.
· There most likely are other efforts to help creators capture value from their work using blockchain and other tech. Is this something you are tracking?
Yes, since we began this research in 2019, we have come across a couple of new and exciting projects that are also asking similar questions of protecting dancers from continuous abuse, ranging from publishing software and applications to do similar work. The problem all of them encounter is still the problem of scalability, independence, and cheap tech, that is youth friendly and accessible to dancers everywhere, especially in third world countries, their solutions are still centralized solutions, where there is a central company that reaches out to dancers to come to their studio to record their dance moves with expensive motion capture studios or smart suites, or for a carefully selected few to join their publishing house.
We are primarily interested in bypassing that bottleneck, that will in our understanding create a similar problem as we now noticed with music publishers like Spotify or apple music, where it is the publishing house that is again, making the best profits on everyone’s content. If we rely on the existing technology for data collection, it will take forever to have a reasonable amount of usable data and will be too expensive to achieve. So, we decided to fully automate the dance annotation process by training different machine learning algorithms, through supervised learning on existing human motion datasets. The trained model will then be tested in annotating already existing videos online. We would like to make our developed methods available on-line with live demos, in order to facilitate reusability and repeatability of our experiments.
This solution, aside from being unique, it is also pushing the limit of knowledge in the AI field, we are making an assumption that Dance can teach AI to learn how to learn, and in turn, we task it to protect motion based IP – in much the way Shazam led to first being able to recognize music that was playing in the environment, and then similar software became used for recognizing misuse of copyrighted materials, the concept and protocols we are developing through this research could have parallel impact on the use of dance. The database and underlying software could be used to recognize unauthorized use and publication of dance moves in video games, in NFTs and social media applications and who knows what is coming next.
· What are the cultural and social implications of Atunda?
Atunda is deliberately drawing on the Afrogenic knowledge and cultural practices, to strengthen the technical infrastructure of our community to fully participate in the digital revolution in our own terms, and by so doing, we are reaffirming our existence in the future and having a part in the current AI race. In other words, the metaverse is already a colonised space. I conclude by returning to the body, as a means of paying homage to all those who use their bodies as sites of revolt, as a museum, as archival evidence of our being here and there, whose bodies have been the metaphor through which others develop an image of themselves and affirms what they imagine to be their identity.
If the ways at which our bodies have been claimed, and the ways we’ve reclaimed our bodies are part of the work of excavation and recovering the past, creating an archeology of the present, and then remembering the future, I will have no need to further state the purpose of my research, because... once upon a time, an enslaved man runs toward freedom, and folks violently ran after him, in a mission to repossess his fugitive body, dead or alive, then I asked, “what has he stolen?” “His body” they replied, “that man wants to own his own body.” Atunda simply wants dancers to repossess their bodies, and own what those bodies produce.