Recently deceased scientist, Stephen Hawking, said to a graduating class, “The next century [21^st] will be the century of complexity” and encouraged them to embrace it.  Yes, we should not fear it, but embrace it by gaining knowledge of a new science, Artificial Intelligence (AI).  It began at Dartmouth University in 1956 when John McCarthy used the term (AI) as the title of a conference. To help us operate successfully in an increasingly complex world, we need to learn the language of complexity.

Why have so many experts who forecasted a market top in the past few years been wrong? Perhaps it is because they have been using 20^th century methodologies in a 21^st century world of increasing complexity.  This is no longer like us playing a game of chess against a few experts and a lot of amateurs, as existed in the 50’s and 60’s.  We are playing against computer programs that have learned to reason from massive amounts of data derived from a very complex system, and are making better decisions than us about everything from driving cars to investing in the stock market.  That means we need to know something about complex systems and the computer programs that understand them better than we do.

The evidence:

The 1^st shot across the bow was when IBM’s Deep Blue beat the world champion chess player, Gary Kasperov, because it knew all the games he and every other expert had ever played and could access them in a flash.  That was the earliest form of AI  which involved 3 steps:

1. Training the model on a massive amount of data shoveled into computers run in parallel on processing platforms.
2. Testing the model to see if it learned how to win in the past.
3. Validating the model to see if it can uncover relationships from the data to predict future outcomes, then, learn from and adjust to future events better than human experts.

Another breakthrough came in March of 2016.  The Google program, named AlphaGo, was taught to play the game of Go that has 10 to the 127^th power possible moves.  AlphaGo scored a decisive win over Lee Sedol, a top-ranked international Go player, winning 4 out of a 5-game series in South Korea.  Then, in 2017 it won the first two games of Go in a three game championship game with the world’s champion, Ke Jie. The Google AI program appeared to think and reason better than human experts. Below is an example of a supervised learning approach which requires a massive amount of historic data to teach the model how to achieve the goal, for example, beat the S&P 500.  This involves some of the language you need to be familiar with.

Neural Network (Supervised Learning)

The X_i are data sets affecting the thing you are trying to forecast, like, turning points in the stock market.  The W_i are the initial weights assigned by experts based on historical results. For example, X₁ could be the 3 year moving average return on the S&P 500.  X₂ could be the 3 year moving average standard deviation.  X₃ could be the upside potential ratio calculated from the first two X_i with the aid of the Forsey-Sortino model.

The tree search in AlphaGo then evaluates positions using statistical methods like Logistic Regression and Random Forests to adjust the weights. The model then makes a forecast and learns from the results how the weights should be adjusted further. This is called reinforcement learning.

The Break Through (Unsupervised Learning)

In 2018, Google updated the AlphaGo model to AlphaGo Zero, which uses an algorithm based solely on reinforcement learning, without human data, guidance, or domain knowledge beyond game rules . This is known as Unsupervised Learning. AlphaGo Zero became a neural network trained to predict AlphaGo’s own moves without requiring massive amounts of historic data. It learned as things changed in real time. How did it do? In 4 days the Zero version learned how to beat the original AlphaGo model in 100 out of 100 games.

What might somebody using an AI model see that you and I don’t see?

Similar to the AlphaGo Zero model mentioned above, AlgoDynamix utilizes an unsupervised machine learning algorithm to process limit buy and sell orders from financial exchanges to identify patterns as shown above. It seems there are certain patterns of buyers and sellers that the AI model has identified in clusters which have predictive power, the rest are noise.  To be sure, I know very little about this company or the methodology they are using.  However, the result shown below leads me to want to know more.

For more information visit www.algodynamix.com. The few graphs in this paper were selected from an introductory course in AI at Stanford University and do not begin to explain the revolution akin to the industrial revolution that is taking place.  I for one intend to find out more about the research being done in AI so I can comfortably surrender my 20^th century driving skills to my new Tesla Model 3 when it rolls off the assembly line.  I encourage you to go to your nearest university to take some introductory classes.  You may want to start by downloading:  Kaplan, Jerry. “Artificial Intelligence: What Everyone Needs to Know”. Oxford University Press. Kindle Edition.

 Conclusion:

It is estimated that AI will have replaced 13 million U.S. workers by the end of 2018 and 1 billion workers worldwide by 2022. Most vulnerable: Truck & Uber drivers, retail workers and… investment advisors ?

“Last year, On March 28, the CEO of BlackRock sacked seven fund managers and shifted billions of dollars to an AI product called Systematic Active Equities (SAE). SAE now employs 80 portfolio managers and researchers, who include more than 30 PhDs in computer science, physics and engineering. This is one of the greatest opportunities in active management that I have ever seen, said Ron Kahn, head of research at SAE.”

Is BlackRock going to need RIA’s to sell their product? Are RIA’s going to need firms like BlackRock to keep their clients? How do they do that if they don’t understand what BlackRock is doing vis-a- vis AlgoDynamix?

The information technology revolution is not coming, it’s here! The winners will be shareholders in high tech stocks, college grads with science degrees and Businesses that adapt to 21^st century technology.