Dr. Jason Corso, a leading expert in computer vision and AI, shares his insights on how AI is revolutionizing the retail industry, from enhancing self-service kiosks to personalizing shopping journeys.
Artificial intelligence (AI) is rapidly changing the landscape of retail, with significant implications for self-service technologies. To understand how these advancements are shaping the future of customer experiences, Kiosk Marketplace spoke with Dr. Jason Corso, a leading expert in computer vision and machine learning, and Co-founder and Chief Science Officer of Voxel51, via email interview. Dr. Corso provides valuable insights into how AI can be leveraged to enhance self-service kiosks, improve operational efficiency, and create more seamless and personalized shopping journeys.
 |
Dr. Jason Corso, CSO and co-founder of Voxel51 |
This Q&A explores the potential of AI to revolutionize self-service in retail, from optimizing kiosk interfaces and product discovery to enabling advanced automation and personalized recommendations. Dr. Corso also addresses the importance of ethical considerations and data privacy in AI deployment, offering practical advice for retailers looking to integrate these technologies responsibly. His expertise sheds light on how AI can empower self-service solutions to better meet customer needs, drive sales and shape the future of retail interactions.
Q: Given the rapid advancements in AI, particularly in computer vision, how do you envision the future of retail experiences being transformed?
Corso:Advancements in AI will transform the retail market giving customers better experiences both online and in-person. Retailers will use their improved insights into what shoppers want in order to increase personalization and maximize touchpoints across the customer experience.
Retailers who deploy AI will be on the front edge of redefining the industry, gaining a powerful advantage over the competition. Visual information plays a critical role in understanding consumer behavior and experience, from how shoppers navigate stores to how they assess the attractiveness and value of a product. The recent surge of investment in AI has led to rapid advances, making them available to a broader range of organizations and applications than ever before. We have also seen a range of camera and sensor solutions designed for retail and increasingly built for use with computer vision to capture and process pictures, video and imaging data from stores, warehouses, and delivery vehicles.
Q: Voxel51 focuses on building platforms for video and image-based applications. How can retailers leverage these platforms to enhance customer experiences, improve operational efficiency and gain a competitive edge?
Corso: Here are a few examples of how computer vision and visual AI can improve customer experience, both in-store and online.
We’ve all had the experience of wandering up and down aisles of a store, unable to find a specific item, let alone someone who can help us get an item that’s hard to reach or under lock and key. Computer vision technology and analysis can address this by processing data from video cameras to identify shoppers who seem to be circling in search of something or struggling to reach an item, and then notify staff so that an associate can be sent to offer the customer assistance. Especially in stores with large footprints, that can dramatically reduce customer frustration while empowering staff to provide more timely and efficient service
Computer vision technology also helps retailers organize their stores in ways that improve customer experience. By analyzing customer movements throughout the store and correlating that information with customer shopping baskets, retailers can improve store layout to reduce the time customers spend walking to different parts of the store. They can also use computer vision algorithms to identify items that are most frequently difficult for customers to reach so that those items can be placed on more accessible shelves.
There are also opportunities for advanced automation of checkout processes, moving beyond current self-checkout systems to more seamless «grab-and-go» experiences where items are automatically detected and tallied. In addition, retailers can offer enhanced virtual try-on experiences, particularly for online shopping, which will help bridge the gap between digital and physical retail experiences.
New examples of ways to use computer vision to enhance the retail customer experience continue to emerge as adoption grows. Amazon and other retail innovators are already using AI systems to create new shopping experiences, such as visualizing how clothes will fit customers or how furniture will fit in their spaces. Integration with other data sources creates possibilities for even more personalized customer experiences, such as enhancing virtual try-on experiences with visual suggestions for complementary accessories. It also creates new opportunities for targeted promotions, including noticing increased traffic for certain products or delivering real-time promotions via digital signage to convert a surge in interest to shopping baskets and sales.
Q: One of the challenges in AI is ensuring fairness and avoiding bias in algorithms. How can retailers ensure that AI-powered solutions in areas like personalized recommendations or customer segmentation are ethical and equitable?
Corso: Before the work begins, the first step is to identify potential bias-relevant groups and elements to build datasets, metrics and evaluations to support the assessment. Examples of groups and elements that are relevant to retail are age, gender and employment strata. Throughout model-building, AI engineers must continue to study model performance across these various groups and elements in detail through metrics, visualizations and interaction. In our efforts to meet ethical and equitable personalization, it’s critical the engineers designing these AI systems understand the relevant domains and data and then leverage the right tooling to support it.
Q: How can retailers effectively integrate AI technologies into their existing infrastructure and systems while ensuring data privacy and security for their customers?
Corso: With protecting privacy paramount, it’s plausible to adapt contemporary AI capabilities to integrate sensor feeds that preserve it. For example, instead of transmitting or working with regular camera data, we can transmit transformations of such data that preserve the relevant semantic information — e.g., the information required to support understanding a shopper’s behavior while browsing in an aisle of a store — while removing identifying information. This could mean blurring faces at the edge (at the camera site itself) or transforming raw RGB pixel data to only transmit silhouettes of the humans, among many other ideas.
Q: What are the key skills and knowledge that aspiring professionals in the field of AI and retail should focus on developing to succeed in this evolving landscape?
Corso: Given how fast this field is changing, collaboration via open source is essential for AI professionals. They can unleash faster innovation in the retail space with open AI systems that feature freely available source code, data, and parameters that can be viewed, modified, adopted, and shared. Open source collaboration has long been the cornerstone of software development, empowering the Linux operating system, for example, to become the foundation of modern computing and power everything from personal computers to large particle accelerators.
By following the example of the software industry and democratizing access to AI models, data, and tools, AI builders won’t get stuck recreating the wheel. A recent IBM study of 2,400 IT decision-makers reflected a growing interest in using open source AI tools to drive ROI, with faster development and innovation at the top of the list when it came to determining ROI in AI. The research also suggested that embracing open solutions correlates to greater financial viability. Open source AI invites the creation of more diverse and tailored applications across retail which might not otherwise have the resources for proprietary models.
About Dr. Jason Corso
Dr. Corso is a professor of robotics, electrical engineering and computer science at the University of Michigan. His research focuses on video understanding, human-AI collaboration and the intersection of vision and language. He received his Ph.D. from The Johns Hopkins University in 2005. He has also co-founded Voxel51, a company that assists other companies in utilizing visual data to create AI models and applications. Dr. Corso has received awards including the National Science Foundation CAREER Award and Google Faculty Research Award.