Samsung subsidiary Star Lab showcases "Artificial Human" project.

2020-01-08

Samsung envisions a world filled with artificial intelligence assistants, but not the kind of chatbot assistants we are already familiar with. At a recent press conference, the Seoul-based company unveiled its subsidiary STAR Labs' Neon project, which aims to provide "immersive... services" and make science fiction a reality. Pranav Mistry, a human-computer interaction researcher and former Senior Vice President at Samsung, explained that the core R3 software engine behind Neon is designed to create lifelike avatars for movies, augmented reality experiences, web, and mobile applications. In a tweet, he wrote, "(They can) autonomously create new expressions, new movements, and new dialogues... completely different from the (few milliseconds delayed) raw captured data." Neon's virtual avatars resemble videos rather than computer-generated characters. This design is primarily intended to make them "companions and friends" to humans, serving as concierges or receptionists in places like hotels and stores. They are designed with strict privacy protection measures, meaning they cannot share personal data without permission. Unlike typical AI assistants, they are not capable of performing general tasks. The lab explicitly stated in a FAQ shared with journalists that Neon does not tell you the weather or play a song. They only engage in conversations with people to help accomplish "goal-oriented" tasks or assist with slightly more complex tasks that require a "human touch." Mistry stated that businesses will be able to authorize or subscribe to "Neon services" when the beta version is released later this year. The second significant component of Neon, which is responsible for the intelligence, learning, emotions, and memory of the avatars, is still under development and may make its debut at a conference later this year. Neon brings to mind the previous Milo project, a typical "emotional AI" developed by Lionhead Studios in 2009. Milo was an AI construct that could respond to voice, gestures, and predefined actions, and it had a procedural generation system that continuously updated an internal dictionary by matching words in the conversation with voice clips. Unfortunately, Milo never made it into the public eye.

RECENT AI NEWS

RECENT AI TOOLS