Understanding Artificial Intelligence And It's Impact On Our Collective Future

Every day I dream. I dream of building intelligent systems that can sense, think, and feel. I dream of building machines and systems that can interact with our physical world, sense patterns, and make decisions. More than that, I dream of understanding consciousness, what it means for a machine to be conscious and what the building blocks of consciousness truly are. To many, I am sure that these dreams sound like the subject of fairy tales or fantasy movies. And yet, as I survey the world and continue to work on this dream every day I become more convinced of the importance of this work; work, grounded in the field of Artificial Intelligence (AI).

What is Artificial Intelligence?

A quick survey of research returns many different perspectives on what AI actually entails. As noted in Wikipedia, Dr. John McCarthy, whom many consider the father of AI, defined AI as the "science and engineering of making intelligent machines". While the definition seems to make sense, a closer look reveals more questions than answers. For example, accepting this definition requires one to understand what "intelligence" is. One must also understand what the presence of intelligence, particularly within a machine, looks like? Is it a passing grade on the Turing test? The ability to quickly recall specific information (something computers can do exceptionally well)? Some other measure? While this question alone deserves follow up in a future post, I will accept this initial definition and lead on from here.

If AI focuses on making intelligent machines, then I suppose that progress in this area would be predicated around the solving of a small subset of centrally unsolved problems. While there are many fundamental problem areas that I am sure researchers would point to, the most fundamental area in my mind relates to learning. To be clear, if we are to unpack and unlock the potential that AI offers, then we must deeply understand learning. In particular, we must stop and understand what it means to learn and answer the basic question: "How does a human learn?" To solve this, we must dig inside the human brain far beyond our current understanding. We must understand how humans store information and how the brain works at a microscopic level. We must also understand memory - both short term and long term - while also understanding how memory systems and networks are developed. The pursuit of these questions leads one naturally into the area of machine learning, an area that I covered in my first Masters of Science in Statistics at Simon Fraser University. While there is sure to be disagreement, this is one area that will get significant attention during my PhD work supported by the incredible Dr. Vicki Lemieux at the University of British Columbia iSchool. Research in this area is also highly relevant to the ongoing daily work that I lead at Cymax. In this work, we use intelligent systems to compete against competitors in many areas, developing machine based intelligent competencies that continue to strengthen over time.

In addition to learning, there are also many other areas that need review. For example, in order to truly understand how a machine learns, we must understand how a human, and thus a machine, senses its external environment. Similarly, items like attention, perception, reasoning, and decision making - areas central to marketing and business - are also critically important. Every day at Cymax for example, when reviewing paid media spending or media assortment planning, my teams think deeply about how consumers perceive various cues, the messages that they respond to, the state of the customer when interacting with our digital and physical properties, as well as the position of advertisements on the page or device. These challenges are a small subset of the same challenges that many of our partners, including Google, Microsoft, Facebook, and others, spend significant time and resources - billions of dollars in fact - developing "intelligent" systems.

What Are The Foundations Of AI Research?

While the paragraphs above describe some of the central problems that the AI community is facing, progress undoubtedly requires practitioners with a unique set of skills in order to tackle the problems that are outstanding. While it has taken me many years to develop differing levels of expertise in the areas I consider foundational to AI (with a lifetime of continuous learning and research ahead), the items noted here provide a high level overview of the main foundational skills that I believe are prerequisites for success.

In order to build intelligent machines, I believe that one must understand robotics, algorithms, cloud computing, statistics and mathematics. One must understanding optimization, calculus and linear / non-linear algebra. Additionally, one must understand neurology, biology and psychology. In addition to the above, one needs to also understand information search, information management & large scale information processing. One needs to understand software architecture, high performance system architecture & design, data modelling and database systems. Moreover, to build live systems, one needs to understand and use various programming languages. While there is always mixed consensus, most of my programming work is focused on Python, Java & R. Finally, one needs to understand the world and have a conceptual understanding of various environments in which AI systems may be most successful. For example, understanding the domain of finance will be critical for those looking to change financial systems using AI. Understanding health will be critical to those looking to do bioinformatics. Indeed, I was privileged to have a physician in last years M.Sc Health Bioinformatics class at Georgia Tech where we built a small emergency medical management application for hospitals that had intelligent capabilities.

The Opportunities That Artificial Intelligent Systems Create

As I look forward and think about the impact of future AI research, there is no question that many opportunities exist. Yes, I recognize the risks that people have raised regarding the dark side of AI systems. That said, in my mind, I strongly believe that the opportunities far outweigh the potential risks. Here are a few of the opportunities that immediately come to mind. 

Intelligent AI systems have the potential to transform agriculture, support the intelligent optimization of food production, while providing food to populations with significant food shortages. Furthermore, intelligent AI systems have the ability to impact  waste management, reduce waste, and monitor resource supplies for contamination. AI also has the potential to help improve our energy use and distribution of energy while also reducing our environmental footprint. Within the home, AI research can impact a family in many ways, including supporting family duties such as the purchasing of food and acquisition of other resources. AI systems have the ability to support learning for school, with intelligent tutors like the one that was active in my previous class at Georgia Tech. Such systems have the potential to change who and how students are educated, greatly increasing the pool of skilled labor.

In addition to this, AI research has the potential to lead to the development of systems that support organizational decision making in much more productive ways, taking us far beyond the current conceptions of what "predictive" forecasting means in businesses.  Furthermore, AI has the ability, and is already enabling, the development of intelligent transportation systems including self driving cars. During my M.Sc at Georgia Tech I had the pleasure of taking a class taught by Dr. Sebastian Thrun and experience first hand his passion for building intelligent transportation systems that can reduce accidents, lower risk, make the world safer (while also giving people time back in their lives during the commute).

Security is another domain that I believe will benefit from future research. AI systems have the ability to create systems that can secure our property and protect both lives and physical assets. I absolutely recognize the counter argument and risk that the production of intelligent weapons systems may bring. That said, with so many people facing unsafe conditions globally, the question of security aided by AI agents is one that I believe cannot be ignored. 

Finally, while there are many other areas, two critical opportunities relate to health and historical information processing. Indeed, within the medical field, the ability to diagnose problems more effectively using AI systems is something that will change the manner in which healthcare is delivered globally. Such acts require systems to synthesize tremendous volumes of data, including historical data that is processed using large scale text analysis. This challenge of large scale text analysis is one that I have found persisting within the archival world. This is why, despite many curious questions from researchers and other professional colleagues, I decided to plant my research within the School of Library, Archival & Information Studies at the University of British Columbia focusing on AI & machine classification. After all, what better a place to deal with the challenge of building giant scale systems than within a research program focused on the handling and mining of large volumes of archival data. Such a program, as well as the larger body of research work that is going on within the AI field, has the potential to make a monumental impact on our collective future in ways that stretch outside the realm of our current imagination!