AI, Machine Learning, and Big Data in the Investment Management Industry


•Artificial Intelligence, Machine Learning, and Big Data are everywhere recently.  It seems like you can’t turn on the news without hearing about them. There is no doubt that AI will have an effect on all industries, similar to the way the internet or electricity did.  It is clearly already creating positive developments in the healthcare industries by helping Doctors to make better decisions and diagnoses.  But Are Elon Musk, Stephen Hawking, and Bill Gates right?  Does this technology pose a threat to our civilization?  (Twilight Zone Music here…)  These men are much smarter than I am, and their opinions obviously are worth consideration, but that is beyond the scope of this presentation.  (At least until we get to the ‘Best AI Movies’ page!)

•There are many confusing issues with regards to these fields. One reason is because there is really no precise definition of what AI is.  Adding to this confusion is what has come to be known as the ‘AI Effect”.  This means that as soon as AI can solve a problem, it is no longer considered AI.  A lot of cutting edge AI is already in general applications in various technologies, but when it is commonplace it isn’t considered AI anymore.  This is why some say that “AI is whatever hasn’t been done yet.”  For example, when Deep Blue beat Gary Kasparov at chess in 1997 it was considered a revolutionary accomplishment for AI.  Now, twenty years later, Deep Blue is considered conventional technology and there are numerous other chess programs that can regularly defeat the top human players. 

•My intent with this presentation is to make this topic less confusing by showing you that AI isn’t as mysterious as some in the media would have you believe, and to help you understand how Artificial Intelligence, Machine Learning, and Big Data can add Alpha by being integrated into investment processes.


What is Artificial Intelligence?

•To understand what Artificial Intelligence is, we need to understand what intelligence is.  Intelligence can defined as the ability to make predictions regarding the future based on observations and inferences, and the ability to set goals and determine ways to achieve them.  There are different types and levels of intelligence.

Animal Intelligence - Any pet owner knows that animals have intelligence.

  1. When I take my dog for a ride, I use a leather leash.  When we are going go for a walk in the woods, I use the long extended leash.  Clifford can predict which action we are going to take based upon the leash I grab.  If I take the leather leash, he runs to my truck.  If I take the extended leash, he predicts that we are going for a walk, so he runs to the woods.  He is making a prediction and that is intelligence.
  2. Clifford can also set a goal and predict a way to obtain it.  If I am eating something and he sets a goal of having some, he knows that if he begs, barks and hits me with his paw enough, I’ll eventually give him some of what I am eating. He makes a prediction that his actions will help him obtain his goal, because these actions have done so in the past.  Clifford set a goal, and thought about ways to achieve it.  That is intelligence.

•Human Intelligence - The next level is human intelligence.  Our intelligence is typically greater than animals because our brains are bigger, and we can evaluate more variables when we make predictions or think about what we need to do in order to achieve our goals.

Artificial Intelligence / Synthetic Intelligence -  AI is software and computers that can do things that we define as intelligence.  They can analyze the past and use the knowledge to make predictions and inferences about what will happen in the future.  They can set goals and take actions that maximize the chances of achieving them.


What is Machine Learning and Big Data?

•Machine Learning is when an AI algorithm has the ability to evaluate results and to adapt and enhance itself as time passes.

•Big Data is a term for sets of data that are so large and complex that conventional or traditional methods are incapable of analyzing them.

•Advanced Artificial Intelligence techniques can analyze these data sets and identify patterns. These patterns illustrate relationships between behaviors and outcomes.  Once they are known they can be used to make inferences and predictions about the future.

•This Data comes from repositories…for example Google has 100 billion searches a month, Facebook has over 60 billion photos, and the types amounts of data that can come from cell phones is staggering.

•The definition of ‘Big Data’ is a moving target.  What was considered Big Data five years ago is no longer considered Big Data, and what is considered Big Data now probably won’t be considered Big Data in five years.

•Alternative Data is data that hasn’t been traditionally used for analysis.  It includes data from phones, social media etc. Like Big Data, this definition is also a moving target.  What would have been considered Alternative Data five years ago may now be considered conventional by some, and what is Alternative Data now may be consider conventional in five years.


Why Now?

•The concepts of Artificial Intelligence and Machine Learning are not new. 

•Hephaestus was the Greek God of metal working and blacksmiths.  In Greek mythology, he made the weapons for all of the other Gods.  He also built self-operating machines out of metal.  These included walking tripods and mechanical servants.  In the early 1500s, Leonardo Da Vinci built a mechanical lion that had the ability to walk.

•The term ‘Artificial Intelligence’ was first coined in 1956 at a Dartmouth University academic conference. By the late 1960s it seemed to be a very promising field which was mostly funded by the Department of Defense.  For various reason in the mid -1970s, interest was diminishing and funding dried up.  The late 1970s are called the ‘AI Winter’ due to the lack of research and funding.

•There was some renewed interest in the early 1980s with the introduction of ‘expert networks’, but for various reasons the era didn’t last long and there was another AI Winter in the late 1980s.

•The late 1990s and early 2000s saw incredible advancements in technology.  AI has been on the rise ever since to bring us to the incredible place where we are today. Recent advances in technology and massive increases in computational ability have made the application of Advanced Artificial Intelligence viable and profitable.

•In 2006 there was 100 exabytes of data on the internet.  Recent estimates put this number at 10,000.

•There have been massive decreases in costs.  For example, a gigabyte of data that was $300,000 in 1981 costs just pennies today and computational power that cost millions of dollars in the early 1980s now can also be purchased for pennies.


Examples in Society

•Healthcare Industry - There are approximately 800 medicines and vaccines that are used to treat cancer.  This negatively affects doctors because there are too many options to choose from.  Microsoft is working on a machine called ‘Hanover’ which will try to study and memorize academic papers in order to predict which combinations will be most beneficial for individual patients.  In addition, it has been proven that AI is better at recognizing signs of cancer on a medical scan than most experienced doctors are.

•Commercial Banking Industry - Credit scoring was one of the first things in the financial services industry to be automated almost seventy years ago.  It used to be that the president of the local bank could decide who could or couldn't get a mortgage or a loan.  Now a sophisticated AI architecture does.  It can make better predictions than conventional methods because it can examine vast amounts of data and find predictive patterns. 

•Gaming and Entertainment - In 1997 Deep Blue beat Gary Kasparov at chess.  In 2011 IBM’s Watson beats two human contestants on Jeopardy, one of which was the record holder for most wins.  This task took five years.  Watson spent four years learning English and then another year reading and retaining every word in Wikipedia as well as thousands of other books.

•Automotive Industry - There are more than thirty companies that are working on the creation of driverless cars and other types of vehicles.  This area is a popular topic of conversation.  People generally say that they will not get into a self driving car.  It is important to understand that commercial aircraft are flown by automated programs, and the pilots are on the plane so the passengers feel comfortable.  This is why flying is much safer than driving a car.

•Sports Industry - Professional sports organizations are employing AI in numerous way.  They are scouting players based on statistics.  There is a movie about this called "Moneyball".  Formula One and NASCAR cars have hundreds of sensors which generate terabytes of data in an attempt to gain an edge.  Such data points could be fuel usage, temperature, humidity, tire pressure and so on.


AI, ML, and Big Data in the Asset Management Industry

•Investment managers across various styles can integrate Artificial Intelligence, Machine Learning, and Big Data into their investment processes in order to enhance performance and reduce risk.

•Managers who can overcome ‘algorithm aversion’ and learn how to integrate these technologies will have an edge.   Managers should not fear being replaced by robots.  

•There are numerous ways that this can be accomplished.  There is no ‘one size fits all’ application.  AI Architectures can be built for specific strategies.

•Artificial Intelligence can help lagging Active Managers compete with and surpass Passive Managers.  Passive strategies by definition do not need to enhance performance so they will not be able to take advantage of the advances in technology. In five or ten years AI will be ubiquitous and ‘baked into the cake’, but until then the active managers who embrace it will have an edge over the ones who don’t.

•Perhaps the most important dynamic for a manager to consider is that it will save time by increasing accuracies and efficiencies.  


Various Investment Styles can be Enhanced

Many managers would like to utilize AI because they understand that once it is developed it is more accurate than humans, it is unbiased, and it is faster and cheaper than its human counterparts.  However, they are unsure of how they can take advantage of it.  Here are some practical examples that illustrate how these technologies can be integrated into different types of investment styles:

•Quantitative Strategies - Can use AI to identify and exploit inefficiencies in the market.

•Traditional Long Only Managers - Can benefit from AI by using it to enhance their portfolios by optimizing sector and industry weightings.

•Screening Process - Consider an International Microcap manager that has a mandate to hold between 50-100 positions and has an investable universe of 3,000.   AI could benefit the strategy by being able to screen out and identify possible investment opportunities.

•Deep Value and Fundamental Managers - Can use AI to read and find predictive patterns in financial statements such as 10-Ks and 10-Qs, in addition to macro data.  They could also incorporate alternative data into the analysis.  Consider the stock of a retailor.  An AI model can analyze traditional data, such as accounting data, financial statements, macro data, etc.  It could also incorporate alternative data, such as google earth images of parking lot traffic, mentions on Twitter, Website visits etc. 

•Event Driven Managers - Can use AI to identify predictive patterns in the markets following meaningful events. These events could be political, technical, macro related, company specific etc.

•Trade Timing - AI architectures that predict short-term movements in markets can be utilized to enhance trade entry and exit points.

•Hedging & Risk Reduction - AI can be used to identify securities that will reduce downside risk and volatility.  These can be used to hedge a portfolio.

•Distressed Debt Mortgage Backed Securities Funds -  AI to can be used to build more robust and accurate portfolio valuation models than conventional ones.  Conventional  models may evaluate between 20 -200 variables.  An AI model could evaluate thousands of variables.  The data can be obtained by zip codes and include things such as interest rates, employment rates, taxes, mortality rates, average age, employment opportunities, public transportation, regional, economic conditions, garbage collection and so on.  It could analyze this data and find predictive patterns.

•Idea generation -  AI can be used for idea generation.  For example, if an investment firm wants to identify and invest in companies that are embracing cutting edge technology, an AI Model can be built to read the financial statements and identify and analyze predictive patterns in Research and Development spending and progress.  Sentiment analysis can also generate profitable ideas.

•Enhance Back-testing and Statistical Analysis - Back testing and statistical analysis can be enhanced with AI.


Checklist for PMs

•Artificial Intelligence, Machine Learning, and Big Data are not as mysterious as they seem, and if used correctly the can enhance portfolio returns.  How can a Portfolio Manager take advantage of these technologies?  

•They need to have a detailed discussion with a Data Scientist who can understand their strategies and needs.  Good data scientists want to deliver useful solutions that work in real world environments, and not to be too caught up in theory.  AI should be viewed as a set of technologies that can automate tasks that have previously been done by people.  

  1. What is the strategy?
  2. How do you wish to enhance the strategy?
  3. What rules and applications will apply?
  4. Where will the data come from?
  5. How will the process be operationalized?
  6. How will success be determined?
  7. What are the initial and ongoing costs?


Quotes Related to Investment Management

“It isn’t some sort of magic pixie dust.  It’s just enhanced statistical techniques that allow you to build better predictive models.”   Adam Duncan, Managing Director at Cambridge Associates

"The industry is on the cusp of a technological arms race where an intelligent program is poised to join your team as a machine co-pilot."  Gary Brackenridge, Global Head of Asset Management at Linedata

"It's all about seeing how far we can push the machine in terms of taking some of the decision-making on the investment side. There are so many variables that people are looking at to be aggregated into the decision-making so the more you can use a machine, the better it is.“  Pierre Lagrange, Man Group

"With AI, advisors will be able to focus on what really matters to clients. AI will change everything and advisors who embrace it will win. I would guess that many advisors are woefully out of date with its technology because they don't understand and are scared of it.“  Larry Miles, Principal at AdvicePeriod

"A couple of years ago when I was at a Talk in Vancouver, I heard Jeremy Howard speak on the future of Artificial Intelligence and how computers were already on the cusp of surpassing human performance in many tasks. Shortly thereafter, I met a fund that was modelled after Graham and Dodd's value investing philosophy but was able to analyze every public filing, such as 10Qs and 10Ks since inception. This further confirmed my belief that this was the future of investing.” Jeffrey Tarrant, CEO and Founder of MOV37 and CEO, CIO at Protégé Partners

“AI is creating a lot of new opportunities. Just as about 100 years ago electrification changed every single major industry, I think we’re in the phase where AI will change pretty much every major industry.”   Andrew Ng, Chief Scientist at Chinese internet giant Baidu Inc


What is Bad AI?

•Predictive models driven by AI can and do get things wrong.

•If there are biases or problems with the data being used to build the predictive models, then the models will be flawed.  

•For example, if you are trying to make an inference about the general population, but the data only comes from social media users, then the predictions and results may be flawed because the entire population does not use social media.

•If the people who build the model make incorrect predictions about how it will be used, then the model will be flawed.  

•Over-fitting could also be a problem.  This occurs when an algorithm searches too far to find correlations and patterns in the data.  There is randomness in all data and if it is over analyzed there may be incorrect inferences.


Ethical and Philosophical Considerations

•Is there a threat to human dignity if AI replaces people in positions that require respect and care, such as nurses, police, and teachers?  Will there be a diminished value of the human spirit if we think of ourselves as merely biological computers?

•Societal versus Statistical Considerations.  Few would argue that race, age, and gender should not be considered when making medical diagnoses, but should they be considered for decisions related to offering someone a job or renting them an apartment? 

•Is AI potentially a threat to privacy?  Can AI speech recognition technologies and the ability to read and understand every email and text in a population effectively allow governments to suppress dissent? 

•Will AI displace jobs?  Will AI take away more jobs than it creates?

•Are intelligent machines potentially dangerous?  Can we ensure that machines behave ethically and are used ethically?  Could AI be designed or guaranteed to be benevolent?  Because AI has no need to share our human motivational tendencies, could the designers program what we know as morality?

•Is there an existential risk?  Could the AI take off on its own and redesign itself at an ever increasing rate?  Could it evolve to the point that humans cannot control it?   Will Artificial Intelligence Weapons used for military combat develop to ability to make their own intelligent decisions of who and when to kill?

•AI is benefitting society in the medical field.  What other positive contributions is it making?


Best AI Movies

•Blade Runner (1982) - The Blade Runner has been tasked with terminating four replicants, android beings that look exactly like humans, who have escaped from an off-world colony.

•Westworld (1973) - This is a story about amusement park robots that malfunction and begin killing visitors.

•Robocop (1987) - In future Detroit, a murdered police officer is resurrected as an experimental crime-fighting cyborg named Robocop. He soon deviates from his law enforcement duties to seek revenge on his killers. 

•I, Robot (2004) - This movie addresses the three laws of robotics as prescribed by the great sci-fi writer Isaac Asimov into its story.  First, a robot may not injure a human being or, through inaction, allow a human being to come to harm.  Second, a robot must obey the orders given it by human beings except where such orders would conflict with the First Law.  Third, a robot must protect its own existence as long as such protection does not conflict with the First or Second Laws

•Terminator 2 (1991) - After the failed attempt at killing Sarah Connor in The Terminator, the Skynet sends a more advanced robot, the T-1000 back in time to kill John Connor. But another terminator is also sent back from the future to protect Sarah and John Connor.

•The Matrix (1999) - The Artificial Intelligence depicted in ‘The Matrix’ is arguably the most advanced of all the movies on this list. The machines are powerful enough to create the Matrix to imprison human minds, while they feast on the body heat of the humans.

•2001: A Space Odyssey (1968) - The computer Hal-9000, aboard the S.S. Discovery is often called the sixth passenger despite being just a computer due to its similarity to humans in thinking and decision-making. Hal is even shown defeating a human at chess which was deemed impossible at the time the movie was made.



As we have seen, Artificial Intelligence, Machine Learning, and Big Data are not as mysterious they seem, are not (yet) a threat to humanity, and can enhance almost any type of investment style.  Investment managers who embrace them will outperform, while those who are in denial and refuse to adapt will be left behind.