Kindred, co-founded by Prof. Graham Taylor, is a robot that can manage warehouse organization.
By Samantha McReavy and Mya Kidson
Artificial intelligence (AI) technology is all around us—although we may not always be aware of it—from self-driving cars to online product recommendations, practical speech recognition, text prediction and even the development of a globally accessible COVID-19 antiviral pill. AI is constantly being used to improve lives.
Dr. Graham Taylor, an engineering professor at the University of Guelph and interim Research Director at the Vector Institute, is working to improve AI technology in a variety of industries. He works to improve machine learning and increase machine independence while still allowing a human to step in and help the system on occasion. Machine learning encompasses multiple strategies to create code without a programmer writing it.
“We use machine learning to train the AI systems,” says Taylor. “The system learns by analyzing data, either provided by human input or collected from the environment, to develop its own code.”
AI technologies continue to be a growing part of our world and play a critical role in improving life. AI helps to make life more comfortable, safe and efficient, he said. These technologies have also shown promise in helping industries pivot during COVID-19.
Taylor says this type of learning allows systems to write software when it becomes too complex for humans—such as software for self-driving cars. The software created through machine learning is made faster, more accurately and more systematically than software designed by humans. Deep learning, the branch of AI Taylor develops quickly organizes the data observed into multiple layers to identify patterns and associations in the data—something that humans do naturally but have a hard time articulating and thus coding it as traditional software.
Bringing AI to e-commerce
During the pandemic, many people turned to online shopping to avoid public crowded spaces. As a result, big companies like Amazon and Gap required more warehouse workers to sort and send products. However, these jobs are rarely sought out by people—and that’s where Kindred comes in to fill the gap.
Kindred, an AI company that builds intelligent robots, has developed a robot called SORT that manages warehouse organization.
Before COVID-19, Kindred’s robots were only picking clothing items. To reduce human contact with grocery items during the pandemic, there was developing interest in using robots to aid in the grocery supply chain.
Ocado, Britain's second largest grocery retailer, recently acquired Kindred systems in November 2020. Locally, they provide the technology behind Sobey’s food delivery service known as Voila. The Kindred team in Toronto and San Francisco now conducts research and development for Ocado.
Taylor, co-founder of Kindred, helped establish the team that built SORT. This intelligent robot is more than 90 per cent autonomous— the machine works and learns on its own. Less than 10 per cent of the time, it calls on a human “robot pilot” for assistance.
When SORT experiences a situation it doesn’t understand, a human needs to intervene. As this happens, SORT collects data on the process to improve its future performance. This allows the robot to continually become faster, smarter and more accurate.
Kindred received funding from many Canadian investors, as well as investments from Silicon Valley in California.
Taylor is the academic director of the founder development program NextAI, part of non-profit Next Canada’s portfolio. He invites leading academics to teach technical courses that will provide participants with the appropriate computer science knowledge to further their AI-focused venture. He also teaches in the program himself.
In addition to technical and business training, the program provides funding, office space and a network of mentors to help new start-ups succeed in a competitive environment.
He is also the academic director of Guelph’s Centre for Advancing Responsible and Ethical AI. The centre affiliates researchers from all seven colleges on campus who are working to ensure that new autonomous platforms are safe and benefit humanity.
Using cellphones to improve wound treatment
The healing progress of a wound can now be tracked via cellphones, thanks to a collaboration between Taylor and the Toronto-based start-up app Swift Medical.
Taylor works to increase the app’s image segmentation accuracy, or the process of outlining wounds and separating them from healthy skin to monitor the healing process.
Current wound measuring strategies—using a ruler and cotton swabs—can be uncomfortable, particularly when measuring wound depth. The Swift Medical app will improve patients’ experience while being nearly two-thirds faster than the traditional technique.
This technology reduces the time required per nurse visits. Limiting the duration of visits can not only reduce COVID-19 transmission risk but also shorten wait times in the health-care setting.
The Swift app can be used across the entire health-care continuum, including long-term care facilities, hospitals, wound care clinics and home health agencies.
“This technology acts as a tool for medical support workers, giving them additional information to monitor patients and discover issues that might have been overlooked otherwise,” says Taylor.
Swift Medical approached Taylor because of his work in computer vision—creating machines that identify useful information from a single image or sequence of images.
Using computer vision, the app was trained to visually differentiate between various stages of healing. It’s a difficult task because image quality can be affected by multiple factors including lighting, hair growth and skin pigmentation.
To overcome these challenges, Taylor incorporated machine learning techniques that enabled the systems to learn how to distinguish lesion severity over time.
This collaboration was funded by the Ontario Centres of Excellence and the Natural Sciences and Engineering Research Council.
Future of Taylor’s research— improving machine confidence
As AI systems continue working closely with humans, the systems need to become aware of their own limitations. AI systems can learn quickly and function more accurately than a human, but they can still make mistakes. These systems need to be able to recognize uncertainties and transfer control back to humans. Taylor wants to make AI systems more aware of their limitations and ensure that systems effectively interact with humans to optimize the technologies’ potential