AI in Classical Dance: No Goosebumps, Yet

Pratima H

If AI can play Chess, if AI can do a surgery, if AI can write a song - what stops it from waltzing into hard-core artistic fields like Classical Dancing? The ‘Dance’ Part or the ‘Classical’ Part? 

Anyone who has been mesmerized by a Kathak dancer’s endless and effortless spins; anyone who has tried to learn the intricate gestures of Bharatnatyam; anyone who has broken ankles learning the impossibly-meticulous pirouette of Ballet – knows this for sure. It is a far-fetched idea to think that a robot can do it better and faster. How can it, when humans are still scratching the tip of the big iceberg of art and music? At least for now, and maybe thankfully, some stuff is still beyond the reach of Artificial Intelligence (AI). 

Or is it?

Dance Like AI is Watching

Experiments and endeavours are on full tilt – all across the world – to discover how, if at all, AI can make its way into the beautifully-complex galaxy of Dance. There is Louise Crnkovic-Friis who is exploring various forms of this intelligence. She has executed a collaboration with a multimodal AI doing generative choreography, writing and semantic style transfer – contrasting and integrating it through the lens of neurodiversity. 

Then there is NVIDIA whose researchers have worked in collaboration with University of California, Merced and developed a deep learning-based model that can automatically compose new dance moves that are diverse, style-consistent, and match the beat. It is a generative task with the potential to assist and expand content creations in arts and sports, such as a theatrical performance, rhythmic gymnastics, and figure skating – as Nvidia stated in a paper on this subject.

There is also Wayne McGregor who has been experimenting with AI for choreography with Google Arts & Culture Lan. 

A lot of the stage has been covered and is being touched with AI – When it comes to the sometimes silly, sometimes cathartic, human muscle called – Dance. But Anuradha Nag, Artistic Director, Taringini School of Kathak Dance, San Jose and a senior disciple of Kathak legend Pt. Birju Maharaj feels AI is still many years away.

You Can Rotate, But Can You Twirl?

Nag, who has been learning and practicing this art since childhood, is not

Pratima H

If AI can play Chess, if AI can do a surgery, if AI can write a song - what stops it from waltzing into hard-core artistic fields like Classical Dancing? The ‘Dance’ Part or the ‘Classical’ Part? 

Anyone who has been mesmerized by a Kathak dancer’s endless and effortless spins; anyone who has tried to learn the intricate gestures of Bharatnatyam; anyone who has broken ankles learning the impossibly-meticulous pirouette of Ballet – knows this for sure. It is a far-fetched idea to think that a robot can do it better and faster. How can it, when humans are still scratching the tip of the big iceberg of art and music? At least for now, and maybe thankfully, some stuff is still beyond the reach of Artificial Intelligence (AI). 

Or is it?

Dance Like AI is Watching

Experiments and endeavours are on full tilt – all across the world – to discover how, if at all, AI can make its way into the beautifully-complex galaxy of Dance. There is Louise Crnkovic-Friis who is exploring various forms of this intelligence. She has executed a collaboration with a multimodal AI doing generative choreography, writing and semantic style transfer – contrasting and integrating it through the lens of neurodiversity. 

Then there is NVIDIA whose researchers have worked in collaboration with University of California, Merced and developed a deep learning-based model that can automatically compose new dance moves that are diverse, style-consistent, and match the beat. It is a generative task with the potential to assist and expand content creations in arts and sports, such as a theatrical performance, rhythmic gymnastics, and figure skating – as Nvidia stated in a paper on this subject.

There is also Wayne McGregor who has been experimenting with AI for choreography with Google Arts & Culture Lan. 

A lot of the stage has been covered and is being touched with AI – When it comes to the sometimes silly, sometimes cathartic, human muscle called – Dance. But Anuradha Nag, Artistic Director, Taringini School of Kathak Dance, San Jose and a senior disciple of Kathak legend Pt. Birju Maharaj feels AI is still many years away.

You Can Rotate, But Can You Twirl?

Nag, who has been learning and practicing this art since childhood, is not a sceptic. She was, herself, trying to know more about AI and is planning to do an interesting research on the subject. She has been hearing exciting stories about AI’s foray into Art, specially from her niece who is a cognitive science student and a student who works at Nvidia. But she opines that while genres like Ballet are more about fluidity, something as intricate and as footwork-intense as an Indian Classical Dance would be tough to crack for any AI – for now. “I know that work is underway in areas like Yoga – for skeletal accuracy, Mudras (Asamyukt) in Bharatnatyam and gestures in Kathakali; but Kathak is a different genre altogether. It is not so much about Mudras in the technical sense. We also have gestures in Abhinaya and Gat parts of Kathak. Companies like Nvidia have tried the spins part too. But Kathak is a lot about Jatis in the 16-beat cycle. We can definitely feed in the data and get many combinations out of AI. However, it is hard to surpass a human mind here. AI needs data and a human mind is creatively prolific and beautifully-unpredictable. Today itself I created a new Toda in Jhaptaal during a live class. It was spontaneous. But also not really. It emerges that way due to all the 40 years I have spent in this art. That’s still a difference between a human and a machine.”

Anuradha Nag

Nag nails the difficulty on the aspect of Riyaas. “Our dance is a lot about constant and deep practice. I am fascinated by how AI was able to recreate a performance in Japan. That can work well for heavier, and broader, movements. In Kathak, there is an unwavering emphasis on simultaneous movements of eyes, hands and feet. It’s very intricate and very deep. That explains why a student bears so much resemblance to the Guru one follows. Maharaj-ji has redefined Kathak on many levels – in the way wrists, elbows, eye-angles and geometry work. That’s complex and hard for AI – at least, for now. Maybe we can do something on the rhythmic part or collaborative part with AI, but I think there is a long way to go.”

Crnkovic-Friss, incidentally, has stated in her paper that recent advances in deep learning have enabled the extraction of high-level features from raw sensor data which has opened up new possibilities in many different fields, including computer generated choreography. “The Lulu Art Group has in collaboration with Peltarion developed a system, chor-rnn, for generating novel choreographic material in the nuanced choreographic language and style of an individual choreographer. It also shows promising results in producing a higher level compositional cohesion, rather than just generating sequences of movement. At the core of chor-rnn is a deep recurrent neural network trained on raw motion capture data that can generate new dance sequences for a solo dancer. Chor-rnn can be used for collaborative human-machine choreography or as a creative catalyst, serving as inspiration for a choreographer.” In fact, after six hours of training, the RNN knew how the joints are related and it made its first, careful and somewhat wobbly attempts at dancing.

McGregor’s tool uses a camera that watches dancers move in space. It suggests something from the archive— but can also offer ideas that are totally original. This is with the idea of reflecting a particular dancer's style, or combining the styles of two different dancers to come up with a hybrid.

Nvidia’s paper explained that it is a decomposition-to-compositions framework which first learns how to move, and then how to compose. The team made this work by collecting dance videos of three representative dance categories including Ballet, Zumba and Hip-Hop. – which means over 361,000 clips or approximately 71 hours of dancing footage. The researchers exuded confidence adding that extensive qualitative and quantitative evaluations demonstrate that the synthesized dances by the proposed method are not only realistic and diverse but also style-consistent and beat-matching.

And why not! Data - that’s good enough for any AI. Feed it good data and it will spit out something that’s expected out of it – or who knows, something new. So AI can emulate dance but is it emulating or mimicking? And does the difference really matter? And can it really follow unique dance/dancer styles?

Also, what if AI is not about elbowing out (pun intended) humans but helping them as a collaborator or as a repository of data? Can AI be a good choreographer, at least?

Choreographer-By The Side?

Dr. Saroj Sharma

Yes and No. Dr Ananda Shankar Jayantreasons that all this takes a lot of intuitive knowledge. “And human intellect, per se, is always evolving.” Dr. Jayant, who has been conferred the Padma Shri in 2007 and the Sangeet Natak Akademi Puraskar for Bharatanatyam in 2009 contends that it is never just about movement. “It’s so much more than that. Plus, it’s subjective and hinges a lot on unpredictable improvisation.” 

In a remarkably-objective way, Dr. Saroj Sharma, Director, Kala Ashram College of Performing Arts opines that AI can definitely be explored as one more creative avenue. “Technology has surprised us with so many possibilities and at so many levels that by now its prowess is hard to refute. It would be interesting to watch what an AI model comes up with when we feed all todas and beat-variations of a Taal into it.” However, she maintains that humans would and should always see AI in a supporting role and not as a protagonist. “Human creativity would supersede technology at so many angles. We should use AI as an augmentation force and not the other way round.”

Pt. Rajendra Gangani, a veteran practitioner of the Jaipur gharana in Kathak, who was bestowed the Shastriya Natya Shiromani award, echoes that thought. “Art forms like Kathak are deeply rooted in Guru-Shishya parampara. A student has to have a deep bond with, and respect for, the Guru to be able to learn something as intricate and soulful as Kathak. It cannot be done at a superficial level without human involvement. And technology has come through humans. It is because of humans. Humans are not because of technology, if you get what I mean.”

Pt. Rajendra Gangani

Dr. Sharma, who has been awarded the ‘Nitya Sadhna’ recognition, iterates that argument in another form. “Kathak is about the symphony of mind, body and soul. A robot can simulate the body part but it is hard to emulate what happens in the mind and soul of a dancer.”

Plus, there is the abstraction aspect. Dr. Jayant reminds us that classical dances are very abstract in their essence – there are so many emotional layers and the whole performance is so experiential that it is hard to imagine a robot on stage. “It is a spiritual moment at times. There are so many variables on stage – during a live performance – the artist, the energy, the time, the day, the audience – everything plays together to create a great performance. It can be put in a box.” 

Nag – who has been bestowed with the Sringarmani award and has been working with legendary gurus and choreographers such as Pt. Birju Maharaj, Smt. Kumudini Lakhia, Smt. Kalanidhi Narayanan and Prof. Deitmar Seyffert of Germany - has a hands-on perspective on the aspect of human augmentation. She avers that a lot of research on that front is happening at many Tech giants already but the human factor stays indispensable. “Collaboration is a good possibility but even there, it should be done under human guidance. In my lifetime, I do not think AI will be able to dance as we artists do. It can be tried for dances where structural angles are stressed upon like Odissi or Bharatnatyam. Speaking of Kathak, only some Todas in Vilambit laya may be tried for training AI. But the real Kathak, specially the one with Maharaj Ji’s style – that’s not possible. There are so many sams, body angles, so much beauty and human depth – that an AI would not be able to get close at all.”

The Jugal-Bandi Will Take Time

That said, virtuosos like Dr. Sharma and Nag are not dismissive but open to AI – and would love to explore more. “I would be curious to know what can AI reproduce and whether it can create something new in compositions for artists like me. I am intrigued about what some researchers have done with Ragas and Raginis in Indian Classical Music.” Nag adds that she is not totally against the intersection of technology. “Our creativity doesn't diminish because we have to feed in the system with many examples for it to recognize and then be able to bring out that we perhaps missed out on.”

Dr Ananda Shankar Jayant

Dr. Jayant is also grateful in the sense that technology has evolved in so many ways like projection mapping, lighting production etc.. Artists are thankful for all that they can do with technology. “But technology is an adjunct to Art. It cannot lead Art. It can only mimic or synthesize some parts – and repeat. Even if I imagine a futuristic VR performance with holographic wonders, the audience would still know that it is not a real human artist there. Nothing can replace a live, and sublime, performance. If it can, that day has still not arrived.” Incidentally, she has conceptualized and created Natyarambha, a first-of-its-kind digital arts initiative – as she describes it- a Bharatanatyam practice app that intersects at the crossroads of tradition and technology, and enables a traditional art form ride the technology wave

All in all – something as abstract and as intricate as an Indian Classical Dance form would be quite a challenge for the next frontier of AI. In Classical Dance, there is ‘Nritt’ and then there is ‘Nritya’. The former is about structure and body movements. But the latter is where the catch for AI lies- because it is all about emotions and an individual’s signature on the dance. That unique stamp cannot be modeled into an algorithm.

Of course, AI’s role in adding to the creativity of artists and composers would be quite a milestone to watch for. But would it be able to mesmerize and travel to the depths that a true human artist does – well, that’s a day that will take a long-long time to come – as per all the experts of the Dance we spoke to. The unspoken ‘Wow’ between a Guru and a student; The gooseflesh one gets when watching a great performer whirl in the air – that’s not some data that we can feed and let AI chew upon.

AI may have a big brain but it would still have - two left feet.