Financial Perspectives
“There are no secrets to success. It is the result of preparation, hard work, and learning from failure.”
– Colin Powell
Friend or Foe?
By Ray Ryan, CFA
Academic recognition of artificial intelligence (AI) dates back to a 1956 conference at Dartmouth College where researchers demonstrated computer programs that could play checkers, solve algebraic word problems, prove theorems, and speak English. Prior to that event, Alan Turing’s “Theory of Computation” claimed machines could “learn” mathematics through the simulation of formal reasoning. Ethical debates regarding “thinking” machines date back even further to philosophers and futurists who challenged not whether such technology was possible, but instead, whether those innovations should be pursued.
Early development in the 1960s and 1970s focused on computational capacity. Gradually, research shifted to teach machines human cognition – i.e., perception and pattern recognition. By the 1990s, enhanced by neural networks, researchers discovered applications in economics, mathematics, and statistics. However, widespread adoption remained limited because there was a dearth of specific problems to solve. There was even a novelty aspect to artificial intelligence as developers demonstrated capability against human Grand Masters in chess matches.
Faster and more powerful computers, access to large databases, and the efficiencies of cloud computing led to advances in machine learning and deep learning methods. Those advances, in turn, contributed to a surge in artificial intelligence deployments since the turn of the 21st century. In fact, Google greatly increased artificial intelligence projects after 2012. Today, numerous studies indicate more than 20% of companies have incorporated artificial intelligence in some manner, and the pace of adoption is accelerating.
There are various forms of artificial intelligence, and we encounter them daily. Some AI programs merely identify patterns, and AI platforms share patterns with other applications. Others classify data and then, according to the classifications, apply rules. For example, email “spam” filters rely on artificial intelligence algorithms that scan messages for keywords or symbols. The program then directs the message to the appropriate folder (i.e., “Inbox” or “Junk”). The dreaded “auto-fill” function of text apps utilizes artificial intelligence to complete messages. The pleasant voice on the end of a helpline is often artificial intelligence programmed to guide callers through a menu of options. Voice recognition AI, such as Alexa and Siri, improves through greater interaction, which is generally the case with artificial intelligence.