The Evolution of AI: Can Investment Morality be Programmed?
May 24, 2018Back form China - InvestGlass was invited by Swissnex and the Swiss Embassy to present its artificial intelligence developments in China. An amazing roadshow organized by Felix Moesner, Swiss Consul & Swissnex China CEO and his team. A three days roadshow, in three cities, Shang Hai, Hong Kong, and Beijing. Each night Swissnex organized a round table confronting Swiss and Chinese A.I specialists. Media Credit - Felix Moesner swissnex China CEO | Consul swissnexChina.org Early fear around robot advisor. The early fear around robot advisor was that technology would replace what bankers do. Of course, it is now given that a robot can filter information faster than human do. For a set problem, a robot will react faster and smarter than a human…yet in reality robots have failed to capture even 1% market share of investable assets! Perhaps because what we called robots were often marketing window dressing. Actual retail robot advisor clients are "self directed", tech savvy, 30-50 YO clients - not exactly anyone. Disruption is visible in both “advisory” and "discretionary portfolio mandate". The game is to speed up the customization of portfolio rebalancing. What we have noticed in Europe : the bloom of institutional robots. A new generation of institutional managers building model portfolio which you connect to investment management platform like InvestGlass. A new generation of independant financial advisors outsourcing investment decision process : Made in Roma, Designed in Zurich… The question of quality control is left aside for now. Speed is crucial. In Switzerland, account opening remains a compliance department dilemma, frustrating digital clients - particularly foreign investors. However facial recognition techniques and do-it-yourself questionnaires can increase the pace of account opening and compliance check. Reducing from few weeks... to few days... In China, WeChat, the equivalent to our Whatsapp chatting application gathers so much information that credit scoring and lending process are reduced to less than 30 minutes! Data and A.I. explains part of it but another reason is slow API adoption in Switzerland as well as traditional IT vendors reluctance to open gates to third party API. As our Chinese colleagues, we believe that robotization is first step to speed up manual, repetitive and low added value tasks. Reducing a dozen staff members to just one with a good bundle InvestGlass + API + Fintech partner vendors. Investors want to be served like they would be in a Starbucks. Modularity of data, applications and A.I is mandatory to offer a mass customised experience. They want their order to feel “unique” - respecting a daily tribe ritual where client is king. Large Americano, Soy Milk, Low fat. Since the new European regulation, MIFID2, financial products’ risks are under scrutiny and should be compared against clients’ key information (KYC and more). Would Starbucks vendor ask you if you are caffeine, milk or aspartame intolerant? Would doctor check your allergic risk, cure success rate and DNA compatibility before he prescribes you a medicine? Well bankers are expected to do so. Without an artificial intelligence this process is just impossible! We found that China’s financial regulator does not require this level of risk control and price transparency…yet. However he is putting pressure on fixed interest products vendors 固定投资 - guaranteeing extremely attractive yields... In Europe, ingredients are standardized and it offers more flexibility for robotic models rebalancing. With one button a trader can rebalance 5’000 clients at once and checking 5'000 KYC at once. Reducing a whole day of work to one or two hours thanks to a banking A.I.. It does not mean that people are going to lose their jobs. It means that administrative tasks will be more efficient. Job description will change - advisors will focus on high value tasks. Call, empathy, storytelling, building communities etc... And what about robots? Can a robot feel empathy? Can a robot entertain a human community? Can we be rational without emotions? We don’t think so yet. Emotions are dynamic valuation with of recognizable similarities or facts. Machines can show some empathy if they mimic our goals. We human are granting “automorphism" to machines. This is the missing link for robots to predict what we need to feel good. Naturally, we are granting some herd empathy to robots. Advisors fears should not anymore be about a compliance KYC issue, cash wire, and portfolio rebalance… It should be about what role they wish to play in this new robotic/human flock... a cyborg environment. Media Credit - Felix Moesner swissnex China CEO | Consul swissnexChina.org Robots are programmed. At each Swissnex evenings in China, audience asked the same question: “Is it possible that one day robots will win over human?” My answer was clear : “Yes they already do on many aspects”. Robots are programmed to win over human at least when we human understand the process to optimize... for the rest it is far from reality just because we still don't understand our whole brain, independant organs etc... Furthermore, robots will be programmed with herd and empathy components to reflect and generate feelings. The field of “affective computing" is currently where InvestGlass focuses its research efforts. The good news is that artificial intelligence could be empathic. Empathic means understanding someone emotions based on someones’ referential. Moral values are also computed inside the machine. Affective computing is used to detect emotions, recording emotions into dialog and then generate sentences inside investment illustration. Mimicking emotions is possible and it increases credibility of an investment solicitation. At InvestGlass, emotional optimization is an important topic of research and development. Understanding moral values is not something self-evident. Media Credit - Felix Moesner swissnex China CEO | Consul swissnexChina.org Ethical decisions are based on our culture. Clearly information is not data. This was the main controversy in China. Chinese speakers where convinced that data quantity would solve the problem, whereas Swiss guests, perhaps with a Cambridge Analytica and GDPR in mind, thought that agile algo and some reinforcement learning could be sufficient. Indeed, context awareness is key. If we say it’s 25°, this data is not sufficient. We need to know if this is °C or °F. 25° C is the real data. But when we say it’s 25°C temperature IS NICE. Saying that it is pleasant means that we place ourselves in a context where 25°C is normal. Humid summer ShangHai versus dry summer in Geneva? At InvestGlass, we are focusing our A.I. algos on supervised and reinforcement learning - leaving unsupervised to more on-demand customisations such as trading patterns recognition, call report patterns recognition to improve clients segmentation. Even with small data versus the “big Chinese data fair”, we achieve to improve financial advisors efficiency by understanding moral values in investing, giving, lending, and saving. Of course it's about math and data but it's also influenced by randomness, moral habits, social norms, financial regulations which ignite an optimal decision. Yes, investment morality can be programmed and individualized.