We are excited to present you our conversation with Amruta Purandare. Amruta is the Director of CINEMATRIX, which focuses on AI-driven entertainment media creation.
In this conversation, Amruta talks about her vision for the role of AI in the entertainment industry in the context of the current pandemic when productions have been halted, her advice for women who want to lead tech startups, and more.
SDD: Amruta, your work with CINEMATRIX is extremely intriguing and exciting. What was your first encounter with artificial intelligence like and can you tell us a bit about your journey so far?
AP: It was back in 2001, long before the wave of AI / ML had disrupted Indian IT scene. I was a 3rd year undergrad student in Computer Engineering, quite fascinated by the field of Artificial Intelligence and Natural Language Understanding. One of my papers on Self-Learning Systems won 2nd prize under AI and Fuzzy Logic category in some college- level competition organized by Pune Institute of Computer Technology (PICT). It presented the idea of mining web-data using text-parser to extract knowledge from web documents. This paper I believe built the foundation for the next few years, when I went to US for my higher studies and got full scholarship to pursue research in this field.
The initial 6-8 years I primarily worked in Natural Language Understanding. I started with Text Mining, then eventually moved into processing Speech and Dialog. In 2006, I presented a paper on Humor Analysis in FRIENDS TV-show at the EMNLP conference, which got me very excited about applying Machine Learning and Natural Language Understanding algorithms to process dialog or audio-visual content in Television Programs and Movies.
My recent work is primarily into Multimedia domain, where I am exploring and applying techniques from Computer Vision, Image Processing, Deep Learning areas, to process rich- media content that involves Audio, Video, Image, Dialog and Music datasets.
SDD: During exceptional times like these when production is halted we saw ESPN air deep- fake ads and surging numbers in online streaming. What is your vision for what AI can do for the entertainment industry in the future?
AP: The future of entertainment industry is no doubt powered and fueled by Machine Intelligence. For end-users or consumers, it is transforming how we listen to music, watch movies, play games or stream online content. Today, we can browse programs or launch apps in our Smart TV using simple voice commands – that’s already AI.
For production houses and media professionals, it is providing new tools and technologies to automate workflows and easily create original content. Whether you are editing videos, adding visual effects (VFx), composing music, or making animation films, AI is already becoming ubiquitous in media industry. You can see deep-learning based algorithms like Video Segmentation, Artistic Style Transfer, Generative Neural Networks, Deep Fake being used to produce Hollywood style special effects, original soundtracks or animations..
Capturing gestures and facial expressions from real videos and applying it on animated characters is I believe a trend we are likely to see in near future. I am also seeing lot of demos these days on creating stunning Visual Effects, purely through Computer Graphics and Machine Learning.
Use-cases exploring applications of AI / Machine Learning in Entertainment are equally and mutually beneficial – not only for Media Professionals and Content Producers, but even for AI Researchers and ML Engineers.. Unlike traditional datasets that come in Excel or CSV formats, here we are talking about rich multimedia content – that has language, dialog, audio, visuals, action, comedy, drama, dance, music, sentiments.. It’s pretty mind blowing when you look at the possibilities it offers for AI / ML research..
While traditionally these two streams (namely Engineering / IT and Cinematic Arts) have run in parallel, we expect to see more synergy in coming future. Today, I already see many engineers prefer to describe themselves as ‘Creative Technologists’ to emphasize that their creative side is just as strong and valuable as their core engineering and technical skills.
SDD: From an entrepreneurial and growth (particularly funding, talent acquisition etc.) perspective, what advice do you have for women who are interested in setting up tech startups?
AP: In early stage, most start-ups are typically boot-strapped or self-funded, so it’s very important to be extra frugal with our resources and budget. As an entrepreneur, we need to evaluate very carefully – what we should outsource vs build in-house, or even what we can delegate vs do ourselves. These key decisions can have a big impact, not just on your budget but even from long-term product / code maintainability standpoint. I see many entrepreneurs prefer to outsource or delegate initial product development to focus only on the sales / management aspect of business. Personally, I am not very comfortable to have this dependency on external vendors or contractors, every time I need to add a new feature or functionality into my product, service, app or website. Understanding the code / platform inside-out is essential for me, in order to scale it afterwards or pitch it effectively in front of investors.
On talent acquisition, I have created student-outreach programs to engage college students looking for internships or final year projects. I like working and collaborating with Creative and Artistic individuals, who are not only strong in technical skills, but also highly creative in how they think or apply those skills in practice.. Fancy jargon somehow never impressed me – almost everyone has access to Git libraries, online training courses, Deep Learning / Machine Learning platforms & frameworks these days. What I like to see is what someone has built or created using those standard set of packages and libraries.
In early stage start-ups, you normally want to hire team members who are multi-talented,e.g. a mobile app or web developer who also has strong front-end design or computer graphics skills, or perhaps built some demos in AR / VR. Writing neat, clean, well documented code is also equally essential, so that you can enhance or maintain it afterwards, even after the team member leaves.
SDD: Do you have any advice for a beginner who wants to get into the field of deep learning (particularly computer vision, image processing etc.) and stay engaged with it?
AP: The best way to learn for me is always by trying it out myself. Whether it’s a new programming language, tool / library, package, framework, or algorithm, I always give myself a fun mini-project or assignment which allows me to try or test something hands-on. Learning is a continuous on-going process, so it’s important to stay up to-date with latest technologies we see trending on social media platforms.
In my experience, it’s usually the fun side-projects we do over weekends or in leisure time that often bring us not only the sense of fulfillment and joy, but also the next big break or opportunity. Since it is something we choose to do voluntarily out of self-motivation, I think the excitement and passion is evident, when we speak or describe it to others.
Bringing your passion or hobbies into projects will make sure learning is far more fun and exciting, than just watching online course videos filled with math formulas. For example, if you enjoy sports, you can experiment with sports-related content while building your computer vision or machine learning models. If you like travel, you can experiment with travel blogs or travel photography, to test your text mining or image classification models. This strategy of giving freedom to choose your own favourite genre or domain has worked out exceptionally well, with some of my interns who were also new to deep learning and computer vision areas, and had never used libraries like TensorFlow, PyTorch, Keras before.
SDD: Are there some of the new research papers or projects in AI that caught your interest recently, if there are can you share with us?
AP: I was quite impressed to see some Deep Fake demos recently, which also inspired me to try a few of my own.. The technology behind Deep Fake to animate still photos or images w/o using any traditional character animation tools is quite fascinating and revolutionary for Animation Industry. A recent demo on Pose Animator which captures hand gestures / body movements from webcam and applies it on animated characters is also quite impressive, and something I see many customers are already interested in.
I am also intrigued by OpenAI’s JukeBox, and Google’s Magenta.JS, which I plan to use in some of my own work on Computational Music Generation. I see this as an opportunity for me to learn techniques like Generative / Sequential Models or LSTM hands-on.
I also see lot of projects and demos trending on Augmented or Mixed Reality these days, built using web / front-end technologies like WebXR, WebGL, Three.JS, Unity, AR Core etc.
SDD: What inspires you everyday?
AP: The best part about Entrepreneurship is you have complete freedom to choose your own path, write your own job description, decide what you want to achieve in the next few years, as well as, define immediate short-term plan, and day-to-day activities to reach those long-term goals.. It’s important to keep in mind long-term vision, as well as, set some short-term milestones to work towards, in order to stay productive and inspired.
Learning is also an essential part of personal growth and development. I always make sure my daily To-Do list includes new skills and technologies I want to develop or explore. These can be programming languages, frameworks, platforms, libraries that are most commonly used in some of the projects I see trending on social media that catch my attention. Keeping 10-15 exploratory project ideas ready on Sticky / Post-it Notes ensures that my daily work never gets boring or monotonous.
Apart from technology development, I find social media engagement as essential part of business development. A post or demo we share on social platforms can bring us important business contacts or leads, and opportunities to engage with like-minded folks or potential collaborators, who are working on related technologies and have similar vision as ours.
Join our She Drives Data community on SHEROES (You can also download the app on iOS and Android) to connect with thousands of women data enthusiasts across the world, hear from data science experts and get updated on the latest tech news every day ❤️️