AI Gets Very Real

Artificial intelligence (AI) is advancing at a furious rate. Personal assistants such as Siri, Cortana and Google Now have moved into the mainstream. They allow us to navigate tasks on our smartphones and in our automobiles—in many cases saving dozens of clicks or taps

At the same time, robots that harness AI are rolling into our lives. The Aloft Cupertino and the Aloft Silicon Valley hotels in the United States and the Henn-na Hotel in Japan have begun using robots. The Henn-na will use “actroids” as receptionists and helpers who will tuck guests into bed at night. They will also handle traditional hospitality tasks.

Yet, one of the most appealing uses of AI—at least for businesses, government and other organizations that provide sales and support—is for virtual agents. Although some Websites now offer basic capabilities, this technology is advancing rapidly.

Consider this: IPsoft has just launched Amelia 2.0, a tool that is very human-like and is able to display near human cognitive capabilities. The company claims this makes her “difficult to distinguish from a human help agent.”

The firm uses cognitive computing capabilities to take the virtual agent where it has never gone before. Designers have tapped episodic memory and semantic memory to duplicate the way the human mind works. IPsoft also claims that it has built in a high level of contextual comprehension and emotional responsiveness.

Amelia aims to completely automate customer support. The company believes that the system brings virtual agents one step closer to passing a Turing test, which requires that a human cannot distinguish between a machine and another human being.

Let’s face it, most human support ranks somewhere between mediocre and pathetic. There is simply too much information and too many complexities for any one person to manage. It would be great if the company could fully deliver on the promise of AI and virtual agents.

But my sense is that tools such as Amelia may deliver an improvement 90 percent of the time, but prove woefully inadequate for the other 10 percent of situations. There are always exceptions that require human intervention. Putting an entire support system on autopilot is a train wreck waiting to happen.

We’re still about 10 years away from a virtual system that can function exactly like a human being. Even then, there will still be the 1 percent of cases that require the help of a person.

Truth be told, if we ever reach 100 percent, then there probably will be no need for humans.