Skip to main content

M.R. Asks 3 Questions: Cupid Chan, CTO, Index Analytics

By July 25, 2019Article

Recently, Cupid gave a presentation to the group called “How do I teach my second grader about AI”? The talk centers on how you can explain AI to a non-technically proficient audience. But it also gives seasoned tech audiences a different take on what AI truly is and does.

CTO of Index Analytics and a Fellow at the University of Maryland, Cupid speaks around the globe on issues of Big Data and AI. We had a great conversation about the future of AI and his ability to make the topic comprehendible for “non-tech” audiences.

M.R. Rangaswami: As a CTO, what’s your prediction on hot technologies in the next 5 years?

Cupid Chan: I like to refer to my predictions as the “ABCDs” of emerging technologies.

Artificial Intelligence – The anticipated applications for AI are broad, ranging from robot-assisted surgery and virtual nursing assistance, to fraud detection and reducing drug dosing errors. Applied AI is also expected to positively impact preliminary diagnosing and identifying potential patients for clinical trials. Even better, with AI, the probability to get an accurate result will keep improving over time when more data is taken into the system to train the model, which will then be applied back to generate more data and form a continuously-improving positive cycle.

Blockchain – I would call AI and blockchain the “Yin-Yang” of Emerging Technology. While AI is probabilistic and the system will keep changing and improving as time goes on, blockchain offers a deterministic mechanism by which once the data is entered, no one can make any change. This provides a foundational “Trust” platform people can rely on without concern about being “cheated” since transactions are transparent. Moreover, due to its distributed nature, no single organization can control and manipulate the result. In practice, blockchain can be applied to drug or device distribution to ensure an item is authentic and from the trusted source.

Cloud – Just like no modern family nowadays has an individual generator to supply its own electricity, it’s much more efficient to obtain the “computing” power as a service from the cloud, as we get electricity from a power plant. This is particularly important as the usage of computer can be high sporadically. It doesn’t make sense to procure a full rack of servers just to handle a once-a-year enrollment period for Medicare or Black Friday weekend sales in the retail industry. Cloud, in this case, provides a scalable and flexible way to support the IT need of an organization.

Data – From the macro perspective, cloud provides an infrastructure in which IT can reside. From the Micro perspective, data provides the building blocks in the cell level to support what the business needs. Years ago, when we talked about data, it usually referred to structured data that traditional relational databases handled pretty well. Now, unstructured data such as emails, notes or audio recordings are abundantly available, and valuable in various forms of analytics. Data is the fuel to spark AI and how the “D” loops back to “A” making “ABCD” the basis of the fast-moving evolution of the IT era. 

M.R. What’s your take on the “older” technologies, like Business Intelligence (BI)?

Cupid: Even though my very first job was working for a BI company, I am not afraid nor embarrassed to say that BI is dead unless it incorporates AI and other approaches into the overall IT strategy.

Technology has transformed to a point where having an interactive dashboard is a basic requirement, though it was state-of-the-art technology at some point in history. We are now in the era where traditional BI can no longer serve the growing needs of modern analysts and executives.

We need more proactive insights harvested from the vast amount of data available. That’s why I proposed an “AI + BI = CI” theory at the recent Conference on Health IT and Analytics (CHITA), organized by University of Maryland. The true intelligence we need to pursue now is Cognitive Intelligence (CI), which combines the speed of how a machine can learn (which AI is very good at) and direction intuited from human insight (supported pretty well by traditional BI platforms). Missing either one will lead you to the wrong destination, or delay progress towards a milestone.

M.R.:  Not many people have a chance to study a full semester about AI or machine learning. Can you tell us in simple terms, what is that?

Cupid: Glad you asked! This is how I explained AI to my second-grade son when we were in the line waiting for a ride at Hershey Park. I borrowed this quote from Alan Turing and said to him:

“Imagine you have a computer to do work for you so that nobody knows whether you or the computer actually does the work.”

This is the very first “definition” of AI. But my son then asked if that meant AI can do homework for him and his teacher wouldn’t find out!

So, then I explained further “Supervised Learning” in terms of how he taught his younger sister the difference between a pen and an apple. After showing her examples repeatedly, she eventually understood different features of “red”, “round”, “long” “pointed” to identify the correct label of “Apple” and “Pen.” This is the same concept for machine learning.

Of course, as a baby grows, she starts to learn and figure out patterns by herself. So, when my wife asks my kids to clean the playground without detailed instruction of how, they categorize various toys together by group. Coincidently, just providing the goal and letting them recognize the underlying pattern has a counterpart in machine learning, called unsupervised learning.

Last but not least, though supervised and unsupervised learning is universal for teaching a child, they are not always the most effective. Another common practice is to reward good behavior and punish a bad one. This “reinforcement learning” also applies to machine learning and is especially helpful in robotics. In summary, supervised, unsupervised and reinforcement learning are not only good for teaching kids, they are also the 3 main concepts of AI machine learning.

Based on this true story of how I taught my son the meaning of AI through these simple, yet critical AI concepts, I have developed and given a presentation at various national and international conferences. I also put up a video on this very same topic on my YouTube channel. 



M.R. Rangaswami is the Co-Founder of Sand Hill Group