By Mike Haley and Tatjana Dzambazova
We’ve been working with NVIDIA for more than 15 years, a long-time collaboration between technology companies. NVIDIA has been busy hosting their GPU Technology Conference (GTC) and we’ve been reflecting on the most exciting aspects of our work together. Here’s a quick look at what’s happening now and some of our future predictions because it’s definitely an exciting time to be in tech.
Let’s face it, AI has the opportunity to be as impactful as the industrial revolution. With machine learning, we can quite literally teach machines to learn how to do almost anything. Virtual Assistants can automate a task and more sophisticated ones can guide design or help optimize making things. Machine learning creates new ways to interact and communicate with tools that make things and this for all industries: manufacturing; architecture, engineering and construction and media & entertainment. Many companies recognize the power that machine learning and AI can bring and are working with the tech. This means there’s a need for powerful workstations, GPUs and software to achieve AI goals.
Data is driving our design decisions and as we capture more data, traditional tools aren’t keeping up. So, new technology is the way to go. In the design world, our workflows are being transformed by abundant compute power. Our changing relationship with technology alters the design of everyday objects, even something as well understood and explored as a chair. Instead of the traditional design process of sketching, modeling and simulating a chair, designers are exploring new software methodologies involving generative design to help them discover design in a fundamentally new way. With generative design, you just need to provide the inputs that describe the problem you’re trying to solve – like goals, constraints and forces – and the tool helps you discover many possibilities. Compute power and smart algorithms help analyze the space and uncover new design options for designers to choose from. It helps you think and design beyond the boundaries of your brain.
To explore this new approach, we’ve been working with Airbus, who continues to reinvent their plane designs. They wanted to redesign a plane partition to be lighter and perform better so those goals were entered into the generative design tool. In return, the tool presented thousands of possible partition designs that fulfilled the goals Airbus defined. Big-data analytics helped Airbus narrow the design iterations and decide on a final, best-performing design for a new 3D printed metal partition. Airbus shared that, this final partition is 45 percent lighter, only used a third of the material and is structurally stronger.
Many industries are starting to see benefits from advances in larger memory and performance. Using GPUs in rendering now allows artists to look at their scene in a drastically faster final rendering view and within the viewport of the application.
Autodesk Maya fluids can be complex to master. In the future, if Maya could learn these settings, then it could help automate what visual effects artists do all the time. It could look at the reference photography and from this, use its intuition to infer the settings needed to recreate the effect digitally.
Another example of how AI can guide the design is traditionally, some of our users are reluctant to apply computational fluid dynamics for all their design iterations because of time and money. With machine learning, we can train a predictor to recognize the correlation between form and flow. We can run this training offline, or in the cloud, for a variety of shapes and input parameters, generating synthetic simulation data. When given enough training data, the predictor acquires a knowledge of aerodynamics, so expensive computational fluid dynamics (CFD) simulation can be skipped. The predictor can now give a customer an idea of their aerodynamic performance. The response is so quick, we could include simulation directly in the modelling and artists can incorporate and further refine their designs.
Powering the AI Revolution
GPU computing is the most productive and pervasive deep learning and AI platform. It begins with the most advanced GPUs and the systems and software we build on top of them. NVIDIA works with the major systems companies and major cloud service providers, like Autodesk, to make their professional line of Quadro GPUs available in laptops, desktops, data centers, and in the cloud. Any screen can be a visual computing experience powered and enhanced by the GPU. NVIDIA creates hardware and software to bring AI to devices like autonomous robots.
Autodesk has the software and expertise to help customers incorporate technology to realize powerful new areas for development. We’re excited to see what we can continue to explore together.