Samsung Dives Into the Uncanny Valley With Neon ‘Artificial Humans’
LAS VEGAS – There have been a lot of debates over how humanoid robots should be designed, and at what point their resemblance to humans becomes a problem. Samsung’s NEON subsidiary (part of the company’s STAR Labs) has taken a deep dive into the same territory by launching a family of 2-D “artificial humans.” The company has made some amazingly bold and possibly scary claims about its plans for the life-like avatars, which they began to reveal at CES this week.
NEON Wants to Make Max Headroom a Reality
For those who aren’t familiar, in the TV series of the same name, Max Headroom is the digital embodiment of his human alter ego, Edison Carter, in the computer realm. He is clever, snarky, a trickster, and a friend and guardian angel for Carter. For full disclosure, I think he is one of the most brilliant creations ever. Now Neon, a company backed by tech giant Samsung, claims that it can create digital beings, initially based on actual humans, that can be “friends, collaborators, and companions.”
Neon’s creations will be able to “understand, converse, and sympathize, just like a real human.” On the surface of it, these seem like pretty outlandish claims. But the company showed some impressively realistic video avatars doing everyday tasks like reading or chatting, on life-size displays at CES this week. So it is worth diving a bit deeper into their underlying technology.
CORE R3: Reality, Realtime, and Responsive
The lower level of the Neons’ software stack is driven by its Core R3 system that models the physical actions and responses of a human in real-time. That means the Neon bots can interact in a fairly natural manner physically. Even in explaining this, Neon waxes poetic: “[Core R3] is inspired by the rhythmic complexities of nature and extensively trained with how humans look, behave, and interact.”
Unfortunately, the demos I was able to see weren’t interactive, and also came with a caveat that they might actually just be based on humans. So while the bots were apparently performing everyday tasks in a lifelike manner, in the limited audience demos, interactions didn’t seem to go much beyond a typical Alexa interaction.
SPECTRA: Intelligence, Learning, Emotion, and Memory for bots
R3 is just a teaser for what Neon promises for the future. Layered on top will be Spectra — a technology that from the company’s description will let a bot not just ace the Turing test, but allow it to serve as a replacement for a companion or colleague. Specifically, Neon says its “Neons” will be able to “understand, converse and sympathize just like a real human.” I’m almost sure I’ve seen that movie, and it doesn’t always end well.
Reality Check: What Neon Really Means
Personally, I think the company’s positioning of its bots is off-base. If they can actually create a “new species of artificial humans” that can be my “friend and companion,” and lead us to a world where “humans are humans” and “machines are humane,” I’m willing to see how it works out. But I’m speculating it’s a fantasy for the foreseeable future. What the company’s bots will almost certainly be able to do within a year or two is serve as a cool replacement for kiosks, digital signage, and other partially effective solutions to helping us live our lives and find our way around.
For example, I can easily see one greet me at an airport (privacy issues aside, maybe it recognizes my face), and point out where I should go to check in. It could then answer questions I might have about where to clear security, check baggage, or find a place to take a nap. Same for hotels, amusement parks, and insanely large trade shows like CES. One could easily be a more believable version of Alexa when I get home in the evening. But replace my human colleagues on wfoojjaec or the friends with whom I play tennis? I don’t think so.