Sexy animated chatbot


06-Sep-2017 04:53

“If it were an actor, Baby X is just improvising, whereas [a digital human] is being told: I want you to walk here, I want you to say this, and I want you to deliver it with this tone.” Crucially, AVA and other digital humans are incapable of feeling anger.

They are built to be bottomless wells of empathy, no matter how nasty a customer gets.

Before founding Soul Machines in 2016, engineer Mark Sagar was a movie special effects guru at outfits like Sony Pictures and Peter Jackson’s Weta Digital.

He won two Oscars, in 20, for his work creating lifelike CGI facial animation in films including Jackson’s .

Soul Machines’ imperfect beauty has to be more than skin deep for expressions to be believable, though, with computer modeling of bone structure, muscle twitches, and other subtleties. Since leaving cinema, he’s been trying to recreate the human nervous system in software.

Analyzing facial expressions to discern a smile, even a subtle one, and analyzing voice to pick up a pleasant tone, Soul Machines’ software provides a hit of virtual dopamine to AVA’s nervous system.

Don’t be mistaken, though: This is very far from artificial sentience, Sagar cautions.

Built on top of IBM’s chatbot software Watson Conversation API, AVA can currently handle about 500 types of requests, identified from running machine learning on recordings of live help calls to identify common problems customers have.

Autodesk is also experimenting with Watson Tone Analyzer to get a read on a customer’s mood, but Rachael Rekart, who manages Autodesk’s “machine assisted service engagement,” hopes for more.

The Autodesk deal is Soul Machines’ first major gig, following a pilot project with the Australia National Disability Insurance Agency from February to September, 2016, and some proof-of-concept demos, like a recent one with Air New Zealand.

Sexy animated chatbot-89

carbon dating decay rate

Greg Cross, Soul Machines’ chief business officer, says the company is working on eight other projects with “major brand names” but declined to name them.

Please help improve my work by reporting bugs and improvements.