Interview with Thor from Rave

What is Rave all about?

The project is all about enabling people to compose novel musical art with AI. Over the past few years, teams of students from the University of Guelph and the University of Waterloo have worked tirelessly at Rave to produce the world’s first AI DJ system. A system that can seamlessly mix a set of songs to produce the optimal mix to listen to as well as unique combinations or mashups of two songs into one. This system is currently producing around 10-20 TB of mashups and mixes for people all around the world through our website

What is your role at Rave?

I am the AI lead at Rave. I wrote the foundation of the system we use for training deep probabilistic models at Prof. Graham’s Taylor’s lab at Guelph and The Vector Institute during my masters. I also love teaching people about mathematics, physics and how to write deep learning algorithms and the environment at Rave is a great place for doing that. I also teach a scientific computing course at the University of Guelph which was designed by Prof. Graham Taylor and is focused on highlighting ways in which we can use computers to solve real-world industry situations.

Who are Rave’s academic collaborators?

The collaboration is between Rave, Prof. Graham Taylor and Prof. Stefan Kramer at the University of Guelph. The benefit of this research in Ontario is that it supports the growth of a local tech company that offers the opportunity to do the very enjoyable work of applying the newest trends in AI and machine learning to generate new and compelling art. The many students who have worked at Rave over the past few years act as a good testimony for the quality of the learning environment that is provided. At Rave, we have an open discussion about many different problems and we actively support each other’s student activities which covers an extremely broad range of problems that can be tackled with AI.

What are your goals for your SOSCIP project?

Our goal is to be able to drastically reduce the computational cost of our current system and provide a platform that can enable everyone to create compelling pieces of art. At the same time, we want to discover new and interesting things about Deep Learning!

How are SOSCIP’s platforms helping Rave achieve its goals?

SOSCIP’s platform will enable us to investigate ways of using bleeding-edge Deep Learning techniques to navigate around existing inefficiencies in our existing system, as well as to take a leap beyond its current level of performance. SOSCIP’s platform is exceptional for helping us achieve these goals as it can support the training of deep neural networks on very large datasets. This is a requirement for us because we want to analyze and generate high-resolution videos at record speed. The platform that SOSCIP provides is perfect for running the experiments that are required to produce models that can perform these tasks efficiently.

What is the benefit of this research in Ontario?

We hope to find hard problems to tackle using our toolkit and we are now looking into anomaly detection for signals and time-series data. There have been some interesting papers claiming that deep generative models outperform existing standards in this area. We are interested in applying this technology to predictive analysis of health records or on critical assets in the power-grid. In the future, I think it will be through efforts like this that our growing team and technology will provide the biggest benefits to Ontario.

How are data science and advanced computing changing/affecting this industry?

There is a revolution happening in how we can use advanced computing to process and generate media content. Our objective is to use all this computing power to revolutionize the way people create and enjoy music and videos. I think that this kind of a platform can affect the music industry in similar ways to how the invention of electricity did in the 20th century – with an explosion of new ideas to make interesting sounds and vibes due to the advent of new instruments.

How has SOSCIP enabled you to facilitate and accelerate your research?

SOSCIP has provided us with the opportunity to make all our research ideas come true. Being able to rapidly run a large batch of experiments is crucial for our daily progress and the support team has been incredible in providing us with a world-class working environment. As an example of how powerful the platform is: we ran 5000 experiments in 24 hours to produce one of the latest updates to our models. Progress in our work is all about being able to quickly iterate through experiments as we try to determine which experiments bring an overall improvement to our system.