How can AI be enabled in online marketplaces? Former Vice President of Technology at IHG and General Manager of Consulting and Integration at Hewlett Packard, Tony Yeoh introduced the technical jargon in AI and machine learning. He spoke at Arcadier Inspire, exploring the different ways technology, design and business can grow into a holistic experience.
AI has become a trending topic – what is it really?
Artificial intelligence (AI), according to Tony, is any intelligent agent or device that perceives inputs from the environment, takes action based on the data collected and maximizes the chance of achieving its pre-defined goal. There are two types of AI, narrow and strong. The former is the current state of AI – what we have today – that is very programmatic and rules based. The latter, however, refers to intelligence that is able to perceive, recognize and has human cognitive abilities. In other words, they are what we see in science fiction films – advanced robots with human-like intelligence.
Tony dives further into the branches of AI, one of them notably machine learning. Machine learning is a subset of artificial intelligence that often uses statistical techniques to give computers the ability to "learn" with data, without being explicitly programmed.
Image ‘Machine learning framework’ taken from spectrumeffect
There are two main methods, supervised and unsupervised learning. Tony explains that supervised learning teaches the system how information can be classified into discrete classes of data. On the other hand, unsupervised learning means that the computer is seeded with unstructured and unlabeled data, and then taught to cluster, define and group the data. For instance, looking at a particular cluster, the computer system learns the characteristics that makes them appear similar. The next time the system comes across data with a similar pattern and nature, it determines that this data belongs to the particular cluster. Tony then brings us through the next steps of feature extraction, machine learning algorithm and how the process provides us with a predictive model.
Now that we have gained knowledge about machine learning, Tony shed light on deep learning, which can be considered as the more advanced subset of machine learning. Deep learning mimics the neural networks of our brain, going through several nodes, and performing tons of calculations and inference based on the inputs. When Google’s AlphaGo AI defeated the world’s best human player in a game of Go, much of the world was taken by surprise. AlphaGo Zero had such powerful artificial intelligence capabilities that it derived thousands of years of human knowledge of the game and invented even better moves of its own. Starting from a blank slate with only the rules of the game programmed, the AI began by placing stones on the Go board at random but swiftly improved as it discovered winning strategies. Basically, the AI capabilities learned purely by playing millions of times over. This is what is called deep learning.
But how can I put AI to use in a Marketplace?
Have you ever caught sight of a beautiful clothing piece in a photograph of a celebrity and really wanted to get your hands on that, but just could not find it no matter how hard you searched online? Rest easy, Alibaba has an image matching service for clothes, where users can buy the clothing they see in advertisements or in celebrity photos. Using image based search, the service attempts to find the best product matches in the Taobao Mail and Taobao Marketplace, Alibaba’s online e-commerce sites. Interesting, isn’t it? Let’s dig into more eye-opening examples of AI’s usage in marketplaces.
There is a common discussion that in about 10 to 15 years, radiologists may be out of jobs. Regardless of the accuracy of the prediction, AI systems are anticipated to be able to proficiently read x-ray and perform radiology diagnosis. Nuance Communications is the first open platform for developers, data scientists and radiologists to accelerate the development, deployment, and adoption of AI for medical imaging.
In a similar industry related to healthcare, Philips is providing access to advanced analytic capabilities to curate and analyze healthcare data, and offer tools and technologies for AI-based solutions. HealthSuite Insights, the AI platform for healthcare, is a marketplace for healthcare professionals to look at analytics in terms of the data that people submit or data collected around patients and healthcare.
Looking at these insightful examples of how AI can be used in a marketplace, it becomes obvious that it may be difficult for the passive and transactional approach of online marketplaces to continue to attract customers. Tony anticipates that with the integration of AI into online marketplaces, next-generation marketplaces are headed to become more proactive and relational in the future.
What’s a good place to start with data collection and analytics?
Tony asserts that when dealing with data analytics, we should always begin with the end in mind. The end goal of any business plan is often to bring about an improvement for customers. The benefit for customers could be to reduce the search time, provide more accurate recommendations of what customers are looking for, decrease the number of abandoned carts and more. Data should be collected towards that particular goal, and in a way that protects the integrity and privacy of customers. The main utility of the collected data is to examine its pattern and behaviour.
Tony goes in depth with data collection and analysis, watch the entire Arcadier Inspire Summit talk by Tony Yeoh here