Raji Arasu, Autodesk’s Executive Vice President and Chief Technology Officer, outlines the company’s cutting-edge approach to AI. Leveraging a decade of research, Autodesk integrates AI to transform design and make workflows across industries. From AI foundation models to sustainability tools and creative aids, Autodesk is redefining productivity and creativity in design.
Autodesk AI foundation models produce 2D and 3D geometry with the precision and accuracy our customers need to create, construct and manufacture with confidence.
I’m very optimistic about the transformative potential of Artificial Intelligence (AI) for our industries and customers. At Autodesk University 2024, with nearly 12,000 customers, partners and employees attending, we shared exciting Autodesk AI developments for our industries.
Autodesk is uniquely positioned to innovate on behalf of our customers and industries, leveraging over a decade of AI research leadership in CAD geometry. This expertise allows us to develop industry-specific foundation models that reinvent workflows in media and entertainment (M&E), architecture, engineering, construction and operations (AECO) and manufacturing. In addition to delivering productivity and automation today, we are paving the way to extend customers’ creativity in the future.
Just like a pace setter in a marathon, Autodesk aims to be a trusted partner to our customers in the AI race, helping them navigate the next tough mile so they can finish strong.
AI should enhance human ingenuity, not replace it.
Raji Arasu
Autodesk’s AI strategy is built on three core beliefs: reimagining the future and the way we work is inevitable; AI needs to solve problems our customers have today; and AI should enhance human ingenuity, not replace it.
AI – now and for the future
Foundation models are fundamental to our AI strategy.
Starting with manufacturing, these models enable productivity-enhancing AI features that automate complex, repetitive and error-prone tasks in the design process. One of these models powers an AI feature that generates fully constrained sketches, minimising errors in parametric modelling. Another aids in object recognition and classification, enhancing similarity searches for fasteners in product design.
We also have plans to expand foundation models that deliver AI features aimed at improving productivity and automation for AECO and M&E.
In parallel, we are working on advanced and transformative AI in research that will enhance creativity and bring ideas to life rapidly in the future. One such pioneering research initiative is Project Bernini, which generates fully realised and detailed 3D objects. It is trained on publicly licensed data, is not available for commercial use and is open to the broader AI community.
Project Bernini and other research projects tackle multi-modal inputs, including 2D and 3D data, sketches, drawings and point clouds to produce 2D and 3D geometry with the precision and accuracy our customers need to create, construct and manufacture with confidence.
We continue to apply techniques from Project Bernini and other research projects towards industry-specific foundation models that are being released now.
In addition to foundation models, we are delivering productivity-enhancing AI features in both our core products and industry clouds through traditional AI.
AI announcements across our industries
Recent announcements highlight our ongoing commitment to continually deliver value to our customers with AI:
- Sustainability outcomes in AECO: Embodied Carbon Analysis in Autodesk Forma is now in beta, allowing designers to test material design decisions and its impact on the carbon footprint of their buildings from day one.
- Manufacturing productivity: AI-driven features like Drawing Automation and AutoConstrain in Fusion streamline manufacturing processes, reducing time spent on creating 2D drawings from 3D models and automatic generation of dimensional constraints between different aspects of a design. Automotive customers can also explore generated realistic new design forms and shapes in seconds with Form Explorer in Alias. And we just announced our intent to acquire NAVASTO, a company focused on the use of AI to accelerate design and engineering.
- Timely insights across workflows: We are enhancing Autodesk Assistant, a context-aware virtual assistant that delivers timely insights and actions in our products and on Autodesk.com. In Autodesk Construction Cloud, Autodesk Assistant helps construction teams quickly find specifications, reducing the need to sift through extensive documentation. In Fusion, it provides expert answers on tasks and best practices in manufacturing. Autodesk Assistant is built with an open architecture where customers can bring their own plugins to get insights across their project lifecycle. And we’re collaborating with companies like DPR Construction in AECO and ISCAR in manufacturing to expand Autodesk Assistant for specific use cases that benefit entire sectors.
Trust and transparency in AI
Trust is paramount in the adoption of AI. And trust starts with transparency.
We are building Autodesk AI on a shared platform capability that provides governance and traceability in how data is used to train our AI models.
Additionally, we are collaborating with policymakers to shape and advocate for AI policies. As members of the AI Safety Institute Consortium (AISIC), run by the National Institute of Standards and Technology (NIST), we’re part of a powerful coalition dedicated to establishing AI safety standards and guidelines across multiple industries. And, we were amongst the first cohort of companies to sign the EU AI Pact, taking proactive and voluntary steps to comply with EU AI regulations.
Data stewardship
We understand there are concerns about data privacy and choice in the context of AI and intellectual property. We train our AI models using aggregated data that is non-identifiable to protect individual company information. These models power AI features like AutoConstrain and Drawing Automation that deliver significant gains in productivity and efficiency, but do not generate ideas that reveal customers’ intellectual property. Just like the features we release every day; these AI features will be integrated into our products and accessible to all users.
In the future, when we release transformative AI features that generate sophisticated ideas and options for our customers to use which could be considered intellectual property, our customers will have choice to use these features and participate with their data. In addition to this choice, the user experience in our products will provide customers with the ability to select, refine and accept the idea or option that best meets their needs.