Understanding the risks of AI
A report released by RIMS, 'Making sense of artificial intelligence and its impact on risk management' broke down AI into two broad types - artificial general intelligence (AGI) and artificial narrow intelligence (ANI).
AGI is a 'thinking machine' that applies intelligence to a wide range of cognitive functions and continue to improve their reasoning abilities automatically. For many of us, "the best examples are the well-known machines from science fiction classics, such as HAL in 2001: A Space Odyssey, R2D2 and C3PO in Star Wars, or SkyNet in The Terminator," said the report. AGI represents tremendous opportunities, as well as risks, but the likelihood of it becoming a reality within the next 20 years is slim.
The second is the type of AI is focused on narrower tasks, such as image recognition, credit card fraud detection and speech recognition. "ANI should be a major concern for risk professionals as it has impacted many aspects of everyday life, and will continue to have growing, and potentially unexpected, impacts," said the report. "The market for narrow solutions is much bigger: Gartner estimates that AI-derived business value will be worth $3.9tn by 2022."
Common implementation scenarios of ANI include advanced process implementation, cognitive engagement such as intelligent agents, chatbots or virtual assistants, using machine to aid in data analysis such as in deep learning algorithms.
The exposures that risk managers must consider, the report said, include:
• Today's ANI innovation is tomorrow's standard product feature that consumers
• Changes introduced by AI innovations can
• Product and service differentiation is not necessarily driven by a computational algorithm. Rather, it is data-driven. Risk professionals should consider impacts to data related to risk domains such as security, privacy and resiliency.
• Distribution systems that rely on third-party developers, partner/agent sales models or other customer contact points can extend your risk beyond the traditional corporate boundaries. AI developed, operated or otherwise influenced by third parties working on your behalf needs to be consistent with your organization's core values.
"Essentially, if you understand the organization's strategy and how it can enhance its operations with ANI or the context around data, then you have something to offer," the report said.