Five ways Generator at Sheridan is using generative AI to power innovation

Although generative artificial intelligence (AI) may be best known for its ability to create original content, the emerging technology also has tremendous potential to enhance research and innovation.

“In the life cycle of a typical research project, generative AI’s powers can be harnessed at every stage — ideation, literature reviews, grant writing, project management, data analysis, knowledge dissemination and more,” says Andrea England, Vice Provost of Research at Sheridan.

The names of two of Generator at Sheridan’s five research and entrepreneurship centres were recently changed to better reflect their longstanding and growing work in artificial intelligence: the Centre for Intelligent Manufacturing (formerly the Centre for Advanced Manufacturing and Design Technologies) and the Centre for Applied AI (formerly the Centre for Mobile Innovation). Meanwhile, Generator at Sheridan’s other three centres — Centre for Elder Research, EDGE and Screen Industries Research and Training Centre — are already incorporating generative AI in various projects.

“Advancing and leveraging generative AI is a priority for Generator at Sheridan because it is a priority for the communities we serve.”

– Andrea England, Vice Provost of Research at Sheridan

“Advancing and leveraging generative AI is a priority for Generator at Sheridan because it is a priority for the communities we serve,” England says, noting steps must be taken to maintain rigour and integrity, and to ensure protection of intellectual property and human participation data, when using generative AI in research. “From small businesses that want to adopt new technologies to entrepreneurs launching AI-focused products to community organizations seeking to better support their clients, we are hearing from our partners that this is an area of rapidly-increasing importance.”

Here are examples of how Generator at Sheridan’s five centres are applying generative AI to support community partners in a current project, each in a different way.

Centre for Elder Research

Imagine if virtual humans (computer-generated moving replicas of human beings) could be programmed to address social isolation felt by older adults, and voice recognition software used to monitor cognitive function and mental wellness. The promise of these singular innovations has the Ontario Brain Institute (OBI) and Sheridan's Centre for Elder Research (CER) exploring the potential of pairing the two to detect cognitive decline in its early stages.

"We've worked with a lot of neurotechnology companies over the years, and one of the constant challenges is how to make their technologies accessible and available," says Dr. Christa Studzinski, the OBI's Director of Business Development and Partnerships. "The concept of embedding voice recognition software into virtual humans is interesting because it turns the virtual human into a distribution channel. Eventually, that virtual human could pop up in an app and remind older adults to do a mental health or cognitive ability check, providing clinicians or nurse practitioners with valuable information in advance of an appointment."

“Generative AI has great potential to support remote health care delivery and monitoring, but it has to be done with care.”

– Dr. Lia Tsotsos, Director, Centre for Elder Research

Before such technology can be deployed into the world, however, it's critical that it be proven both viable and safe. CER's role in the project includes conducting a feasibility study, gathering user feedback and validating claims about digital humans' capabilities. "Generative AI has great potential to support remote health care delivery and monitoring, but it has to be done with care," says CER Director Dr. Lia Tsotsos, "recognizing each individual's unique considerations and multi-faceted experiences."

Early returns from CER’s research — which includes bringing older adults into the centre on multiple occasions to interact with the virtual humans and complete validated cognitive and mental health tests — have been promising.

"We had a hunch about how it might go using digital humans to interact with older adults who didn't grow up with computers, and we've truthfully been blown away," Dr. Studzinski says. "We've also seen insightful feedback, such as requests for the digital humans to be customizable so they can look more like the older adults they're interacting with."

Centre for Applied AI

What's the one place in town where you can borrow a lifejacket, create a laser-cut project and receive emergency access to food — all for free?

If you aren't aware that the Oakville Public Library (OPL) offers all these services and more, you aren't alone. "We're the town's best-kept secret," says Joseph Moncada, the OPL’s Director of Innovation & Integration. "Even with a user-friendly new website that features a robust keyword search, it can be quite difficult for our customers to find everything we offer because we have so many resources."

Sheridan's Centre for Applied AI (CAAI, formerly known as the Centre for Mobile Innovation) is evaluating ways to resolve that challenge by leveraging generative AI. Working under the mentorship of Faculty of Applied Sciences and Technology professor Dr. Nick Sajadi, students Parth Jigneshkumar Patel and Kunal Bajaj have trained a ChatGPT model with unbiased and relevant data from the OPL's website. The OPL is now testing the model to determine whether the chatbot can be used to enhance its website's Frequently Asked Questions system by providing customers with personalized support.

“The CAAI has been an unbiased expert, guiding us to the right tools... and demonstrating that this could be doable.”

– Joseph Moncada, Oakville Public Library Director of Innovation & Integration

"Training open-source models can be very challenging because their capabilities often differ greatly," Dr. Sajadi explains. "We had to find the right model for this project and then guide it to work within the boundaries of the OPL's website."

"The CAAI has been an unbiased expert, guiding us to the right tools that best fit our needs and demonstrating that this could be doable," adds Moncada. "Now it's a matter of us taking the model and determining how to best integrate it within our OPL.ca environment. Ultimately, we're hoping this will help customers quickly and seamlessly access library information, resources and services."

Screen Industries Research and Training Centre

Generative AI has already brought efficiencies to character animation through automation of facial animation and voice synthesis processes. Now, Sheridan's Screen Industries Research and Training Centre (SIRT) is optimizing synchronization of characters’ multi-modal movements, ensuring the actions of their bodies, hands and faces align seamlessly with their speech.

M-body, an open-source bilingual applied research project led by SIRT and funded by the National Sciences and Engineering Research Council of Canada (NSERC), is designed to enhance the distribution and advancement of generative animation tools. The project leverages SIRT's expertise in performance capture, character animation and interactive software development in collaboration with researchers from Durham College's Mixed Reality Capture Studio and AI Hub, Le Centre de développement et de recherche en intelligence numérique (CDRIN) and Le Laboratoire en innovation ouverte (LLio).

"Animators currently have to hand-animate the movements of these characters' bodies, which is very time intensive, or use a motion capture volume, which is very expensive," says SIRT software engineering lead Mike Darmitz. "This project is about increasing efficiency and lowering costs while maintaining the quality of animation required by creators. In addition to supporting larger production companies, greater access to generative animation tools will ultimately enable more people to be content creators."

The research project began earlier this year with identification of common pain points in character animation and an assessment of the current state of generative character animation research and commercial adoption. Those insights were used to determine three deliverables that were both technically feasible and would provide real value to industry stakeholders: i) producing generative character animation tools; ii) developing software that eases machine learning integration into content creation tools; and iii) building a dataset of interactions between multiple humans.

“We're closing the gap between where research is on this and what people are actually able to use.”

– Mike Darmitz, SIRT software engineering lead

M-body researchers plan to release an open-source beta version of generative character animation tools and a complete dataset of multi-agent conversational animations later this year — both of which would be free for commercial use — and expect final versions to be available by the summer of 2025.

"Not only do we want to help people adopt these tools, we also want to help advance the tools," Darmitz says, adding that most of the early generative animation technology and data has only been available for academic purposes. "We're closing the gap between where research is on this and what people are actually able to use."

EDGE

Sheridan EDGE has supported Project Management graduate certificate student Majdi Nawfal's generative AI start-up every step of the way, including serving as its first "client".

Nawfal's product — an AI-driven dynamic chatbot capable of providing personalized guidance and support to website visitors — was inspired by his experiences as a Sheridan Peer Mentor, when Nawfal often found it challenging to quickly provide answers to student questions. He initially pursued the idea as a capstone project with EDGE, where student engagement manager Dr. Nadine Sinno helped him see the product's broader business potential.

“Feedback really helped me improve the project's focus and development.”

– Majdi Nawfal, Project Management student and Aidvisor AI Solutions Inc. founder and CEO

Now participating in an entrepreneurship-focused co-op at EDGE, Nawfal has created two minimum viable products (MVPs) to serve as prototypes: AskEDGE, which fields general questions about the centre's services, and EVA (EDGE Virtual Advisor), a platform that provides entrepreneurial advice. Both chatbots are capable of conversing in many languages.

"We need a solution that can create equity in the service we provide and prevents students from having to wait for an appointment or a reply to their email," says Dr. Sinno. "AskEDGE and EVA could both help us reach more students and have multiple engagements with the same students."

In addition to meeting weekly with Nawfal and his team, EDGE also arranged two idea validation labs that provided feedback from a panel of experts. "That feedback really helped me improve the project's focus and development," says Nawfal, who went on to become the founder and CEO of Aidvisor AI Solutions Inc.

"One of the panel members also suggested that I could sell this product at a higher price point than I had in mind because automated 24/7 customer support could be such a great solution for start-ups. Small businesses spend a lot of money to attract prospective customers to their website, but if those people can't find what they're looking for, they'll bounce and are gone forever."

Centre for Intelligent Manufacturing

Nearly half of lower-limb prosthetics end up being abandoned by amputees, largely due to discomfort and sores caused by inadequate fittings.

Robotic Gait Simulator, a collaborative project between Sheridan's Centre for Intelligent Manufacturing (CIM) and global assistive device provider ProsFit Inc., aims to reduce that rejection rate by developing a digital twin to aid prosthetic fitting.

A six-axis industrial robot that mimics human gait movement patterns is used to identify pressure points of discomfort or blistering. Those results are then fed back to the digital twin platform with a finite element model to refine parameters, ultimately enabling better outcomes in the prosthetic manufacturing and fitting process.

“The application of generative AI presents a fascinating opportunity to further push the boundaries at the intersection of rigorous empirical research, robotics and AI — with the mission to improve human lives.”

– Christopher Hutchison, ProsFit co-founder and Chief Technology Officer

In phase two of the project, researchers are exploring the potential of generative AI to enhance data inputs that can account for a broader range of gait patterns. "Work that has been published does not provide equations or descriptions of how the gait pattern depends on a person's weight, height or terrain nature," explains Sheridan Faculty of Applied Science & Technology professor Dr. Nadim Arafa, the project's Principal Investigator. "AI can be useful in generating such data from different experiments and analyze these to get the exact effect. This work can also be extended to handle complicated cases such as walking down a ramp or climbing stairs."

"Apart from the value to be created in improvement for amputee populations, this exploratory research with CIM is highly innovative with the potential to unlock new areas of investigation related to biomechanics, prosthetics and rehabilitation," adds ProsFit co-founder and Chief Technology Officer Christopher Hutchison. "The application of generative AI presents a fascinating opportunity to further push the boundaries at the intersection of rigorous empirical research, robotics and AI — with the mission to improve human lives."

Learn more about Generator at Sheridan