To date, there is no crisp, universal definition of artificial intelligence. John McCarthy, the Stanford University computer scientist who coined the term in 1956, defined it as, "the science and engineering of making intelligent machines." Subsequent definitions have been similarly broad, but the idea is that artificial intelligence involves the creation of machines that think and behave like people along a spectrum of capability, where intelligence is defined in relation to human intelligence. Its capability spectrum ranges from aiding mathematical decisioning with calculators to augmenting medical diagnoses with tools that parse inordinate volumes of data to identify in minutes what might otherwise take doctors years to discover.
I'll refer to artificial intelligence at the higher end of the spectrum as cognitive computing. Cognitive computing is IBM Watson's approach to AI, and it's being leveraged broadly throughout industries today. Given that mankind is a long way from understanding the mechanics of consciousness, self-aware machines are outside the scope of this article. They remain very much in the realm of science fiction.
Cognitive computing represents a new form of computing in which machines are no longer programmed. Instead, they are trained by non-technical subject matter experts, like lawyers. Cognitive computing represents a benign partnership between man and machine where, as we'll see, man remains very much in control.
Cognitive systems are noteworthy because they mine unstructured data for insight. Unstructured data is data that people create for other people to consume, like narrative text (think judicial orders, briefs, expert reports, contracts, etc.). To this day, most computers can only analyze structured data, or information stored in relational databases (think keywords, fields, metadata, numbers, dates, etc.). Importantly, 80 percent of all data is unstructured.
It's been impossible for computers to parse unstructured data in any meaningful way until very recently. As a reflection of human language, unstructured data is full of nuance. Houses burn up as they burn down. We fill in forms to fill them out. In order to mine language for insight at scale, cognitive systems perform three essential functions: They understand, they reason and they learn.
In understanding, cognitive systems comprehend human language, discerning intent, tone and personality. In reasoning, they infer and extract concepts, forming hypotheses...