Leading Digital Transformation – Artificial Intelligence – Jack Hong

Attended this event hosted by HeadHunt, as part of a brown bag lunch and learn series to promote lifelong learning. This learning session was conducted by Jack Hong, who is a co-founder of Research Room, a management consulting and data analytics company, as well as CEO of JCube Capital Partners, a company that uses AI in portfolio selection and management.

The aim of this session was for participants to learn how to execute some of the new business drivers using the platform business model, artificial intelligence, and rapid experimentation.

Synopsis of the backdrop against which this session was forumlated:

Digital technologies have virtually disrupted every industry and caused permanent changes in the way we work and live. A common misconception is that businesses can be ‘miraculously’ transformed by adopting new technologies in their existing business models. The hard truth is that impactful digital transformations have little to do with technology and everything to do with the mastery of the new business drivers: new customer behaviours, co-opetition, data and analytics, innovation, and enlightened leadership.

Presentation by Jack Hong

  • Important takeaways from the session
    • Digital transformation is about the business, not the technology
      • Business drivers have changed
      • Companies can use digital thinking and technology to enter any industry that they so wish
        • Platform business model
        • Co-operation – platform businesses connect 2 or more consumer types and are focused on building trust, convenience, availability, speed, scalability and stickiness, and competitors must come together
        • Data, analytics, and AI
    • Thus, digital transformation is the responsibility of the CEO and he cannot outsource this task
    • Successful digital transformation also requires a new way of working and a new mindset shift
      • This always clashes with the old culture
  • The priority for digital transformation should then be Leadership (business model) > Team working culture > Technology
    • Unfortunately the prevailing mindset on digital transformation is still about spending on technology rather than redefining leadership, refining business model, and changing the working culture
  • What is the impact of digital technology?
    • There is an unprecedented pace of technology innovations
    • With digital technology, the greatest threats to an industry come not from incumbents but from people outside who dream differently
    • Think Uber (transport), Airbnb (accommodation), Kickstarter (fundraising)
    • For the first time in history, Asia is a key contributor to the technology development of the 4th industrial revolution
  • What is digital transformation?
    • it is the fusing of digital and physical (IOT etc.)
    • The locus of Asia’s digital technology focus is slightly different from western
      • Rather than focusing on developing cutting edge tech, Asia’s digital leaders and unicorns are innovative in implementing digital solutions to tackle local and regional challenges and opportunities
      • E.g. Visa and PayPal go through other consumer and business transaction-based ecosystems (B&M or Online merchants) to reach consumers, while in Asia, users interact with WeChat or the Alibaba ecosystem to reach other consumer and business transaction-based ecosystems
    • One example of how disruptive these businesses are is Alibaba’s Hema (河马), which promises 30mins delivery within a certain radius of the store – this disrupts the refrigerator industry
    • Another example is a traditional eCommerce spectacles shop on TaoBao that has $10M annual turnover and is ranked one of the top eCommerce retailer on the platform, but in contrast, on WeChat chat where there are referral-based chat groups, there is a business that sells only cosmetic contacts and generates $100M of turnover a month
    • For transport-originated companies like Uber, Grab and Gojek (in order of present-day focus on transport), the greatest friction to them is the availability of drivers, and a lot of Uber’s valuation is underpinned by the technology that it possesses and its long-term plan in possessing future technology (e.g. self- driving cars or helicopters) and so it will be hard for Grab and Gojek to achieve the same level of valuation multiple
    • The responses from 191 senior executives in Southern Asia Pacific show that IOT, AI and big data analytics are the top priorities for them
  • The key ingredients for business transformation – data
    • There is a need for IOT and complementary technology
      • IOT: sensors that collect data for intelligent processing and execution, via the cloud, for e.g. i) surveillance cameras and computer vision, ii) climate, energy, light, geolocation, movements, etc. iii) Arduino (Arduino is an open-source electronics platform based on easy-to-use hardware and software)
      • Complementary technology: i) Big data and analytics, ii) Cloud computing and infrastructure, iii) Robotics: self-driving cars and drones, iv) virtual and augmented reality
  • What is the difference between big data, analytics, and artificial intelligence (AI)?
    • Big data is not “massive amounts” of data, it is unstructured data
    • Big data analytics is the capability to make sense of unstructured data
      • requires non-tabular type of data storage and retrieval technology
      • Files are typically large and may require batch parallelized processing
      • Techniques to quantify speech, video and photographs (so we can apply statistical and machine learning)



Explaining the intersection of key enablers. Source: Jack Hong presentation.

  • Digging deeper into AI
    • AI is a set of research designs and computational techniques used to train a machine
      • it is based on known / observed data and used to predict outcomes with the accuracy exceeding that of a bunch of domain expertise
    • There is a need to train AI for specific use cases – even though it can be particularly powerful/accurate in any given one, it might not be immediately applicable to another, and that is where domain expertise comes in
      • For example, AI can be used to train to allow a player in Super Mario to go through a game avoiding dangerous elements (fire) but this cannot be immediately applied to Zelda where fire is present to
    • The “ancestor” of deep learning is Perceptron, which was the first algorithm for pattern recognition, created in the 1950s by Frank Rosenblatt
    • Neural networks, like Perceptron, was a hot topic in the 1980s and 1990s, especially using it as a predictive model for finance, healthcare and control systems, much like today
    • What has changed and the critical technology elements supporting AI and its application today
      • The availability of data (vs. surveys in the past)
      • Next-gen computing architecture – the availability, cost and capability of computing power
      • Advances in algorithms
    • Critical use-case elements supporting the rise of AI
      • Rules-based or autonomous
        • in the past, a rules-based approach to real life problems was preferred
        • However, the computer problems we see today cannot be effectively solved just by using deterministic rules-based systems
      • We do not know what we do not know
    • The case of Google DeepMind Challenge Match – AlphaGo vs. Lee Sedol
      • AlphaGo was originally trained by mimicking humans and observing their plays and then playing itself
      • In the challenge, AlphaGo lost once to Lee Sedol and Google engineers realized the flaw was that it was trained using humans
      • Hence they took that element out of training, and only told AlphaGo Zero the rules of Go, allowing it to learn by itself
      • AlphaGo Zero (and then AlphaZero) became much better versions of its predecessors
    • AI missteps – many machine learning and deep learning algorithms are known to be black boxes – researches know how to make a machine reach a solution but we have not seriously questioned how it arrived at it
      • Microsoft’s Tay AI chatbot generating racist tweets
      • Tencent’s Baby Q and Little Bing chatbot generated anti-government responses (and was subsequently taken down)
      • Facebook AI bots started talking with each other in a language of its own
      • Uber self driving car knocked down and killed a pedestrian in Arizona
    • AI technology simply models whatever guidance it is given and without proper guidance, potentially extremely detrimental missteps could happen
  • Another exciting algorithm – Generative Adversarial Network (GAN)
    • It is the culprit behind DeepFake
    • Involves 2 deep neural networks acting against each other
    • The Generator network attempts to create forgeries of an actual thing (e.g. painting)
    • The Discriminator network tries to call out the fakes
    • Initially the forgeries are of low quality, but after iterations, it is hard to tell real from fake


Harnessing the power of AI. Source: Jack Hong presentation.

  • We can, and should harness the power of AI together with human capital
    • We want to automate repetitive tasks
    • Augment the decision making process, applying tools, models and data to drive decision making (this is where the best value for money is) – this will be a compulsory managerial skill set in the near future
    • Then if the decision maker wants to go a step further, they can choose to amplify the process, which is to use AI to enhance domain experts’ skill sets, to create AI tools and models that is feedback and applied in the augmentation process
  • The innovation process has to change
    • Traditionally, it followed the steps of: Strategic meeting -> Market research -> working meeting -> develop -> internal testing -> launch -> market feedback
      • However, at many times this is way too slow before a product even gets to market and the emphasis should be to test and iterate
    • The approach should be to go through a ideation round and have a few ideas, with each resulting in a Minimum Viable Product (MVP) and then getting market feedback to improve and test again, ultimately then deciding whether to continue or terminate quickly
      • This means that in parallel, different ideas and products are being tested
  • There needs to be a mindset shift of rethinking objectives from the value proposition perspective
    • The first critical input to the Scrum is the backlog, a list of everything there is to do in order to complete the project
    • Then in the Rapid Experimentation process, we want each iteration to come out with deliverables that has demonstrable value
      • To allow customers to give accurate feedback
      • To allow us to test whether the deliverable is on the right track
      • To show management demonstrable progress
      • To allow the team to understand why we are doing something
    • Hence we have some expectations as to the quality of the backlog
    • In the Scrum, we do not use “task”, but “stories” and the size of the story is known as “story points”
    • The “who” and “why” are included in the task description in addition to the “what”
    • To increase productivity of scrum teams, rather than setting deadlines, the team is gathered to think about 3 wastes, as defined within the Toyota Production System (waste through unreasonableness, inconsistency, and overburden) and it is a process by which waste is eliminated

Further reading