Big data and artificial intelligence have been slogans of innovation for a long time, but in recent years we’ve seen how these technologies are actually disrupting industries across the board. At Rakuten Optimism 2022, a panel of experts showed how the latest tools in data and AI are changing business practices from startups to multinationals.
Four big trends in AI
Ways to exploit data and AI can be broadly grouped into four main trends, according to Ting Cai, Chief Data Officer, Group Senior Managing Executive Officer, Technology Services Division, Rakuten Group. For one thing, they have become an increasingly important and practical part of our everyday lives, as seen in services like automated driving, driver alerts and language translation. For another, AI is being applied to all industries. Rakuten, for instance, is using AI to leverage data for marketing, to improve logistics and get products to customers faster.
In addition, AI is becoming more creative, with dedicated programs producing songs, paintings and other content, not to mention generating software code, Cai noted. Finally, AI is becoming more generic in that its less specialized and more of a general-purpose tool, a longtime goal of AI scientists.
“AI is changing the world, but our dream is even bigger,” said Cai, who was Google’s Senior Director for Geo Search and Assistant before joining Rakuten earlier this year. “We want to help every business, big or small, to realize their potential, and to help every individual to realize their dream.”
Nuts, bolts and cancer cells
Dr. Ewa Szymanska, Global Head of Rakuten Institute of Technology, Executive Officer, Rakuten Group, gave some detailed examples of how AI is being deployed for industrial applications.
“AI needs to be explainable. We need to combine human intelligence with artificial intelligence so we can enhance each other.”Ting Cai, Chief Data Officer, Group Senior Managing Executive Officer, Technology Services Division, Rakuten Group
For instance, AI-powered drones are inspecting cellular network towers, not only checking for damage but actually identifying individual bolts holding towers together and verifying their tightness. This eliminates the need for workers to manually check thousands of bolts, focusing resources on the few components that need attention. In a similar labor-saving innovation, AI-equipped mobile devices can be used to automatically detect whether marketing materials and merchandise displays at Rakuten retail shops are up to date and properly arranged. Meanwhile, Rakuten Institute of Technology is developing AI software to automatically detect tumor cells and biomarkers in cancer screening imagery.
“Doing all this with the naked eye is an incredibly time-consuming if not impossible task,” Szymanska said of the cancer screening process. “With AI, we can process thousands of images, generating invaluable data for understanding biological mechanisms supporting clinical trials and ultimately moving us closer to a future of personalized and effective cancer treatments.”
Understanding language naturally
Aside from more accurate and practical image-processing and analysis, AI is being used to improve our lives through better understanding of language with something we do every day: online searches.
California startup Divinia has developed natural language processing technology that can enhance online searches, eliminating many irrelevant results. Users input precise descriptions of what they want instead of the conventional basic keywords, and the Divinia search engine sifts through millions of reviews to find the best results.
“As part of our secret sauce, we have unsupervised learning technology that will allow us to pull out almost everything that consumers care about from user-generated content such as reviews,” said Divinia founder and CEO Ji Fang, who did a live demo at Optimism by searching for hiking shoes. “It does so at scale and low cost. Rather than huge models at high cost, our models can be built with a few small machines.”
There’s a graph for that
Another AI tool known as graph neural networks (GNNs) is also making searches more effective. GNNs are better than conventional deep learning techniques when it comes to dealing with unstructured data, for instance relationships between people in a social network. Toyotaro Suzumura, a professor at the University of Tokyo’s Graduate School of Information Science and Technology, told participants how GNNs can be used to learn how consumers are related to people they know or products they tend to buy. Using GNNs for search, for example, can better connect people with what they’re looking for by understanding what’s popular in their social networks.
“Deep learning is a tool to automatically extract features such as recognizing a cat or dog,” says Suzumura. “In graph neural networks, we have methods to learn how you connect to your friends or products…so we can recommend very similar items.”
Cai pointed out challenges surrounding further AI adoption including making it affordable for smaller businesses and startups, making it effective with limited data and making it explainable. He compared the importance of explainable AI to finding a cooking video online that not only shows how to make sushi, but explains why every step in sushi making is necessary.
“That’s the spirit I want to learn from AI as well,” Cai told Optimism participants. “AI needs to be explainable. We need to combine human intelligence with artificial intelligence so we can enhance each other.”