Stated very simply, AI is the name given to computer systems that attempt to replicate human intelligence and learning. There's great difficulty in finding a satisfying definition for AI because the definition of intelligence itself conjures big questions of consciousness and being that have not yet been resolved by science and philosophy. A practical way to approach AI is to consider why it exists in its current manifestations.
AI is typically sorted into two groups — strong (broad) AI and weak (narrow) AI.
Strong AI is a closer representation of what most people understand true intelligence to be — that is, a reservoir of utility and knowledge that can be drawn upon and adapted to handle whatever task is at hand. Strong AI, however, does not yet exist, so when people talk about AI today, they're usually talking about weak AI applications.
Weak AI systems draw from algorithms that learn, but whose capabilities do not extend beyond a narrowly defined scope. These are often computers that play chess or try to predict future crimes, for instance.
Today's weak AI systems are used for two overlapping reasons — to complete tasks that people are capable of, but that are tedious or time-consuming, or to synthesize data and produce conclusions from that data with a precision and clarity that the human mind is not designed to comprehend. In short, today’s AI fills the computational gaps in human ability, while humans maintain status as nature's safe keeper of executive function. People create AI systems not as separate entities, but for the purpose of extending their own intelligence and capabilities.
Artificial intelligence goes by many names, and among them are game theory, machine learning, deep learning, reinforcement learning, neural networks, fuzzy logic, data analytics and data mining. In most cases, these terms each have their own specific meanings and purpose, but they generally all relate to the same field of study; and when pressed, most of even the foremost experts are unable to articulate definitions that separate one term from the next. The term “AI” can be rightfully applied to any system that inhabits those corridors typically reserved for human wisdom and experience.
A more thorough understanding of AI can be gathered from examination of government's use of such systems. Humans are poor at recognizing certain kinds of patterns in discrete data, and computers are exceptionally good at this, so any task that requires pattern recognition tends to be a good fit for an AI application. Researchers at the University of Chicago’s Computation Institute, for example, are training AI systems to help the city of Chicago find which houses are most likely to contain lead pipes and harm the inhabitants.
Government holds out hope for AI research to solve a broad range of problems, from social issues and public safety, to more efficient management of back office functions. Firm IPSoft developed a "cognitive agent" named Amelia to help governments get the most from their limited budgets. AI can be used to automate entry-level tech jobs and run help desks, explained Bob Beck, general manager of IPsoft’s government solutions.
"We believe wide-scale adoption of AI by government will provide a great deal of support in a number of different areas," Beck said. "I think first and foremost is budgets. There's a lot of angst these days with using funds appropriately and using them wisely, and doing more with less. Using AI gives the government the ability to use dollars they have to do more."
AI fits closely with today's trend of data-driven decision-making. After all, using data to make sound decisions has another more common name: thinking.
