IPsoft's 'Amelia' virtual assistant just got a whole lot more human

Designed to help companies automate customer support, Amelia on Wednesday graduated to version 2.0, bringing the technology another step closer to passing the Turing test, maker IPsoft said.

The Turing test requires that a human user must be unable to distinguish the machine from another human being.

Amelia’s physical appearance and expressiveness have been transformed to create a more human-like avatar for deeper customer engagement, IPsoft said.

The artificial intelligence platform has also gained maturity and core understanding capabilities, thereby broadening the range of roles it can assume. Included in version 2.0 are new advances in comprehension and emotional engagement through improvements to memory, contextual comprehension and emotional responsiveness.

Amelia’s memory now more closely mirrors the way human memory is organized, for example, allowing the technology to hold more natural conversations, IPsoft said.

With contextual comprehension, meanwhile, Amelia can quickly and reliably retrieve information across a wider and more complex set of knowledge, IPsoft said.

Finally, Amelia now has richer mood and personality vectors, enabling her to personalize the customer service she provides.

Numerous architectural changes are included in Amelia 2.0 as well. For example, the entire backbone has been rewritten so that the technology can scale seamlessly and remain resilient throughout extreme peaks in volume.

Amelia 2.0 is scheduled to be demonstrated for the first time this week at the Gartner Symposium/ITxpo in Orlando.

Amelia 2.0 signals "the next step in the evolution of what's now being called the cognitive category of tech and services," said Frances Karamouzis, a vice president with Gartner. "It's becoming harder and harder to discern when we're communicating with a machine -- when insights are coming from an algorithm in something like Amelia."

Amelia is currently set apart by the fact that she speaks 20 languages and has an emotional quotient, Karamouzis said. Also new is that the technology can interpret the user's facial expressions as conveyed by a camera and generate appropriate ones in return, she said.

The biggest improvement in version 2.0, though, is what's called semantic understanding, Karamouzis said.

"It's starting to not just follow a process but understand from a semantic level what is your goal -- for example, if you're trying to open a new account or close one," she explained. "It's creating a cognitive system that's more goal-oriented than process-specific."

Tools like Amelia are becoming cognitive-like, Karamouzis said, but they still can't pass a Turing test.

"A lot of these tools are very close," she said. "They don't incorporate all the elements of emotion yet, but they're the closest you'll get out of the market."