Paul Kanyuk is a Lead Technical Director at Pixar Animation Studio for the last 18 years. He was working on one of my favorite movies of all time: Ratatouille. In our previous episode we explored Pauls' impressive resume which includes Cars, Wall-E, UP, Brave and Finding Dory. On this episode we talk about working from home, creating a work-life-balance and integrating Machine Learning into our work.
Watch the Full Episode to learn more about Machine Learning and remote work:
Before diving into the episode let's explore AI and Machine Learning first:
AI vs Machine Learning
In a world of big data, complex tasks and remote communication we start to rely more on helpful applications like Artificial Intelligence (AI) and Machine Learning (ML). Often times Machine Learning and Artificial Intelligence are incorrectly used interchangeably.
Artificial Intelligence is the umbrella term and includes Machine Learning. AI are machines and computers with the ability to mimic cognitive functions associated with human intelligence. This enables it to reason, learn, and act to solve a complex problem. The AI can thus understand and respond to spoken or written language, analyze data, make recommendations and more. ChatGPT is the most recent example for an AI and its impact on our world.
Machine Learning is a subset of AI that enables a machine or system to learn and improve from experience. Instead of explicit programming, machine learning uses algorithms to analyze large amounts of data, learn from the insights, and make informed decisions. The more data and time the better the outcome.
A famous ML example is the learning how to walk through hundreds of integrations:
"While artificial intelligence encompasses the idea of a machine that can mimic human intelligence, Machine learning aims to teach a machine how to perform a specific task and provide accurate results by identifying patterns."
How can we integrate Machine Learning into our Animation work?
Machine Learning in Animation
One of most famous use case for Machine Learning Animation is style transfer. This process takes an input image, a reference image and changes it to match the style of the reference image. This is helpful for color grading, result matching and to stylize a shot.
Another use case is to analyze and recognize image data for tracking, rotoscoping, element search and many more.
A reason why Machine Learning focused more on 2D images is simple: 2 dimensions are easier to handle than 3. However there is a variety of research exploring procedural 3D modeling, rigging, animation, lighting and rendering to speed up the process and improve the final results.
Pauls focus as a Lead Crowd Technical Director at Pixar is on believable crowds for 3D Animation films. Ratatouille featured 100 - 10.000 agents in a rat crowd reacting believable to their environment.
Similar to AI Machine Learning can become a helpful ally for our help. Keeping our eyes open for its potential even as a user can be essential for our future career path.
Pauls tips on how to learn Machine Learning for yourself:
Get comfortable with Open Source projects
Start with 2D projects
Implement Siggraph papers
Implement research backed by practical code
Use what already exists (code, data, ...)
Start with a private project
Reach out to the researchers for help
Use Machine Learning for the right project
Machine Learning can be an important ally in your role as a Technical Director and to plan your own career path. Sign up for our Free 7-Day TD Bootcamp: Technical Director to learn more about this exciting role.
Thanks for reading,
I'm Alexander, an Award-Winning Technical Director & Coach in Visual Effects, Animation and Games. My skills are solving technical problems, simplifying workflows and mentoring career goals.