News & Events
Highlights of the Symposium on Risk Management and Business Intelligence 2020 - Applying AI in Business
The Symposium on Risk Management and Business Intelligence 2020 was successfully held on 9th May 2020 (Saturday) in webinar format; the theme of this year’s symposium is ‘Applying AI in Business’. The Symposium is an annual occasion for industry experts to exchange their invaluable ideas on the development and application of Risk Management and Business Intelligence (RMBI). The event mainly comprised speeches by guests and a panel discussion, which greatly enlightened the participating professors and students on the new trends of AI.
At the start of the event, Professor Kai-Lung HUI, Director of RMBI program, HKUST, in his opening remarks , said it was a pioneering move to conduct the Symposium virtually under the COVID-19 pandemic. He added that the event had been supported by more than 50 organizations with over 2500 participants and had invited a diverse range of prestigious speakers from various industries over the past 10 years. Then, he briefly reviewed the topics of previous Symposiums and stated the purpose of the event - to combine the strengths of industry practitioners, professors and students from various fields so as to cover the best practices in RMBI holistically from both academic and practical perspectives. He ended the welcome speech by introducing the speakers and their topics.
The first session commenced with a presentation ‘Be Smarter and Be Quicker to Manage the Risk’ by Mr. Jay WANG, Head of Equity Derivatives Market Risk Management. He began with a scene from the classic movie Margin Call which showed AI beating the world champion in chess, thereby asserting the dominance of AI. He then gave a few applications of AI in the financial industry, such as improving documentation processing efficiency, investment advisory and risk control. He then detailed three applications: enhancing loan review automation using two generations of AI; measuring risk consistently and quantifying misbehavior using the trader risk analyzer; and using an AI risk radar to generate and process signals for decision-making. Mr Wang explained that AI technology is experiencing explosive growth along the rule-based, machine learning, deep learning and self-evolving stages, and that it has been reshaping the financial world and significantly impacting the job market. He used GPS driving as an interesting metaphor for AI trading - while GPS gives the most efficient route but may cause traffic congestion, AI trading makes trading more systematic but could result in a high market volatility and risk exposure. He ended his speech by expressing his optimism for the future of AI and encouraging the audience to keep an open mind.
Mr. WANG took three questions from the audience. The first question was whether FinTech or lending companies might potentially erode banks’ traditional market share through AI technologies. The speaker stated that AI was beginning to move into mainstream, so banks would need to upgrade their technologies so they could be more adaptable to change, such as adding more components of borrowers’ qualities. The second question concerned how to cope with the challenge of explaining AI model decisions to traders and regulators, given that AI technologies such as deep learning often lacked interpretability. The response was that there might not be a straightforward answer, but he believed that such explanations should start from basic levels, simple versions or small parts. The final question was whether AI was more useful in a stable or volatile high-risk market setting. His answer was the latter, because the challenge in a volatile market was detecting market directions for better portfolio management, and AI could predict them based on past experience at a level humans could not.
The second speaker of this session was Dr. Chao HE, Director of Wisers AI Lab. His topic was ‘Mining Business Risks and Opportunities from Media Big Data using AI & NLP’. He first explained how AI and media big data could facilitate smart enterprise decision making. Data, whether they were structured or unstructured, internal or external, had to undergo processing by AI algorithms and Big Data infrastructure before applications in various fields such as risk and compliance, investment, business intelligence and marketing could be developed. In light of our multiple roles of reading, publishing and disseminating information in the new media ecosystem, which consisted of the government, regulators, corporations and the general public, he mentioned that large amounts of business values could be derived from media big data. While AI and Natural Language Processing (NLP) could convert unstructured data into structured data for easier application, he cautioned that Chinese NLP was not easy because of the challenges in word segmentation and tackling text ambiguity. He then went on to discuss five intelligent financial solutions powered by his company, in turn, outlining how they addressed the customer pain points on the domains of Risk Management and Crisis Alert, Investment Research, Quantitative Trading, Policy Intelligence and Smart Compliance respectively. Before ending the presentation, he described how Wisers led a technology revolution and highlighted some of the company’s greatest achievements using AI.
Mr. HE answered six questions raised by the audience. The first few were technical in nature, inquiring about the limitations of NLP in deriving insights from public opinions, as well as the ways to resolve overfitting in machine learning, which he answered in brief. Mr. HE gave a particularly insightful reply about the technical challenges of NLP, pointing out that Cantonese contained many slangs and lacked proper grammatical form, and that social media posts might be biased towards the younger generation, and not representative of the whole society. The latter questions were more about industry trends and data privacy issues brought about by AI. In his reply, he said whether the financial industry preferred deep learning or traditional machine learning more was ambiguous - clients who understood AI more might be inclined towards the former, but in general clients were simply more concerned about how AI solved their business problems. Therefore, he argued that businesses should educate their clients on how AI could be combined with their operation workflow. He also warned the audience that AI had its limitations, and that some institutions might be better equipped to embrace it than others. In response to data privacy concerns, he emphasized that the correct practice was to focus on relevant resources in the public domain only in order to prevent privacy breach.
Following a short break, the second session began with a rather technical speech by Professor James LEI, Senior Director of Hong Kong Applied Science and Technology Research Institute and Adjunct Professor of the Department of Mathematics, HKUST. The presentation topic was ‘Opportunities and Challenges in DeFi’. He began with a poll for the audience, which revealed that 83% never owned a cryptocurrency, and that the members came from a mixed background in terms of their level of understanding of and exposure to blockchain. He then explained how blockchain had disrupted the market, revolutionized our usual databases and disintegrated the monopoly in Internet data. He brought forward the concept of Decentralized Finance (DeFi), which broke barriers and redefined the boundary between quantitative science and information technology in our era. Not only did he cite numerous applications such as Virtual Bank, Facebook’s Libra cryptocurrency and Chinese Central Bank’s DCEP, he also supplied a few examples of cryptocurrencies which included Bitcoin, ETH, Ripple and others. He stressed that the decentralized architecture of blockchain was of paramount importance because of its greater efficiency, lower costs and lower vulnerability to attacks than a centralized system. After outlining the limitations and security issues regarding blockchain, he emphasized that personal privacy and institutional integrity were both crucial for DeFi and thus recommended the use of Domain Specific Language (DSL) and Zero-Knowledge Proof (ZKP) for greater security. He concluded by reiterating that decentralization was advantageous and necessary.
A question from the audience to Professor LEI was whether China’s DCEP digital currency was actually against decentralization since the government could trace transactions easily. He responded that DCEP was controlled to some extent, but this would prevent it from achieving its full potential under decentralization. He opined that the key issue was determining where to draw the line, which he perceived as difficult since it involved governance and politics instead of technology alone.
The fourth and final speaker was Mr. David LIU, Managing Director and Head of Asia Pacific, Compliance, Risk and Diligence, Kroll | Duff & Phelps; the topic of his presentation was ‘The Growing Role of AI in the World of Risk & Compliance’. He complimented the previous speakers before proceeding to a brief introduction of his company. After that, he outlined the challenges faced by the risk and compliance industry, discussing how wide-ranging financial crimes conducted by threat players had posed substantial risks deep beneath the ‘tip of an iceberg’, including data breach, information theft and monetary losses. After revealing some stunning statistics on their huge financial impacts, he shifted his focus to the role of AI in combating them. He reminded us that AI had a wide array of sub-disciplines, although he would narrow it down to machine learning only. He detailed the Anti-Money Laundering (AML) due diligence process and elucidated how AI could overcome the hurdles of huge data volumes, high costs, tedious information processing and inconsistent results in traditional processes. Researchers could harness result clustering, result classification and ‘Learning to Rank’ for subject identity resolution, outlier identification and content screening. The speaker asserted that AI technology was advancing at an unprecedented rate and scale, and that many of its significant benefits were manifested in the risk among other business functions. He went on to explain how AI had brought benefits in regulatory compliance, finance and financial crime compliance, despite its challenges such as high costs and a lack of skilled talents. In conclusion, he advised businesses to combine machine learning, rule-based systems and human analytics into a whole in order to get the most out of AI, while emphasizing that the human element was still essential.
The last part of the symposium was a panel discussion on the topic ‘Risk Management in AI World’, moderated by Professor Kai-Lung HUI. He sided with Mr. David LIU on the viewpoint that human analytics was still necessary in AI and was inspired by the first three speakers’ vision on how AI would benefit the world. Before the actual discussion, he shared a few slides to point out a few salient trends of AI adoption. For example, AI was perceived to impact the insurance industry more than banks in terms of risk management. In addition, China’s breakthroughs and developments in AI would probably make her a leader within a decade from now. Nevertheless, it is undeniable that there would be significant risks from mass AI adoption, given that most firms did not engage risk specialists in AI implementation. He mentioned privacy breaches, cyber-attacks and discrimination as some of the potential negative outcomes.
To start the discussion, Professor Kai-lung HUI expressed his concern about how the massive amount of training data in AI algorithms might lead to privacy breach or discrimination issues. To illustrate this, he supplemented an example of unlawful media reporting of the bankruptcy of individuals. Hence, he would like to invite insights in data governance and scrutiny. Mr. David LIU commented that it was vital for companies that had implemented AI strategies to think from a holistic perspective and draw every department together for more comprehensive checks and balances. This would create a robust security system and eliminate insider threats more efficiently. He claimed that AI was not a bad thing itself, but we had to supply information cautiously and use AI appropriately. Professor James LEI led us to see through the lens of cryptographic technology as a privacy-preserving technology. He cited ZKP as an example, claiming that a trusted environment and infrastructure would overcome the privacy problem. He added that AI did not always need much data when a myriad of AI fields was moving away from deep learning towards Artificial General Intelligence (AGI). Continuing on from the privacy point, Mr. Chao HE seconded Mr. LIU’s idea that AI had to be combined with human knowledge. To reduce bias in judgement in this ever-changing world, he believed that data evaluation and the continuous evolution of AI models were critical, and that businesses should continuously acquire skills in efficient AI usage. Mr. Jay WANG thought that ‘technology is innocent’ and concurred with Mr. HE on the need for continuous improvement and appropriate data interpretation to reduce bias and errors. He added that businesses should build their models in a gradual manner, starting with easily understandable ones first.
The panelists reached a strong consensus that AI was here to stay, ushering in an era of digital economy. They generally opined that businesses should utilize AI to leverage available information wisely, develop data insights, facilitate decision-making and differentiate themselves. Professor HUI asked the panelists for advice in managing the risks from AI and preventing disasters. The guests agreed that the risks and dangers of AI were not negligible, but they were optimistic about its future and its benefits to mankind. Mr. WANG and Mr. LIU recalled respectively how humans had upgraded technology after the 2001 market crash and invented a computer for the first time. These interesting examples illustrated that humans have the potential to grow with disruptive inventions and put in place controls when necessary. Mr. HE stressed the need for regulating how AI and its systems were deployed, while Professor LEI recommended that AI needed to be decentralized to prevent it from developing into a monopoly to avoid its dangerous downside.
Three questions from the audience were addressed by the panelists in the end. The first one concerned whether AI only learns from historical events and whether high volatility might result from history not repeating itself. In response, the panelists mentioned that AI could learn the less evident latent factors behind data, such as the symptoms in social media, to capture signals and triggers for events. Cognitive-based learning would be a good methodology to merge data-driven machine learning with human knowledge, preventing excessive reliance on past data. The second question was how machines could distinguish fake information and news from real ones. While their patterns, behavior and language could be analyzed, the panelists recognized that it was a challenging task, suggesting that more incentives, more computational power and continuous improvement would be required in this area. Although AI was a tool to help minimize risk, it was still evolving and might not easily detect manufactured news published online. The final question was how companies could leverage AI to become competitive. According to the panelists, businesses should exercise caution and focus on their stage of development and assess whether the technology could solve their business problems. Apart from paying attention to budget control, talent investment, business culture and domain knowledge, businesses should ponder how AI could fit slowly into their operations instead of ambitiously viewing AI as a sacred tool that could automate everything immediately.
Professor HUI delivered some final remarks to conclude the symposium and thanked the speakers for bringing new perspectives, thoughts and inspirations to the event. The valuable occasion witnessed a passionate knowledge exchange among the panelists and audience around the topic of AI. Overall, the RMBI Symposium 2020 was a great success.
YUNG Chun Nok, Jorge, Year 3 in Risk Management and Business Intelligence (Class of 2021)