Recently on the Unstructured Unlocked podcast, host Tom Wilde sat down with Andrew Carr, founder and chief scientist of Cartwheel, to explore how generative motion can provide new perspectives on unstructured data. Carr’s insights into using AI to transform animation and motion, particularly in gaming and entertainment, offer a novel look at the intersection of creativity and technology.
Listen to the full podcast here: Unstructured Unlocked season 2 episode 10 with Andrew Carr, Co-founder, and Chief Scientist, at Cartwheel
From machine learning to generative motion
Carr’s journey began in machine learning, eventually leading him to work at some of the most prestigious AI companies, including OpenAI. His early experiences with machine learning focused on complex problems like predictive physics at Lyft and program synthesis at Google Brain. However, it was his time at OpenAI that introduced him to generative AI, laying the foundation for what would become Cartwheel.
He explained the core concept of Cartwheel’s offering: “Fundamentally, you type text as input, and you get motion animation in 3D as output that you can edit.” This technology, which empowers artists and animators to create and manipulate motion effortlessly, underscores the growing influence of AI in the creative fields—and could have a place in other industries as well, including insurance.
From code to motion: empowering everyone to animate
One of the most exciting aspects of Carr’s work at Cartwheel is its democratizing effect on animation. While traditional animation can take years and cost hundreds of thousands of dollars, generative AI tools are making it possible for anyone to create high-quality motion with just a few keystrokes. “We’re able to turn every person into an animator,” Carr noted. This statement underscores the power of AI to make specialized tasks accessible to non-experts. And AI’s equipping effect here is mirrored by Indico’s Intelligent Intake solution—significantly increasing data intake and processing efficiency for insurance companies through advanced AI tools.
By integrating motion-generation capabilities with user-friendly interfaces, AI helps bridge the gap between technical expertise and creative expression. This also creates new opportunities for the gaming and entertainment industries. Carr discussed how game designers can now iterate faster while maintaining high-quality output, a key factor in keeping up with the fast-paced demands of the industry. “Enabling them to move much faster while maintaining quality, I think, is an appealing push for them,” he added.
The technical challenge of motion in 3D space
Carr dove into the technical side of generative motion, explaining how Cartwheel’s approach differs from conventional large language models (LLMs) like GPT. While LLMs excel at generating text, they struggle to handle the complexity of 3D motion, making Cartwheel’s work even more groundbreaking. The company built its own data models and hardware to better predict motion over time, offering unique advantages for animators and motion designers.
Unlike text or images, motion is a unique data type that requires specialized models. Carr explained that building these models from scratch allowed Cartwheel to ensure high-quality, editable output—something not easily achievable with existing general-purpose AI models.
Related content: Drive precision in risk decisions and business growth through automated underwriting
Business applications beyond entertainment
Although Cartwheel is currently focused on the gaming and entertainment industries, Carr and Wilde see vast potential for AI-driven motion in other fields. For instance, industries like insurance and manufacturing could benefit from using 3D motion to model risk, simulate workflows, and enhance training programs.
“I can imagine in insurance workers’ comp recreating accidents or looking at the geometry of individuals on a manufacturing line and looking for either areas of danger or areas of improvement risk areas. ‘Oh, those motions are going to lead to back injuries,’ that kind of thing,” Wilde remarked, pointing to the broader applications of Cartwheel’s technology for the insurance industry. Motion data could help identify areas of risk, enabling companies to improve safety and optimize workflows.
The future of AI in motion and beyond
As Carr discussed the future of generative AI, he emphasized that while the field is progressing rapidly, there are still significant challenges to overcome. One of the most pressing issues is the complexity of reasoning and planning in AI systems, particularly as it relates to agent-based architectures.
Carr’s view on the evolution of AI tools was cautiously optimistic. He acknowledged the immense potential for these technologies to revolutionize various sectors, from animation to insurance, but also noted the fragility of early-stage systems.
“Who knows where the copper is coming into the system,” he quipped, referencing an anecdote about an issue in the manufacturing process of an early computer chip. It turned out that an employee had a copper doorknob at home, and he was putting copper into the system just through his fingertips. That’s certainly a relevant story for any company building an AI model—the model could be affected by any number of variables, and it can be exceedingly difficult to identify them.
Related content: Transforming insurance with data: how telematics and AI are shaping the future
A glimpse into the future of AI-driven creativity
Andrew Carr’s work at Cartwheel offers a fascinating glimpse into the future of AI and its potential to reshape not only the creative industries but also sectors like insurance, risk management, and manufacturing. By making animation more accessible and enabling faster, more efficient workflows, Cartwheel is setting a new standard for how we think about data and motion.
As generative AI continues to evolve, we think that the line between creativity and technology will blur even further. For companies like Indico, which specialize in unstructured data, the lessons from Carr’s approach to motion and 3D modeling offer valuable insights into how AI can transform business processes across industries.
Interested in learning more about how AI is reshaping industries? Subscribe to Unstructured Unlocked on your favorite platform:
Subscribe to our LinkedIn newsletter.
Frequently asked questions
- How does Cartwheel’s AI differ from other generative AI technologies like GPT? While GPT models (like the one used here) focus on generating text, Cartwheel’s AI is specifically designed for motion in 3D space, which is significantly more complex. Cartwheel had to build its own data models and hardware because traditional AI models struggle to handle motion’s multidimensional and dynamic nature. This makes Cartwheel’s AI more specialized, focusing on motion prediction and animation rather than text or images.
- What practical applications of generative motion technology exist outside of gaming and entertainment? Although Cartwheel is primarily focused on gaming and entertainment, industries like insurance, manufacturing, and risk management can also benefit. For instance, 3D motion could be used to simulate accidents in insurance cases or analyze workers’ movements on manufacturing lines to identify potential safety hazards, optimize workflows, and prevent injuries. This can lead to more effective risk assessments and process improvements in these industries.
- What are some of the key challenges in developing AI-driven motion technology? One of the biggest challenges is handling the complexity of 3D motion, which involves multiple dimensions and temporal predictions. Existing AI models, like those for text, are not equipped to manage this complexity. Building specialized models from scratch, as Cartwheel has done, is necessary to ensure accurate, high-quality, and editable output. Additionally, AI systems face broader challenges, such as reasoning and planning, and are prone to fragility due to unexpected variables, as Carr highlighted in his anecdote about copper contamination in an early computer chip.
