January 15 news, according tomedia reports, Samsung’s Star Labs lab in 2020 CES brought the latest virtual human project Neon, caused a stir in the industry. But the demo project doesn’t integrate the artificial intelligence brain, and the functionality is a bit thin. There’s a lot of work to be done for the Neon team in the future.
Here’s the translation:
CES has been a wash-out event, and the big news at this year’s show is the Neon project from Samsung’s Star Labs Labs. Samsung began promoting its “artificial man” long before it took part in the Las Vegas show.
Self-hyped speculation is often dangerous and often requires special courage to make a commitment and demonstrate your own skills in a live demonstration. Pranav Mistry, president and chief executive of Star Labs Labs, stood in front of a group of “virtual people” to introduce the Neon project to reporters and the audience. In Mistrier’s words, Neon is a virtual assistant who looks, speaks and behaves like a human being. But after watching a concept video about the Neon project, it can be seen that The Neon-like chatbot is still a failed pantomime.
During the presentation, Mistry chats with a female virtual person on the screen, asks her questions, and uses the app to transform various facial expressions. Most of these virtual people look real and can answer some questions successfully, but the problem is that the presentation is tightly controlled by the team. Even in this case, Neon seems clumsy. In the demo, questions can only be asked by Neon team members, and the virtual person’s performance is also stiff. It can be said that this kind of virtual person is not useful at all.
While the entire demo was conducted “on-site” and Neon team members randomly generated pre-rendered unique questions with computers, these were not the artificial intelligence that Samsung touted in the demo. So far, there is little evidence that the company has been able to achieve its vision.
The long-term goal of the Neon project is to make virtual people act and react like real people. Unlike voice assistants such as Alexa, Neon is not designed as an inexhaustible source of knowledge, but is used for specific tasks such as concierges, guides and guides, or as a digital companion for users. Want to learn guitar? Maybe a virtual person can teach you.
Mistriri has been chief executive of Star Labs Labs since October, leading the project. He was also involved in the development of Ballie Robot and Samsung Robotics Chef Bot Chef. Prior to that, Mistry was vice president of research for the development of the first smartwatch, Galaxy Gear. Mistry said Neon’s idea had been in the works for two years, but the project had only begun six months ago. “This is a preview of the technology, not even a beta version. It’s far from ready to roll out,” he admits. In a tweet on CES, Mistry also called Neon “a young start-up project for just four months.”
The avatar in the demo is based on the body and face of a real person. The Neon project is based on a real-life avatar and then overlays an artificial intelligence layer on it to generate “millions” of possible animations, gestures, and sound feedback.
The Neon project has two main components. The first, which represents “real-time response to reality” Core R3, is the rendering engine behind Neon’s virtual human actions and expressions. At the moment all virtual people are based on real people, but Neon has only extracted the basic image from it, and all the rest of the work is done by Core R3 technology. “Virtual people’s physical characteristics come from real looks, but behavioral characteristics are not necessarily theirs, ” says senior technology writer Angie Chian, introducing the Neon project. “We can give them a new posture. So the way the lady laughs is not necessarily the way she laughs in real life. I really don’t know how she laughs. “
Neon’s ultimate goal is to be able to build virtual people from scratch without imitating any real people. Mistry says they should eventually be able to do just that.
Third parties will be able to insert their own characteristic information into Neon to provide a “brain”. For example, a hotel can create a digital receptionist by ordering a Neon project and then using Neon technology to provide specific information for its business. Mistry even believes that one day we will see Neon used as a news anchor to broadcast big news. However, he insists that “the significance of the Neon project is not to replace humanity”.
The second component yet to be released for the Neon project is Spectra. This software engine will become a virtual person’s memory, able to learn about users from interactions, and become smarter over time. That’s where Neon needs to deliver on its promises, but it’s also where Samsung, and many companies, have suffered setbacks.
Neon’s team points out that virtual people don’t currently use Samsung’s existing technology, so it’s not based on Samsung’s Bixby voice assistant. Mr Mistry said Samsung had not joined Neon’s road map, at least for now. It’s easy to recall Samsung’s failed attempts at artificial intelligence. This year, Google Assistant and Bixby will be integrated into Samsung TVs this year with Alexa and Amazon Alexa, both of which are more powerful and popular than Bixby, which could be seen as Samsung’s admission of its failure. In addition, the release of Samsung’s first smart speaker appears to have been delayed indefinitely.
Viv is another case of failure. The artificial intelligence platform Samsung acquired in 2016 was eventually integrated into Bixby 2.0. Viv was a compelling product at its debut, more successful than Neon, but to no avail. It’s tempting to wonder if Samsung should go after a project like Neon’.
“Spectra will give users the kind of experience they want to have on the phone, ” says Mr Mistry. He was optimistic about this, but did not provide details. “Spectra will give the virtual person the ability to know you, I know your preferences, I know you’re tired, I know you don’t like this movie, so I won’t talk about it again.” “
Neon plans to test with some partners by the end of the year, but it won’t be fully available for some time. In fact, the presentation on CES was done on a powerful computer. When the Neon project is finally online, it may be moved to the cloud for processing.
Samsung’s biggest mistake was to show the Neon project prematurely without spectra components. For now Spectra looks like a well-designed chatbot. If there was a brain, Neon might be more special and useful. Especially at a time when Amazon is working to make Alexa’s voice response more human, Neon seems to think it makes more sense to be able to make users see an interactive avatar.
Even in the customer service market, Neon seems to be coming late. Soul Machines, a start-up focused on providing digital avatars for customer service, recently raised $40m, and companies such as Procter and Gamble, Royal Bank of Scotland and Google have had some success in deploying robots to customer service. Soul Machines’ system also has the upper hand, as it uses IBM’s Watson Artificial Intelligence system for some voice processing.
It’s too early to say what will happen, either if Neon may never be able to get out of the lab or be sold to someone else. “If you look at the gaming industry, it’s been trying to make characters look more real for the last 15 years. If you go to the film industry, the same thing will happen,” Says Mistrey. “People want this. “
Misterley also admits there is still a long way to go before Neon really goes live. Neon virtual people do look like real people, and the potential is huge. But the real test is yet to come, and it remains to be seen whether Neon can combine virtual human technology with artificial intelligence.