I was recently asked a question by my mentor saying, “there’s a lot of hype on what AI/ML/Deep learning can do and will do. I have two simple questions.
- What are these computing paradigms good at and not so good at?
- What next?
This question was not an uncommon one by my mentor. I have a few friends, who are CXOs and board members of many large corporations who said, that these are the real questions that every CXO is asking and within closed-door do not really have any real clarity of what next? Why in spite of billions of Pounds being invested in technology and computing, banks fail, investing in stock markets plummet and assets implode across all classes?
My initial response was the new “intelligence” is really “Heuristic and Evolutionary Intelligence” and went on to demonstrate a few but very clear slides. The common responses were to say this is very exciting and potentially could redefine the next work not only in computing but also across business and governance.
I started work in the area of heuristics a long while ago and developed on it looking at two key areas of overlapping spaces with incredible and exciting applications in computing;
- Heuristics & Heuristic Intelligence
- Evolutionary Computing & Intelligence
Why Heuristics and not a typical NP-Hard problem.
Why 40 Trillion (40 x 10^12) simulations to find 1 or 2 trading desks with AI, when heuristics could solve that challenge better, example Index investing? Classic case of an NP-Hard problem. Common to all on Wall Street companies and for that matter every organization from Fortune 500 to startups to governments. does that scale? According to SPIVA over 75% to 90%+ underperform their capital compared to the index!!! The BFSI is not alone. Or Markowitz and his Nobel Prize-winning work on Modern Portfolio Theory who in the end used a Naive 1/n heuristic for his own investments. As Gerd Gigerenzer said it would take 500 years of data to prove the whole model!
The goal is in a few basic slides demonstrate what we believe is our fundamental view on AI and what we believe will be the next big area of unbundling value across ecosystems.
I worked in AI, ANN and expert systems when in my teens as an undergrad on my own passion and time (over 80%) on AI and robotics school and computing models for applications that were looking at the aviation and aerospace industry, MSDF (Multi-Sensor Data Fusion), dynamic supplychains, etc. Those were early days and one has not seen any major work as most of the math and statistical theory has evolved but still static in many ways. What has changed are the laws of the telecosm making computing, network storage and bandwidth a utility like electricity.
A simple heuristic, Index Investing by John Bogle is now a $4.9 Trillion AUM company called Vanguard Group and the fastest growing asset class of investing. Does not need complexity for complexity sake.
A simple heuristic saved hundreds of lives, not backtesting and complex computing in an uncertain environment. Would you depend on a simple heuristic or complex AI/ML models that in the movie suggested that they could have landed back at the airport.
So the key is VUCA (Volatility, Uncertainty, Complexity and Ambiguity). This is where past modelling do not and consistently prove that these models, while good at repetitive tasks are poor are evolving scenarios. So having made the case for why we need a dynamic and constantly evolving model of computing based on what is next or maybe, we need to look at why Evolutionary Computation will potentially be the game changer.
Gerd Gigerenzer at the Max Planck Center for Human Development is a brilliant mind and leader in the space and a future Nobel Prize-winner (in my mind for work done). We will be publishing our work on Heuristics & Evolutionary Computing that we will share publically as most of our work is for internal consumption only. Love to connect and hear your thoughts.