Meet Amelia, a machine that thinks like you

Amelia on Laptop (handout image provided by Feature Photo Service) Feature Photo Service

Here's what it's like to have a conversation with Amelia, the nearest thing yet to a real-life version of Samantha, the artificial-intelligence operating system in the movie "Her."

"Where does Christopher Mims work?" types Ergun Ekici, the lead architect of Amelia, into her (for now) text-based interface. And then, well, there's really no other way to put this: Amelia responds like a person. She's read my bio online. She knows all about me. But her real talent isn't regurgitating information; it's solving problems.

Here's a typical conversation with Amelia, this one in another area of her expertise, diagnosing car trouble.

Joe: "Hello, I'm stuck here and my car won't start."

Amelia: "I'm sorry to hear that. Could you take a look at your dashboard? Is the battery light on?"

Joe: "No."

Amelia: "OK. Are any of the lights in your car on?"

Joe: "No."

Amelia: "It could be an issue with your battery. Do you have jumper cables with you?"

On its own, this dialogue isn't remarkable. It could be scripted, though it's not. And Amelia can talk like this on pretty much any topic once she's trained. She gets questions right nine times out of 10, says Jinho Choi, who took a sabbatical from Emory University to help build her brain, more than any other system that has been made public.

Natural-language conversations with computers aren't new. In 2013, Google unveiled "conversational search," which lets you talk to the search engine, engaging in short dialogues about topics present in its "knowledge graph" of people, places and things. Thus, it's possible to ask, in spoken English, who is the U.S. president, and to follow up with questions like "who is his wife?" and "when was she born?" without having to reference the original subject of the conversation.

And IBM's Watson, famous for winning "Jeopardy," excels not only at answering individual questions about facts in its database but also the probability that the answer is correct.

Amelia, though—incubated for the past decade at a privately owned, relatively obscure IT services firm called IPSoft—is different. She learns from textbooks, transcriptions of conversations, email chains and just about any other text. As long as the answer appears in the data she gets, she can solve problems.

IPSoft's main product is another bit of artificial intelligence, a software suite known for automatically resolving IT infrastructure issues known as IPcenter. A Gartner analyst wrote that IPSoft is a "stealthy newcomer," competing with the likes of IBM. IPSoft was founded in 1998 and has offices in nine countries.

Amelia is already being tested—in some cases, alongside Watson—by companies in surprisingly diverse industries, from telecommunications to energy. She embodies a new approach to artificial intelligence called cognitive computing. Its defining characteristic is machines that can learn. Yet because of the complexity of their understanding , the knowledge they contain can't be programmed into them. Like all software, these systems are first built by programmers, but like children, they must be taught to do the things for which they are intended.

The fundamentals of cognition

Amelia is the product of an attempt to understand how people think, rather than to copy the means by which we do it. Many traditional AI efforts try to map the human brain, or the brains of less complicated animals, like fruit flies. But Amelia is all about turning what psychologists and linguists know about how thinking happens—a high-level understanding of it, rather than how it's carried out by our neurons—into software.

"We didn't achieve powered flight by copying birds," says Chetan Dube, president of IPSoft. "First, we had to understand the principles of flight." And so it is with cognition.

One potential application of Amelia is in the call center. Fortune 100 companies are already testing Amelia in this role. The goal is consistency—every time anyone calls, that person should get the same, correct answer. And the reason Amelia knows the correct answer is that she is ingesting every single support request the company receives and learning from the answers humans dispense to customers.

Another way to put it: Soon, the ultimate repository of technical knowledge in a company could be the cognitive computer at its heart.

Amelia is the product of a consortium's worth of partnerships with academics in the fields of language processing and artificial intelligence.

IPSoft is privately held and profitable, giving it the luxury of pursuing the interests of its founders, many of whom started as computer scientists working in universities.

Amelia is still in the pilot phase, and whether or not she has real utility has yet to be publicly proved or disproved. But she is already having hundreds, and soon thousands, of simultaneous conversations with IPSoft's customers.

"The near future application is, we're trying to replace humans with this—especially in a customer support type of situation," says Kazu Gomi, CEO of NTT America, which is testing Amelia.

Automated emotions

Despite her abilities, the one thing Amelia is not is sentient or alive. She remains, like all synthetic intelligences, merely a clever way to automate tasks. As an IPSoft spokesperson put it, Amelia "has no free will."

She does, however, have feelings. If you tell Amelia "I hate you," the three variables that define her emotional state— arousal, dominance and pleasure—are negatively affected. "Our goal here is not to just model emotions but to use what we detect of those sentiments in decision making," says Mr. Ekici.

We've yet to create an operating system that anyone might fall in love with, like the humans in "Her." But it can't be long now.

Follow Christopher Mims on Twitter @Mims and write him at christopher.mims@wsj.com.