Is it real or is it hype? Any fast-emerging, highly touted technology attracts that kind of skepticism. Artificial intelligence, as hot and hyped as anything in today’s tech marketplace, is no exception. But neither does AI fit neatly into any pre-existing patterns or assumptions about IT hype and reality.
There are indications, both anecdotally and in market research, that AI is enjoying a rush of commercial interest, fueling expectations that it will be game-changing, groundbreaking, even revolutionary in its impact on business and other aspects of society. To be sure, AI in its various forms – including machine learning, natural language processing and robotic process automation – is still, in terms of business applications, early-stage.
Also, perhaps paradoxically, the technology is not exactly new. As a science, AI is well over a half-century old, gaining early notoriety through popular culture (2001: A Space Odyssey) and for winning chess games. However, as computing power has improved in the era of big data, the excitement now is the potential of AI to disrupt nearly every aspect of our lives, from automated vehicles to cancer treatment.
Those high expectations do not seem unrealistically distorted. That is one takeaway from a recently conducted survey by the Global Association of Risk Professionals (GARP) and analytics leader SAS: four out of five respondents (81%) said that AI technologies are already benefiting their institutions.
The GARP/SAS survey, conducted online in December 2018, drew more than 2,000 total responses from across the financial services industry, including banking, investment banking/securities and wealth/asset management. The most common departmental functions were risk (48%), finance (14%), and IT (9%); and executive levels included a healthy mix of director-level and above titles (28%), team leader/senior manager/manager (36%) and analyst (31%).
For the survey, AI included machine learning, natural language processing, computer vision, forecasting and optimization. In terms of current usage of AI technologies, forecasting (54%) and optimization (51%) were followed by machine learning (34%), robotic process automation (29%), natural language processing (23%), computer vision (23%) and virtual agents (22%).
Sizable percentages of current non-users said they are planning to use the technology. That would, for example, bring the machine learning penetration up to 80%, and natural language processing 66%.
There are of course plenty of practical and operational challenges, ranging from the need for a basic familiarity with these systems, to finding the necessary technical talent, to managing the quality of big-data inputs and being able to understand and explain how AI models produce their outputs.
As one survey participant summarized: “In my field of work [model risk], AI is currently viewed as something unavoidable, and with many potential benefits but also many potential issues regarding model understandability, good modeling practices, etc.”