Artificial intelligence continues to push boundaries for tech titans to sustain their disruptive innovations and competitive advantages.

Artificial intelligence continues to push boundaries for several tech titans to sustain their central disruptive innovations, competitive moats, and first-mover advantages. In recent years, the professional management consultancy firm PwC predicts that artificial intelligence can add more than $15 trillion to the global economy by 2030. This new economic output spans mega banks, tech titans, pharmaceutical giants, and e-commerce platforms such as Alibaba, Amazon, and Flipkart etc. In fact, rival prognosticators at McKinsey, Deloitte, and so forth offer similar predictions in the broad range of $13 trillion to $17 trillion.

Google CEO Sundar Pichai describes the recent modern developments in artificial intelligence as more profound than fire and electricity. Both smart computers and algorithms that can help perform the high-skill jobs of radiologists and lorry drivers etc may inadvertently cause a new wave of unemployment. This tech mega trend tilts the reversal of fortune toward public corporations and high net worth people whose work leverages artificial intelligence and smart robotic automation.

There is no doubt that artificial intelligence has made substantial progress in recent years. Specifically, machine-learning algorithms pervade technological advances in Internet search, software, online advertisement, cloud storage and computation, mobile connectivity, e-commerce, social media, and many more. Smart computers, algorithms, and cloud applications have become dramatically better in many fields such as facial recognition, natural language, financial fraud detection, econometric analysis of bank loan default and delinquency likelihood, economic recession risk prediction, or even stock market valuation etc. With vast cloud resources and data oceans, many econometricians, data scientists, and statisticians etc apply modern machine-learning algorithms and econometric methods to power search engines and voice assistants, suggest email replies, and further underpin most algorithmic systems that help identify unwelcome posts and comments on social media. These tech advances are quite remarkable. Many econometricians, data scientists, and statisticians who can harness the arcane intricacies of these quantitative methods, algorithms, and techniques etc can attract great fortune in the modern era of high-tech digital services.

Perhaps the key high-profile ubiquitous display of artificial intelligence is the recent algorithmic system built by DeepMind (a British subsidiary of Alphabet). DeepMind can manage to beat the best Chinese and Korean grandmasters at Go, an ancient Asian board game, in the rather short time span from 2016 to 2020. This high-tech breakthrough surprises many global megatrend observers and commentators as artificial intelligence continues to expand into several other fields, niche specialties, and blue-ocean markets. 

As Google CEO Sundar Pichai tries to compare artificial intelligence with fire and electricity, most machine-learning algorithms serve as general-purpose technology that can affect entire economies worldwide. The main machine-learning algorithms excel at recognizing big data patterns. Astronomers apply these algorithms to hunt for planets in glimmers of starlight. Banks use these smart systems, algorithms, or other quantitative tools etc to better assess asset market value fluctuations, credit defaults and delinquencies, and operational risk events such as fraud and litigation. In China, Europe, and India, regulatory agencies use machine-learning algorithms to monitor social welfare payments for mass surveillance.

Artificial intelligence heralds express cautious optimism that better transformations have yet to arise in the broader business context. Both computer developers and software engineers of autonomous cars predict that key robotaxis can revolutionize road transport. An artificial intelligence computer scientist, Geoffrey Hinton, points out that smart algorithms can perform most tasks of radiologists in the next decade. With calm optimism, Eric Schmidt, Former Google CEO and Chairman, suggests that artificial intelligence can help accelerate scientific research and discovery for key scholars and experts to keep up with the current deluge of big data. Elon Musk, serial entrepreneur and chairman of Tesla and SpaceX etc, welcomes the modern business transformations of artificial intelligence and automation with sustainable energy in road and space transport. At the same time, however, Elon Musk worries that artificial intelligence may inadvertently lead to the creative destruction of low-skill jobs in due course.

 

Artificial intelligence starts to stretch its data limits and cognitive restrictions.

Although modern artificial intelligence techniques can be quite powerful, they are sometimes narrow in scope and hence start to reach their inevitable limits. In some occasions, it can be hard and troublesome to deploy these smart machine-learning algorithms and other artificial intelligence quantitative tools and methods. Artificial intelligence experts, data scientists, econometricians, statisticians, and software engineers etc must confront at least 2 major problems. First, data are not always readily available for subsequent quantitative analysis. For instance, it is often hard to apply artificial intelligence to monitor Covid-19 airborne transmission without a comprehensive database of human movements worldwide. Even when data are available (in addition to high-performance computers and algorithms), the raw data can contain hidden assumptions, selection biases, and survivorship issues etc that can trip the unwary from time to time. For this reason, data scientists and software engineers often take time to cleanse these raw data in preparation for subsequent rigorous quantitative analysis or data visualization. The latest artificial intelligence systems and cloud machinery can be prohibitively expensive. These concerns and considerations can slow down the ubiquitous adoption of both artificial intelligence systems and algorithms due to data constraints.

Second, smart machine-learning algorithms often cannot embed human common sense into their artificial intelligence systems and algorithms in a reliable manner. Smart machine-learning algorithms use thousands and even millions of data points for econometricians, data scientists, statisticians, and software engineers to train new models. From neural networks to principal components, these models mimic the neural architecture of the human brain. The resultant systems and algorithms can perform some specific tasks and jobs (such as speech and image recognition etc) far more efficiently than traditional decision rules. Nevertheless, these artificial intelligence systems are still not intelligent in a universal way. Specifically, these artificial intelligence systems and algorithms can be powerful tools for recognizing statistically significant patterns in big data, but these systems and algorithms lack cognitive capabilities that biological brains inherit from generation to generation. Artificial intelligence systems and algorithms still mechanically reason, discover, and generalize with no common sense and basic intuition. In this unique fashion, most artificial intelligence systems and algorithms can excel at specific high-skill niche tasks and jobs, but these new systems and algorithms can inadvertently offer erroneous forecasts and predictions in out-of-sample tests. In many high-skill fields, we still need human experts to apply their institutional domain knowledge, common sense, and basic intuition to help guide actionable business insights and decisions with modern machine-learning tools and techniques.

In light of these 2 major problems (i.e. data constraints and basic human insights), artificial intelligence has yet to overcome these pervasive pitfalls to pass the proof of concept. At any rate, artificial intelligence continues to push boundaries for top tech titans to sustain their new disruptive innovations, competitive moats, and first-mover advantages. The recent artificial intelligence boom starts to stretch its data limits and cognitive restrictions. A full-blown machine-learning recession seems unlikely at this stage. After all, smart machine-learning data scientists now need to ensure the seamless integration of artificial intelligence and domain knowledge. In this important way, artificial intelligence and automation can become more general-purpose technological advances in the next tech boom.

Amazon GO stores are impressive places. In these cashier-less shops, customers use Amazon apps, pick up grocery items, and simply walk out with them. The main algorithmic system uses many sensors, panoramic cameras, artificial-intelligence-driven mobile devices, and wireless connections to track all of the items available on the shelves to ensure premier user experiences. This high performance accords with what Amazon CEO Jeff Bezos regards as customer-centric delivery services. Once the shoppers leave with their products, Amazon GO stores can automatically charge these shoppers retail prices and discounts with cloud software and artificial intelligence. Most Amazon GO stores demonstrate impressive real-time accuracy and so show what artificial intelligence can perform to help enrich the daily lives of many people with tons of continuous video data.

In theory, the world is awash with data, and these data serve as the main life blood streams of modern artificial intelligence. Some market research firms such as IDC and McKinsey predict that the world can generate more than 87 zettabytes of data from 2020 to 2025. Nonetheless, data issues are perhaps one of the most common blocks and obstacles in most artificial intelligence projects. In the case of Amazon GO stores, the necessary data may not exist at all. Alternatively, the tech company often has to simulate these data with virtual assistants and customers. Sometimes these data are only available in the vaults of close rivals such as Alibaba, Flipkart, and Walmart. Even when the relevant raw data become available, the tech giant still has to hire key data scientists, econometricians, and statisticians etc to cleanse the raw data into standard forms and relational tables for subsequent quantitative analysis. In due course, this quantitative analysis can lead to actionable business insights, decisions, and strategies. At any rate, the dual data transformation often requires much grunt work and time.

From time to time, domain expertise and institutional knowledge are vital for the vast majority of artificial intelligence systems and algorithms to become practically relevant and useful in real life. There are many examples of artificial intelligence systems that can beat human performance in some particular tasks and jobs. For instance, an early smart computer Deep Blue can beat the world chess champion. Another smart computer Watson outperforms a human champion in the quiz show Jeopardy. In recent years, an autonomous driverless car company Waymo builds its first prototype vehicle. Apple designs and improves its artificially intelligent voice assistant Siri available on iPhones and iPads. A new machine-learning computer program AlexNet makes substantive progress in image recognition and so sparks modern interest in deep neural networks. AlphaGo now defeats and dominates the South Korean grandmaster Lee Sedol at the ancient Asian board game Go. From 2017 to 2020, several artificial intelligence systems and algorithms start to match human performance at detecting some cancers and eye diseases.

Nevertheless, sometimes the data themselves can contain hidden traps. Machine-learning systems and algorithms correlate key inputs with outputs through multiple hidden layers. These smart machines, systems, and algorithms blindly follow the letter of the law, but often remain oblivious to its spirit in the broad business context. Selection bias can be another main problem in most machine-learning applications. In recent times, U.S. hospitals test more than 200 facial recognition algorithms and then empirically find that many algorithms turn out to be significantly less accurate at identifying black faces than white ones. This major problem seems to reflect a preponderance of white faces in the training data. Another recent study from IBM reports that about 87% of faces have light skin in 3 common and pervasive training data sets.

Several tech titans strive to ameliorate these data deficiencies. A quick fix entails substantially expanding the datasets to include more rare events and observations. Amazon designs virtual shoppers and grocery items to simulate customer-centric user experiences in Amazon GO stores. Facebook, Google, and Twitter etc deploy smart systems and algorithms to detect both fake news and misinformation in the months prior to the American presidential election. The new GPU microchip maker Nvidia comes up with a fast method of building synthetic data for autonomous cars. These key synthetic data often turn out to be more informative than real data. Apple and Microsoft apply smart artificial intelligence systems and algorithms to encrypt sensitive mobile devices for both business and home users. In essence, this smart encryption helps ensure better data quality and privacy.

Privacy is another fresh attraction of synthetic data. In law, finance, and medicine, artificial-intelligence-driven business organizations must comply with U.S. Health Insurance Portability and Accountability Act, California Consumer Privacy Act, and European General Data Protection Regulation (GDPR). It can be difficult for most tech companies to properly anonymize big data. Artificial intelligence systems and algorithms that train on synthetic data can circumvent this main problem in practice. Monte Carlo simulations and historical simulations are often close enough to reality. From time to time, tech titans learn to add a fair bit of statistical noise to real data to build sufficient synthetic data. Synthetic transactions are therefore fictitious, but it is quite plausible for machine-learning experts to ensure that these synthetic data share the same statistical characteristics of real data (e.g. mean, median, standard deviation, skewness, kurtosis, and interquartile range). Many artificial intelligence models are subject to drift or deterioration in model performance over time. Outside structural changes can often cause adverse changes in the quality and accuracy of both synthetic data and their smart decision rules. Languages evolve over time; customer behaviors change in response to mega trends, tastes, and preferences; and regulations further alter how firms and consumers interact in several business transactions. Covid-19 lockdowns, for example, dramatically shift the aggregate demands for products and services in some industries such as air transport, travel, tourism, retail fashion, food and beverage trade, and many more. In the business world, change is often the universal constant.

One of the recent main Monte Carlo simulation studies shows that the conventional industry practice often embeds a downward bias in the macroeconometric default probability adjustment for bank capital measurement. With hundreds of thousands of synthetic data points, several researchers and practitioners design a fast and accurate default probability adjustment through the macrofinancial cycle. This core alternative capital bias adjustment serves as a more pragmatic solution in contrast to the computationally intensive baseline adjustment by brute force. This capital bias adjustment can lead to several hundreds of billions of dollars of both retail and wholesale equity capital shortfalls for each of the mega banks worldwide such as Bank of America, Citigroup, Goldman Sachs, JPMorgan Chase, Morgan Stanley, Wells Fargo, and so forth.

This evidence has profound public policy implications in the broader context of the recent proposal for banks to hold substantially more equity capital. Specifically, the quantitative results bolster the business case for revisiting the 3%-6% core capital requirements under the new Basel bank capital regime. In stark contrast to these rather lenient regulatory equity capital requirements, the Monte Carlo simulation results suggest that the typical bank equity capital as a proportion of the total asset base must be as high as 13% to 26%. This broad range accords with the qualitative recommendations of the new proposal for banks to substantially raise their equity capital ratios. In practice, the higher equity capital ratios would be commensurate with open financial risk exposure that banks would face in a rare severe economic recession such as the Great Depression of the 1930s and Global Financial Crisis of 2008-2009.

With large-scale synthetic data, this Monte Carlo simulation study can help design a macroeconometric stress test for effective bank capital shortfall management. This simulation study further prevents most banks from actively seeking to engage in the strategic manipulation of regulatory core equity capital requirements. With this precaution, global bank regulators now need to impose effective limits on some risk parameters (such as probability of default (PD) and loss given default (LGD)) in internal bank risk models. In a nutshell, this recent research offers quantitative support for more robust total capital adequacy. This research endeavor serves as a new scientific microfoundation for the core thesis that most banks can become more stable by holding greater equity capital buffers to safeguard against extreme losses in rare times of severe financial stress such as the Global Financial Crisis of 2008-2009 and the recent rampant corona virus pandemic outbreak. When push comes to shove, the law of inadvertent consequences counsels caution.

 

Tech titans and mega banks apply smart machine-learning algorithms to ensure better user content curation and credit default likelihood analysis.

Almost 1.8 billion people use Facebook from day to day. These active users write posts and comments, share photos, and upload video clips on Facebook. On this global scale, social media operations become so large that only smart machine-learning algorithms can police original user content curation. In practice, Facebook uses both smart algorithms and human moderators to spot posts and comments that violate either country-specific laws and regulations or the social media policies on consumer privacy and misinformation. Therefore, these smart algorithms show many advantages over their human counterparts. These smart algorithms are fast, accurate, automatic, and iterative on a continuous basis. Social media algorithms scan thousands of posts, comments, notes, and messages etc per second and so help save substantial time and personnel costs.

In addition to Facebook, Google uses smart machine-learning algorithms as cloud software tools for many specific purposes. These smart algorithms refine Internet search results, target online ads, and track specific user profiles etc. Also, Amazon, Apple, Disney, HBO, and Netflix now apply smart machine-learning algorithms to recommend new products and TV shows for global end users. Twitter and TikTok both deploy smart machine-learning algorithms to suggest new users to follow their probable favorite key opinion leaders and influencers. These tech bellwethers use both smart machine-learning algorithms and cloud software systems to provide all of these online services with minimal human intervention. This pivotal competitive advantage serves as one of the main reasons why these tech bellwethers can achieve high share price multiples (e.g. 25x to even 33x P/E ratios, P/B ratios, and P/S ratios etc) with relatively small workforces. In this unique fashion, several tech bellwethers often convey their irrational exuberance about comprehensive artificial intelligence deployment.

However, there are several good reasons for the inexorable reality check on this comprehensive artificial intelligence deployment. First, it is often difficult for large enterprises to institute organizational culture changes that help incubate the next disruptive innovations. Even though these disruptive innovations can revolutionize future industries, large enterprises often lack both business agility and lean startup mentality to adopt artificial intelligence. Industrial firms continue to focus on heavy capital investments with no or little artificial intelligence, whereas, Internet giants often apply smart machine-learning algorithms to analyze huge amounts of user data for investor sentiments, demographic shifts, and other structural changes in consumer behaviors and preferences.

Second, key artificial intelligence experts and machine-learning data scientists are scarce and so can command luxuriant salaries, bonuses, and other cash rewards and incentives. Only tech titans and large hedge funds can afford to employ these high-skill workforces. For this reason, academia has become a fertile ground for artificial intelligence recruitment. Stanford, Berkeley, NYU, Columbia, Princeton, MIT, and Chicago Booth help the next disruptive innovators fulfill their tech-savvy dreams and aspirations etc. More and more academic economists, data scientists, statisticians, and software engineers serve as senior advisors and consultants to tech titans such as Facebook, Apple, Microsoft, Google, and Amazon (FAMGA) as well as international organizations such as the International Monetary Fund (IMF), World Bank, OECD, and Bank for International Settlement (BIS).

Third, a subtle but common problem relates to the main purpose of comprehensive artificial intelligence deployment. The major Canadian Moravec paradox suggests that even though smart machines can handle complex arithmetic logic, these smart machines and their concomitant algorithmic systems often still struggle with simple human tasks such as body language and locomotion. Another technical problem relates to the generic explainability of smart machine-learning algorithms. Because artificial intelligence systems learn from big data instead of following clear rules, it can be hard and tricky for subject matter experts to figure out why these algorithmic systems reach some particular conclusions from a fundamental perspective. In the worst-case scenario, many subject matter experts who harness high-skill domain knowledge often need to apply sound qualitative judgment to justify the quantitative results and outcomes that people derive from key artificial intelligence systems and machine-learning algorithms. Correlation cannot imply causation, so both human judgment and justification may inadvertently involve some sort of circular logic. As these smart machine-learning algorithms spread into nascent blue-ocean market niche specialties such as law, finance, and medicine, it has become substantially more important for subject matter experts and specialists to delve into explainable artificial intelligence.

For instance, almost any human can staff a customer support helpline. Very few can play the ancient Asian board game Go at the grandmaster level. However, it is much harder for artificial intelligence experts to design an interactive customer service chatbot in stark contrast to the superhuman algorithmic machine AlphaGo. The ancient Asian board game Go has only 2 binary outcomes (i.e. win or lose). It is thus relatively easy for smart machines to identify sequential steps that lead to these binary outcomes. Individual games often play out in zillions of unique ways, but the basic rules are few and crystal-clear. These highly specific problems are a good fit for artificial intelligence. By contrast, a customer call after flight cancellation can involve many alternative ways, routes, and options. For this reason, it is often highly complex for artificial intelligence experts to develop an interactive customer service chatbot (whereas, nowadays smart machines can beat the grandmasters in most ancient board games such as chess and Go).

Another good fit for artificial intelligence is the behavioral credit score FICO by Fair Isaac Corporation. The credit card fraud detection system Falcon aims at banks and credit card companies (such as Visa, MasterCard, and American Express etc) and therefore relies on smart machine-learning algorithms. The smart algorithms excel at identifying irregular patterns in huge piles of credit card transaction records. With predictive analytics, the behavioral credit score system can provide accurate credit card fraud detection results in light of both clean and readily available data. Subject to annual macrofinancial stress tests by the Treasury and Federal Reserve System, most banks now include the behavioral credit score FICO as an essential explanatory factor in addition to the current loan-to-value ratio, debt-to-income ratio, delinquency status, and macroeconomic time series such as the prime interest rate, unemployment rate, core inflation rate, key institutional fund growth rate, corporate default spread, Treasury term spread, S&P 500 stock return performance, stock dividend yield, and so forth. These explanatory factors help predict the probability of default (PD) and loss given default (LGD) as the primary risk parameters in both prime loan interest rate determination and equity capital quantification for first-lien mortgages, home equity loans, home equity credit lines, auto loans, credit cards, specialty loans, small business loans, wholesale corporate loans, and commercial real estate loans etc. In loan default predictive analytics, smart machine-learning algorithms can complement most logit regressions to meet both credit consumer demands and regulatory requirements. In light of all of these new tech advances, both mega banks and other financial institutions (such as insurers and credit card companies) learn to invest substantially in relational databases, high-skill human resources, and smart data analytics etc. These corporate tech hurdles remain high, so most smaller financial institutions still rely on qualitative expert judgment in new loan default assessments with few capital investments and relational databases.

 

Artificial intelligence starts to stretch the limits of holistic cloud software design with both big data and human content curation.

In light of the recent rampant corona virus crisis, most tech titans Facebook, Apple, Microsoft, Google, and Amazon (FAMGA) experience a huge deluge of both online fake news and conspiracy theories. This disinformation mishap demonstrates the major benefits of always keeping human moderators in the loop. Because human moderators see sensitive and private user data, these human moderators typically work in corporate offices with stringent security protocols and policies. For instance, knowledge workers cannot use personal emails and smart phones to handle these sensitive and private user data in most tech companies. When Covid-19 airborne transmission spreads worldwide, it is difficult for tech companies to enforce these security rules when their team members work from home for an indefinite period of time. This interim transition means greater reliance on smart machine-learning algorithms that help ensure safe benign online user content curation. Sometimes smart algorithms inadvertently remove harmless or innocuous posts, comments, and video clips on Facebook, Google, Twitter, and YouTube etc, whereas, these informative items seem to violate neither privacy rules nor security policies. Less human supervision would likely mean longer response times and more mistakes due to the imperfect nature of most artificial intelligence systems and algorithms. Artificial intelligence can often accomplish much with less manpower. At the same time, however, human moderators are essential helpers for smart algorithms to work wonders when this collaborative coordination comes to fruition in time.

A common combination of computer complexity and competition can suggest that computational costs often tend to rise sharply within a relatively short time frame. Google designs and develops the artificial intelligence language model BERT for search engine optimization with about 350 million internal parameters and almost 3.5 billion words of text from the key open-source online collaborative encyclopedia Wikipedia. These big data can help better train the artificial intelligence systems and algorithms with many more nuances and connotations etc. More big data often mean more computational costs. From Facebook to Twitter, one round of training data for the biggest models, systems, and algorithms etc often can cost millions of dollars in electricity consumption. A Californian artificial intelligence research firm OpenAI substantially scales up these smart machine-learning algorithms for video games to run thousands of microchips non-stop for several consecutive quarters.

All of these big data complications pose a conceptual challenge to the Moore law (or the master metronome for the high-tech semiconductor industry). This industry standard predicts that the cost-effective amount of computational power available per microchip almost doubles every 2 years. In the modern age of both big data and human curation, artificial intelligence starts to stretch the limits of holistic cloud software development with respect to the Moore law.

An influential U.S. venture capital firm Andreessen Horowitz points out that several artificial intelligence startups rent their computational power from external cloud software service providers such as Amazon Web Services (AWS) and Microsoft Azure. The resultant cloud costs sometimes represent at least 25% to 35% of total revenue and so serve as one main reason why these artificial intelligence startups may make for less attractive investments than old-school software companies (the latter of which often enjoy hefty 80%-90% gross margins and at least 50%-60% net profit margins). Nvidia CEO Jensen Huang suggests that the Moore law cannot be plausible any more. It is harder for new disruptive innovators to excel at artificial intelligence without cloud software support as the Moore law begins to run out of steam. For this reason, Stanford computer scientists Christopher Manning and Fei-Fei Li launch the recent National Research Cloud to help U.S. artificial intelligence startups, teams, and solo researchers etc keep up with the prohibitively expensive cloud software usage.

 

Tech titans now run their core business operations in accordance with new rules and principles for the ethical use of artificial intelligence.

Autonomous cars can now often illustrate the current limits of artificial intelligence. Driverless cars work in the same way as most other applications of smart machine-learning algorithms. Computers crunch huge piles of data to extract general rules about how autonomous cars work well on the road. More big data can help improve both the accuracy and reliability of key artificial intelligence systems and algorithms. With sustainable energy, for instance, Tesla electric cars beam big data soon back to headquarters that strive to attain key iterative continuous improvements in smart software for road transport. (As of late-2020, Tesla now surpasses Toyota in terms of total stock market capitalization: the former reaches almost $419 billion stock market cap, whereas, the latter can sustain no more than $203 billion stock market cap. With the same calculus, Tesla is now worth more than the total stock market valuation of BMW, Fiat Chrysler, Ferrari, Ford, GM, Honda, and Volkswagen etc. Many economic media co-anchor commentators such as Jim Cramer, Becky Quick, and Mark Cuban point out that Tesla is worth much more because the tech titan combines artificial intelligence software, autonomous navigation, and sustainable energy into each Tesla electric vehicle.) On top of the millions of real-world miles, Waymo autonomous cars continue to generate billions of miles worth of data with driverless road transport simulations in virtual environments.

A major problem arises from this context. Smart machine-learning algorithms are fundamentally statistical and so link inputs to outputs in specific ways in light of key in-sample training data. These smart algorithms inevitably cannot cope with what most software engineers view as unusual and unforeseen circumstances that are not long prevalent in those training data. Human drivers can usually deal with these unusual and unforeseen circumstances with no or few further thoughts and efforts. These circumstances can involve black swans, gray rhinos, or even light airplanes on the road etc. Computer vision systems pause when snow obscures road signs; stickers can cause autonomous cars to misidentify both road signs and even speed limits; and sometimes motorbikes and parachutes can baffle artificial intelligence systems in driverless cars. Smart machines and algorithms still struggle with these atypical circumstances. For this reason, these out-of-sample cases can pose real dangers and challenges for autonomous cars that heavily rely upon real-time smart data analytics. It is often hard for smart machines and algorithms to apply strategic human foresight to anticipate many unusual and unforeseen contingencies.

Humans are better able to cope with these oddities because top-down logic helps find out some unique way the world works when key bottom-up signals sometimes turn out to be incomplete and ambiguous. Most artificial intelligence systems and algorithms still lack these distinctive capabilities. Even the most tech-savvy neural networks can only mimic specific parts of the human brain. These smart machines and systems can be competent in their comfort zone, but even trivial changes can become problematic. From a fundamental perspective, brittle artificial intelligence systems and algorithms still lack the cognitive human capacity to reason outside the in-sample training data.

Similar problems are practically relevant to natural language processors such as Amazon Alexa, Apple Siri, and Google Translator etc. These smart machines and systems cannot understand the basic structure of human language. Some basic concepts such as verbs, nouns, and passive voices etc seem alien to these smart machine-learning algorithms. Artificial intelligence constructs statistical rules that connect the dots between keywords and phrases in different languages. However, these statistical rules and connections often tend to overlook the basic concepts, connotations, and implications of those keywords and phrases. In effect, artificial intelligence often blindly follows the letter of the statistical law of natural language but not its spirit. Current machine-learning systems and algorithms adapt new rules in a narrow way. Many artificial intelligence applications often need a lot more data to learn new tasks and jobs that human experts can master within reasonable time frames. Most human experts can learn from their past mistakes, setbacks, even epic failures, whereas, smart machines and systems cannot. It is quite ironic that artificial intelligence often cannot learn from human stupidity over time.

Artificial intelligence models and systems often learn solely on large amounts of text, speech, or imagery. By contrast, biological brains learn from far richer data than smart machines. For instance, babies can rely on sounds, tones, and voices etc as well as the body language of their parents in the richer physical environment. In this unique fashion, all these peripheral clues help anchor abstract concepts in the real world. In other words, most artificial intelligence systems and algorithms lack cognitive senses and capabilities that most humans can apply from day to day. The key substantive contents of human minds are so complex that smart machines often cannot embed such dense contents. In the foreseeable future, subject matter experts should learn to combine smart machine-learning algorithms with symbolic artificial intelligence systems to emphasize formal logic, hierarchical capacity, and top-down human cognition.

In recent years, the American top tech titans Facebook, Apple, Microsoft, Google, and Amazon (FAMGA) publish their own sets of artificial intelligence principles. In accordance with these principles, key artificial intelligence systems and algorithms should be socially beneficial enough to avoid reinforcing unfair bias for the safety and welfare of humankind. Some other tech companies such as Tesla, IBM, Intel, TSMC, Huawei, Foxconn, and so forth have made similar pledges and promises. Artificial intelligence safety remains nascent and thus continues to grow over time. Among the Internet giants, smart artificial intelligence incubation means that further investigations into its broader ripple effects seem to have lagged behind the bottom line. In the future, these tech titans should shift their current focus from high-skill improvements in both analytic accuracy and prediction to the ethical causes and consequences of artificial intelligence. From Facebook and Google to Apple and Amazon, the recent U.S. and E.U. antitrust scrutiny helps probe into these broader ethical, social, and economic concerns and considerations with respect to end user privacy, data safety, monopoly power, and job security etc.

 

Smart automation and artificial intelligence can cause the next creative destruction of low-skill jobs worldwide.

The International Monetary Fund (IMF) predicts that smart machines can cause the creative destruction of almost 100 million low-skill jobs in hospitality, tourism, food provision, construction, and transportation in most OECD countries by 2030. Both artificial intelligence and robotic automation power these smart machines. In due course, these smart machines replace young, female, and low-skill workers who face the high risk of layoffs or pay cuts. These new technological advances tilt higher pay and better employment toward high-skill knowledge workers in law, finance, medicine, and information technology etc.

As the IMF further shows, fewer than 20% blue-ocean niche specialists worldwide are in teleworkable occupations or live in rich countries with the tech infrastructure for effective remote work arrangements. This average figure tends to disguise wide disparities. The proportion is almost 35% in North America and Europe and about 6% in Brazil, Russia, India, and China (BRIC). As of mid-2020, the recent rampant corona virus pandemic outbreak does as much damage to low-skill employment as artificial intelligence and robotic automation can inflict over decades. Disruptive innovations destroy some jobs, but can create many others and so free people to accomplish many other tasks. For instance, the key advent of ATMs frees financial practitioners to carry out more complex jobs and tasks instead of dispensing cash to bank customers. Also, online banks empower most end users and consumers to transfer money without having to visit bank branches. In recent years, big data and smart machine-learning systems and algorithms further enable each financial institution to reduce the number of local branches.

Over time this creative destruction has been unfavorable for people. The new jobs cannot perfectly match the headcount losses in terms of both skills and locations. High-skill labor mobility is often much lower than blue-ocean market expectations. Many economic media observers and commentators agree that smart automation accounts for the mass destruction of low-skill jobs in recent decades in most OECD countries such as America, Britain, Canada, France, Germany, and Japan etc. In recent years, artificial intelligence seems to accelerate this mega trend that smart machines empower high-skill solo specialists to outperform mid-size workforces. This smart automation may not completely replace people, but artificial intelligence can substantially reduce both the number and quality of jobs worldwide.

Artificial intelligence systems and algorithms can empower many econometricians, data scientists, statisticians, and software engineers to apply large-scale synthetic data for running tests, simulations, and experiments in only days. Smart machines often complement scientific work, and this complementary work is quite important for high-skill subject matter experts, specialists, and senior advisors etc. As a result, smart automation and artificial intelligence can help free up human work time for disruptive innovators to focus on better business solutions. Nonetheless, this smart automation still lacks face-to-face social interactions and team efforts. At any rate, high-skill teams, solo specialists, and other management consultants etc can apply artificial intelligence systems and algorithms to work in a more efficient manner.

When new Covid-19 airborne transmission comes into play, smart automation and artificial intelligence can further empower high-skill teams and senior specialists to work from home. With effective remote video conference calls, the business agility factor prevails in prime professions such as law, finance, health care, information technology, Internet communication, and virtual reality etc. For instance, remote general practitioner appointments soar from 1%-3% to more than 90% during the recent rampant corona virus pandemic outbreak in Britain. In America, one health insurer finds that the online appointments substantially increase from about 10,000 per month to more than 210,000 in only one state during the Covid-19 pandemic outbreak. With artificial intelligence and automation, teleworkable health care can be as simple as a video conference call via Zoom or Skype. This smart technology facilitates substantive behavioral changes that the corona virus pandemic outbreak supercharges in recent times. Doctors can bill key online appointments in the same way as physical visits. Most patients can schedule remote appointments with no physical presence at local hospitals. Hence, online general practices can probably need fewer nurses, technicians, receptionists, and medical project managers etc. This contingency adaption further applies to shops, salons, hotels, restaurants, e-commerce platforms, and cloud software services. Without effective vaccines and medications, the global pandemic outbreak can lead to the more widespread use of both artificial intelligence and automation due to the essential needs for social distance and business resilience. After all, smart machines are less susceptible to global pandemic outbreaks.

Most modern economies must make technology work in their favor. As digitization continues to reshape the economic interplay of both soft skills and smart systems, governments should strive to invest in high education, broadband Internet access, and 5G digital infrastructure for new business adjustments. In time, these business adjustments help transform the common use of both artificial intelligence systems and algorithms into greater economic inclusion worldwide.

 

The current corona virus crisis highlights the imminent urgency of strengthening the global financial architecture.

The Covid-19 pandemic outbreak is the root cause of all macro stress tests for the global economy. With airborne transmission, the corona virus threatens most parts of the world. From early-2020 to present, more than $100 billion capital funds flow out of frontier financial markets. This figure is almost 5 times as much as the capital exodus during the Global Financial Crisis of 2008-2009. Global trade falls fast, soft currencies weaken against the American dollar, and commodity prices collapse in response to the global economic recession.

The current corona virus pandemic outbreak is a major reminder of how much the global economy relies on the American dollar for global liquidity. Banks borrow and lend to one another in the international interbank market, and this key market runs heavily on American dollars. The greenback entails about 85% of foreign exchange transactions worldwide. The greenback further serves as the most pivotal currency vehicle for cross-border trade and interbank settlement. Also, the American dollar remains the common disproportionate denomination for the vast majority of bonds sold to foreign investors. When banks refuse to extend credit, most countries can safeguard themselves against sudden liquidity shortages by holding sufficient U.S. dollar reserves. Amid Covid-19 economic policy uncertainty, the U.S. Treasury and Federal Reserve System both buy dollar swaps and bond repurchase facilities for foreign central banks. Such emergency actions can help bring credit spreads back down to pre-crisis levels. Proper Treasury yield curve control can help lessen the adverse impact of the current global macro recession that results from the recent rampant corona virus pandemic outbreak.

In response to Covid-19, the International Monetary Fund (IMF) moves quickly to create the short-term liquidity lines for disbursing $1 trillion cash assistance. The IMF continues to negotiate both bilateral and multilateral core credit arrangements worldwide. There is no requisite agreement of a supermajority of countries on the new allocation of special drawing rights (SDRs) despite widespread calls from the official communities. As of mid-2020, the IMF offers short-term debt service relief for at least 6 months to 29 low-income countries (or previous IMF loan recipients). As the IMF sister institution, the World Bank points to pandemic bonds as its fresh contribution to weathering the current corona virus crisis. For both low-income and rich countries, this ex ante instrument serves as the ideal form of insurance against public health shocks. As part of the G20 action plan, global finance ministers and central bank governors can now call for greater fiscal and monetary stimulus in the form of both paycheck protection programs and large-scale asset purchases etc. Global fiscal and monetary policy coordination proves to be vital and important in rare times of severe financial stress such as the current global macro recession.

 

The butterfly defect of globalization suggests that many systemic risks are endemic to our hyperconnective world.

A new corona virus carrier flies from Wuhan to New York, a computer virus invades a wireless hot spot, and U.S. subprime mortgage defaults trigger a global financial crisis. The super-spreaders of the main benefits of globalization (such as airport hubs, fiber-optic cables, and global financial centers etc) sometimes turn out to be the super-spreaders of the prohibitive costs of globalization too. With this butterfly defect of globalization, even small actions in one place can spread rapidly to cause substantial global effects, systemic risks, and ramifications in our hyperconnective world. At the same time, trying to stop globalization cannot remove these sudden threats and risks. Rather, this attempt may inadvertently amplify global threats and risks. High protective walls often undermine team dynamism for global cooperation that helps manage our common threats and risks. Protectionism can reduce capital investment, trade, tourism, and tech-savvy innovation. In comparison, multilateral cooperation helps create new jobs and high wages in both low-income and frontier countries that build better resilience in response to sudden threats and risks such as the current corona virus crisis and Global Financial Crisis of 2008-2009. These resilient systems can only be as strong as their weakest links. Stopping the current corona virus crisis must remain one of the top priorities for most countries in the next few years. Through the better governance, key staff, and capacity of the World Health Organization (WHO), most frontier countries continue to incubate effective vaccines and medications to alleviate the recent global public health shocks.

In recent decades, globalization has led to revolutionary changes that outstrip the slower evolution of supranational institutions such as the International Monetary Fund (IMF), World Trade Organization (WTO), World Health Organization (WHO), and World Bank etc. This mega trend causes an inexorable wedge between our complex systems and our methods for managing global threats and risks etc. From Covid-19 to artificial intelligence and smart automation, global systemic risks can overwhelm our otherwise robust economic architecture with sufficient labor market mobility and asset market stabilization. Specifically, the current corona virus crisis highlights our lack of immunity to natural threats and rare disasters. However, this pandemic crisis offers a new opportunity for G20 world leaders and taskforces to reset our global institutions and economies for better global risk management.

Our financial centers, trade hubs, digital hot spots, and other connective systems interact with one another through complex global networks. These global hubs and nodes often tend to concentrate in some particular geographic locations (such as financial centers and major ports and airports etc). This concentration makes the global networks both fragile and vulnerable to sudden unusual unforeseen threats, risks, and circumstances. Greater geographic diversification can help build better business resilience and faster information exchange. However, these key benefits have yet to find their way into antitrust regulations and risk management strategies. Multilateral cooperation can be quite vital and essential in this regard.

The Pareto principle suggests that 80% of the consequences often come from 20% of the causes. A small set of actionable business insights can help resolve a large part of global risks and threats. From time to time, supranational institutions such as the International Monetary Fund (IMF), World Bank, World Trade Organization (WTO), and World Health Organization (WHO) etc often manage to form top-down coalitions to tackle global financial crises, pandemic outbreaks, carbon emissions, climate changes, socioeconomic racial disparities, and so forth. These top-down coalitions require greater team firepower and coordination.

With better economic inclusion, these top-down coalitions often learn to take into account the important role of global tech titans in the global architecture. Amazon Web Services (AWS), Google Cloud, Microsoft Azure, Alibaba Cloud, Baidu Cloud, Ant Financial Group, and Tencent WeChat connect to most systemically important financial institutions (SIFIs) with both smart data and artificial intelligence systems and algorithms. Amazon Marketplace and Alibaba Taobao Tmall now serve as the biggest online platforms of global trade and commerce for face masks, ventilators, and other personal protective items. Facebook, Google, and Twitter have become the dominant default distribution channels and online media outlets for Covid-19 public health information. Apple and Google now both lead Western attempts at app developments for tracing corona virus contact and social distance. The current corona virus crisis cannot conform to our old mental maps. For this reason, it is vital and important for global institutions and economies to establish robust cross-border partnerships to better cope with Covid-19 airborne transmission, poverty, gender bias, racial inequality, environmental degradation, and many more.

 

Many central banks and treasuries now propose new fiscal and monetary stimulus programs in response to the economic woes of the current corona virus crisis.

Economic history illustrates that it is not unusual for countries to keep borrowing money even when sovereign default risk is high. A comprehensive review of more than 90 sovereign default episodes from 1827 to 2017 shows the usual experience to be sharp increases in both external and domestic debt in the gradual run-up to subsequent sovereign default. On average, it takes about 7 years for governments to resolve most sovereign default episodes. Unfortunately, debt reorganization can become a new game in town where the country debtor seeks to exchange higher future deficits for lower debt payments now. This strategic delay helps both sides bargain for larger capital infusions from most official creditors. These creditors can renew evergreen sovereign debt in order to temporarily make their balance sheets look better. In the worst-case scenario, the current corona virus crisis can lead to another lost decade in economic development with long delays in sovereign debt resolution.

In response to the economic woes of the current corona virus crisis, many central banks and treasuries now propose new fiscal and monetary stimulus programs for economic revival. These policy instruments include greater fiscal deficits, large-scale asset purchases (QE or quantitative easing unconventional monetary policy actions), near-zero and even negative interest rates, and substantial government expenditures in education, infrastructure, and capital investment accumulation for filling the output gap. The current global pandemic outbreak is a once-in-a-century rare shock that merits a generous response from many central banks, treasuries, and private creditors toward both rich and poor countries. Desperate times call for desperate measures. The resultant fiscal-monetary policy coordination often helps preserve global trade and finance for most countries to better weather sovereign debt problems.

 

The corona virus pandemic outbreak causes deep divisions in the global economy.

Covid-19 can cause adverse effects on the poor health of people whose daily lives expose themselves to greater contact with others. America has been afflicted with the highest numbers of both infections and fatalities because the U.S. has subpar public health standards of major frontier countries in light of lower life expectancy and substantial health inequality. Covid-19 and its attendant health ramifications cannot go away soon. The public fear of another pandemic outbreak can continue to linger in the next few years.

Covid-19 exacerbates economic inequality and so broadens the threats and risks from artificial intelligence and automation. Smart automation can replace low-skill workers in law, finance, healthcare, education, information technology, and so on. Zoom and Skype video conference calls can replace airline travel and face-to-face team dynamism. Amid corona virus economic policy uncertainty, there is greater public demand for face masks, ventilators, and other personal protective items etc. The global public healthcare regulators cannot fully contain Covid-19 until effective vaccines and medications become available to the general public.

In response to substantial income and wealth disparities and labor market frictions, it is vital and essential for most countries to better train high-skill workforces. These comprehensive work programs help reduce both the income and wealth disparities over time. Nowadays, the global economy is rife with market power, stock market ownership concentration, and labor exploitation etc. The current rules of the blue-ocean market game weaken central constraints on corporate power, deflect worker worries and concerns, and sometimes even exploit users, consumers, borrowers, students, and workers due to managerial entrenchment and rent protection. For instance, both the American and European regulatory agencies delve into greater antitrust scrutiny of potential anti-competitive behaviors of top tech titans such as Amazon, Apple, Facebook, and Google. Apple now keeps the exclusive license to extract 30% hefty taxes on apps sold via the iOS App Store, retains its dominant power in the global sale and distribution of mobile devices such as iPhones and iPads, and continues to expand into the blue-ocean market for multimedia services such as Apple TV, digital music, and virtual reality etc. Amazon shares the blame for exploiting major information about third-party sellers on the global e-commerce platform from America to Europe. Amazon Web Services (AWS), Microsoft Azure, Alibaba Cloud, and Google Cloud continue to dominate the nascent niche market for fast cloud software and high-performance computation. In recent times, the U.S. Department of Justice initiates an antitrust lawsuit against Google as the search engine giant manages to monopolize more than 80% of organic search traffic in America. In social media, Facebook and Twitter both have yet to alleviate antitrust concerns about user privacy, misinformation, and consumer protection etc. Tesla remains cautious about road safety as the tech titan combines artificial intelligence, big data, smart software for autonomous tech navigation, and sustainable energy etc into the electric car. All of these regulatory concerns and considerations pose new risks, threats, and challenges for global end users, consumers, shareholders, and other stakeholders (such as upstream suppliers, employees, and institutional investors etc).

Global institutions often learn the essential need for rewriting the current rules of the blue-ocean market game. Monetary policy decision-makers should shift focus from price stability, Treasury yield curve control, or inflation containment to asset market stabilization and maximum sustainable employment. Fiscal policy actions must help maintain progressive tax systems, which emphasize not only the market distribution of both wealth and income but also the fair redistributive transfers from the rich to the poor. Financial rules and regulations must constrain both the political clout and power of mega banks in order to circumvent the next subprime mortgage crisis, currency crash, domestic debt dynamism, or bank interest rate manipulation. Corporate governance laws must recognize the best interests of all stakeholders (not just shareholders). Labor legislation has to better protect workers and unions with greater scope for collective class actions. The current corona virus crisis calls for both structural reforms and culture changes in light of these key policy concerns and considerations for better economic revival, equality, and redistribution. When push comes to shove, the law of inadvertent consequences counsels caution. One size usually cannot fit all. There can be many different ways for both policymakers and regulators to skin the cat, and all roads ultimately lead to Rome. Covid-19 now reveals enormous cleavages across many countries. Policymakers and regulators must both strive to coordinate their fiscal, monetary, and other stimulus programs to pace the next global economic revival.

 

As of mid-2020, we list our proprietary dynamic conditional alphas for the U.S. top tech titans Facebook, Apple, Microsoft, Google, and Amazon (FAMGA). Our core proprietary alpha stock signals enable both institutional investors and retail traders to better balance their key stock portfolios. This delicate balance helps gauge each alpha, or the supernormal excess stock return to the smart beta stock investment portfolio strategy. This proprietary strategy minimizes beta exposure to size, value, momentum, asset growth, cash operating profitability, and the market risk premium. Our unique proprietary algorithmic system for asset return prediction relies on U.S. trademark and patent protection and enforcement.

1. 15.37% 6-factor dynamic conditional alpha for Facebook;
2. 14.21% 6-factor dynamic conditional alpha for Apple;
3. 15.58% 6-factor dynamic conditional alpha for Microsoft;
4. 14.90% 6-factor dynamic conditional alpha for Google; and
5. 13.83% 6-factor dynamic conditional alpha for Amazon.

 

Our unique algorithmic system for asset return prediction includes 6 fundamental factors such as size, value, momentum, asset growth, profitability, and market risk exposure.

 

Our proprietary alpha stock investment model outperforms the major stock market benchmarks such as S&P 500, MSCI, Dow Jones, and Nasdaq. We implement our proprietary alpha investment model for U.S. stock signals. A comprehensive model description is available on our AYA fintech network platform. Our U.S. Patent and Trademark Office (USPTO) patent publication is available on the World Intellectual Property Office (WIPO) official website.

Our core proprietary algorithmic alpha stock investment model estimates long-term abnormal returns for U.S. individual stocks and then ranks these individual stocks in accordance with their dynamic conditional alphas. Most virtual members follow these dynamic conditional alphas or proprietary stock signals to trade U.S. stocks on our AYA fintech network platform. For the recent period from February 2017 to February 2020, our algorithmic alpha stock investment model outperforms the vast majority of global stock market benchmarks such as S&P 500, MSCI USA, MSCI Europe, MSCI World, Dow Jones, and Nasdaq etc.

 

 

This analytic essay cannot constitute any form of financial advice, analyst opinion, recommendation, or endorsement. We refrain from engaging in financial advisory services, and we seek to offer our analytic insights into the latest economic trends, stock market topics, investment memes, personal finance tools, and other self-help inspirations. Our proprietary alpha investment algorithmic system helps enrich our AYA fintech network platform as a new social community for stock market investors: https://ayafintech.network.

We share and circulate these informative posts and essays with hyperlinks through our blogs, podcasts, emails, social media channels, and patent specifications. Our goal is to help promote better financial literacy, inclusion, and freedom of the global general public. While we make a conscious effort to optimize our global reach, this optimization retains our current focus on the American stock market.

This free ebook, AYA Analytica, shares new economic insights, investment memes, and stock portfolio strategies through both blog posts and patent specifications on our AYA fintech network platform. AYA fintech network platform is every investor's social toolkit for profitable investment management. We can help empower stock market investors through technology, education, and social integration.

We hope you enjoy the substantive content of this essay! AYA!

 

Andy Yeh

Chief Financial Architect (CFA) and Financial Risk Manager (FRM)

Brass Ring International Density Enterprise (BRIDE) ©