Samsung's Neon explained: Just what exactly are these 'artificial humans'? – CNET

neon-keynote-ces-2020-020

Neon CEO Pranav Mistry introduces the company’s new AI to an audience at CES 2020.


Sarah Tew/CNET

This story is part of CES 2020, our complete coverage of the showroom floor and the hottest new tech gadgets around.

While other companies at CES 2020 looked to the future of artificial intelligence, Samsung brought a bit of futuristic AI to the present when it revealed Neon, an AI chatbot emerging from the company’s Star Labs advanced research division. But chatbots have been around for well over a decade now, so you may be curious what the hubbub surrounding Neon is all about.

CNET’s Shara Tibken had questions, too, but the answers she got from Neon CEO Pranav Mistry left her more confused than ever. And when Andrew Gebhart got the chance to interact with a Neon, the experience left him impressed more with the possibilities than the reality.


Now playing:
Watch this:

Neon, the lifelike AI from Samsung, answers my questions…



6:22

Even still, the buzz around Neon is palpable. To help introduce you to Neon’s new “artificial humans,” as the company calls them, here’s everything we know about the Neon project, as well as a few lingering questions we have about these next-generation AIs.

Everything we’re told Neons are not

Like any good philosophical paradox, the best way to approach the question of what a Neon is is to start with what they are not. According to the company, a Neon is not meant to replace or improve upon the technology used in the current generation of digital assistants, like Apple’s Siri, Amazon’s Alexa or the Google Assistant.

A Neon is not, in other words, meant to fetch answers to simple questions about the weather or sports scores. Nor is it there to control your smart home devices, set reminders and alarms or play your favorite music.

According to the company, Neons are also not “androids, surrogates or copies of real humans,” although they may exhibit similar physical or behavioral traits to actual people. They aren’t meant to inhabit robots, either.

neon-ah-01

The first wave of Neons are modeled after real people, but Neon CEO Pranav Mistry says future Neons can be modeled to look like — but not identical to — actual humans.


Neon

What Neons are (rather, what they could be)

According to Neon CEO Pranav Mistry, “Neons are more like us, an independent but virtual living being, who can show emotions and learn from experiences.” He goes on to explain that Neons are meant to converse and behave like humans, as well as remember and learn. They’re like a related — yet different — new species of life from humans.

Neons are meant to act as “friends, collaborators and companions,” according to the company. “They can serve as an individualized teacher, a personal financial advisor, a healthcare provider, … a concierge … an actor, a spokesperson or a TV anchor.”

What Neons might be able to do for you

Although we’re curious about the practical uses for Neons, Mistry sees the AI being used in a variety of specialized applications, tailored to individual needs. Say you wanted to learn yoga. You could potentially learn from a Neon yoga expert that could show you various poses and, much like a human teacher, incrementally increase the level of difficulty as you master new skills.

neon-keynote-ces-2020-057

Reporters had a chance to interact with Neons at the company’s CES 2020 booth. 


Sarah New/CNET

If you wanted to become fluent in a new language, a Neon could teach it to you, as well as translate in real-time when you need help.

A Neon could also serve as something akin to a friend or a confidant. All language processing is said to occur on-device, so Neons will reportedly keep your secrets, even better than a human person, one presumes.

Neons can trace their lineage back to Samsung

Neon is the brainchild of longtime Samsung researcher Pranav Mistry. The project emerged out of Star Labs (aka Samsung Technology and Advanced Research Labs), and is funded by Samsung, but it’s not actually a Samsung company. Mistry told Tibken that Neon isn’t meant to replace Samsung’s digital assistant Bixby, it won’t be installed in Samsung products and that the company operates independently of Samsung. 

013-big-booths-of-ces-2020-central-hall

Neon emerged from the Samsung Technology and Advanced Research Labs (aka Star Labs) division. It’s funded by Samsung, but operates independently from the technology giant.


Sarah Tew/CNET

But having the backing of Samsung also means that Mistry doesn’t have to have a clear business model yet for Neons, he said. Ultimately, Mistry envisions Neon having other investors alongside Samsung.

Where (and how) Neons will exist

Mistry told CNET that Neon doesn’t intend to turn the avatars into physical robots, but he speculates that eventually they may exist as holographs. In the meantime, expect to interact with Neons using the same technology you use to connect with friends and family — on smartphones, tablets, PCs and even smart TVs, similar to using apps like Skype and FaceTime.

When you do, Neons should appear lifelike and life-sized, at a similar scale to the people you already video-chat with. The first phase of Neons are based on real human likenesses, but Mistry said future Neons can be modeled to look like — but not identical to — real people. The Neons will still look like humans, but not any actual, living people.

No, you can’t build your own bot

Unlike in games such as The Sims (or even Apple’s Memoji feature on iPhones ($699 at Apple) and iPads ($330 at Walmart)), you won’t be able to choose how your Neon looks. They will arrive on your screens fully developed, and you won’t be able to change them. The implication is that the company wants to keep the focus more on the interactions you have with Neons, and less on superficial aspects like appearance. 

As Mistry told CNET, “When you meet a friend, you build that friendship, not that you build that friend. In the same way, businesses who hire Neons can’t decide what they look like.” They even come with their own names, like Hana, Natasha, Jordan and Johnny.


Now playing:
Watch this:

Going paws-on with a robotic cat that listens to your…



5:09

You won’t own your bot, either

Neons likely will be sold under a subscription model, meaning you can essentially rent but not own your Neon. Businesses will be able to “hire” Neons for specialized tasks, like translation, but will be unable to license the core technology that powers Neons. It’s unclear, however, whether you can have a Neon with you all the time or if they only show up for, say, watching Netflix for a few hours on Fridays or teaching you Mandarin on Tuesdays. 

Speaking of core technology, here’s the techy stuff

NEON AI at CES 2020

An ‘artificial human’ demonstrates Neon’s Core R3 graphics engine at CES 2020.


Sarah Tew/CNET

There are two fundamental technologies driving Neons. First is Core R3, which stands for “reality, real-time and responsiveness.” Think of Core R3 as the graphics engine that powers Neons’ natural movements, expressions and speech.

Then there’s Spectra, which will drive the AI’s “spectrum of intelligence, learning, emotions and memory.” It’s what gives a Neon its mind, heart and soul, if you will. Spectra is not quite ready for prime time just yet, but the company says it will preview the technology sometime later this year.

Feelings can be hurt, hearts can be broken

Another way the company says Neons are different from Alexa, Siri and the rest is that Neons are not programmed to remain passive in the face of indignity. If you’re mean or nasty to a Neon, you run the risk of upsetting it, or even making the AI angry at you. Presumably, if you’re pleasant and polite to your Neon, it will respond in kind.

ces-2020-privacy-panel-5954

This CES panel included Facebook’s privacy chief Erin Egan (second to left) and Apple’s privacy chief Jane Horvath (middle).


Angela Lang / CNET

Expect a higher degree of privacy than you’re used to

Privacy concerns around voice-interface AI are growing, and it’s with that in mind that the company behind Neon claims that privacy is at the core of the AI’s design. All language processing occurs on-device, the company says, and all records of your interactions with Neon are kept private. The company stresses that it will never share your private information without your permission. 

Like a real, human friend, the memories of interactions are tied to one specific Neon. If you interact with a new Neon, it won’t be able to pick up where you left off with the previous one. You’re starting over, as if it’s a new person you’re meeting.  

Neons still have quite a ways to go

If you’re noticing that Neon leans heavily on potential and much less on present-day performance, you’re not wrong. The company’s CES presentation was meant to demo the Core R3 technology that powers the Neons’ movements and interactions, but the Spectra engine driving “intelligence, learning, emotions and memory” is still under development.

Some remaining curiosities

  • Mistry said Neon will see a beta release later this year, but when might it be ready to deploy at scale?
  • The Neons on display at CES 2020 appear as both male and female avatars, but are they meant to be gendered? Are their preferred pronouns “he” or “she” or “it” (or something else entirely)? Will there be transgender Neons, too?
  • How much will a subscription cost? Will some Neons, possibly those with more specialized skills or knowledge, cost more than others?
  • Chatbots have a sordid history. A notable recent example was Microsoft’s Twitterbot, named Tay, which turned into a sex-crazed, Hitler-praising racist. It’s unclear what — if any — failsafes Neon will have to prevent similar corruption.
source: cnet.com