Korean web large Naver is exploring bots, synthetic intelligence and autonomous driving
Picture credit: by no means
In July of final yr, by no meansThe search engine producer often known as Google South Korea has opened its new second headquarters. Naver 1784. The constructing’s identify refers to a yr normally related to the Industrial Revolution. Inside you will discover a testomony to the rising international affect of automation, within the type of 100 autonomous robots.
Robots, known as rookies, You possibly can transfer between individuals, ship packages, espresso and lunch to Naver staff, all with 5G. Additionally they have their very own robot-only elevator, the Roboport. Naver Labs, the corporate’s analysis and growth division, launched utilizing the brand new headquarters as a check mattress for learners. And in November this yr, it’ll prepare the bots at its new information middle, Gak Sejong, in South Korea.
Picture credit: by no means
I visited the positioning final yr to see the robots firsthand. Rookies transfer via each inch of the headquarters, whereas… Robotic arms painted photos. Naver already owns a big group of subsidiaries, providing companies comparable to e-commerce, messaging, funds, digital comics, cloud, and Metaverse. And simply final week, the corporate Launched generative AI companies Constructing on its hyperclova X language mannequin, it joins the house AI race.
Does the web large need to broaden robotics, synthetic intelligence and self-driving expertise past its present enterprise to drive its future progress? When Naver was requested about her ongoing tasks, she stated that she is engaged on them A bunch of tasks has centered on robotics, self-driving expertise and synthetic intelligence, via its R&D division since 2013. Six years in the past, it was Acquires Xerox Analysis Middle Europe (XRCE) in Grenoble, France, which makes a speciality of analysis in synthetic intelligence, machine studying, laptop imaginative and prescient and pure language processing, to boost its analysis and growth. After the acquisition, it grew to become XRCE Nefer Labs.
The corporate believes that individuals who reside in such revolutionary environments will profit from Naver Labs tasks – together with robotics – within the not too distant future. The web search engine needs to participate in a brand new sort of platform, bots, which it believes may someday substitute cellphones.
Naver Labs’ analysis targets 5 areas: robotics (Round, Round G and Ambidex), synthetic intelligence (ARC system), digital twins (each high-resolution and hybrid mapping), laptop imaginative and prescient/AR (AR indoor navigation and ahead navigation) and autonomous driving . Driving expertise.
On the Naver Synthetic Intelligence Convention final week, TechCrunch sat down with Naver Labs Europe (NLE) Science Director Martin Homburger to find out about Naver Labs’ mission and ongoing tasks.
“Our long-term imaginative and prescient is to create robots which might be helpful in on a regular basis life,” Hommenberger tells TechCrunch. “This contains assistive robots within the house, hospitals, museums, malls, airports, and so forth. Within the quick time period, we’re specializing in last-mile supply functions in good cities and buildings.”
Most of its long-term tasks haven’t but been commercialized. Nonetheless, its ongoing analysis and the most recent applied sciences are always being utilized to enhance the efficiency of Naver’s robots, Rookies, in Constructing 1784 and the Gak Information Middle. As well as, it gives Robotic as a Service (RaaS) to By no means Cloud platformHomenberger provides.
One instance of RaaS is ark system Powered by Naver cloud platform. Arch, an intelligence system for robots, allows Naver learners to see, acknowledge and act. ARC isn’t just a management system for robots. It is the eyes and thoughts of the robots, which implies learners do not want entry to a GPS; As an alternative of that, Arc eye It helps them get to know their environment, whereas offering probably the most environment friendly route. That is in keeping with the chief of the Imaginative and prescient Group at Naver Labs Korea, Donghwan Lee, who leads the ARC Eye undertaking. In the meantime, ARC Mind connects the service infrastructure in real-time and updates information and algorithms within the cloud for robots. Lee added that Constructing 1784’s starters do not want a built-in working system due to their cloud connection that powers the ARC system that controls the robots. Naver says the ARC system can scale back robotic manufacturing prices and battery consumption by shifting the duties the robots must deal with to the cloud.
Picture credit: by no means
Homenberger says each Naver Labs in South Korea and France have some clear widespread targets, however there are others working in parallel. Collaboration depends on data and sharing of progress till widespread targets are recognized.
whereas Naver Laboratories Europe (NLE) focuses its analysis on “the basics of synthetic intelligence and find out how to overcome present and future challenges in synthetic intelligence,” and Naver Labs Korea (NLK) researches and develops robotics, Methods engineering And Platform structure He provides that operations.
The first analysis space of NLE is AI expertise for robotics. NLE is creating “next-generation laptop imaginative and prescient strategies for robotic notion,” which assist robots higher perceive the atmosphere, and “new strategies for sequential decision-making and combinatorial optimization, which allow robots to navigate unknown environments or carry out advanced processing duties.” Homenberger explains. NLE can also be constructing expertise that permits efficient interplay between bots and people by way of language.
Picture credit: Naver Labs/ Martin Humenberger, Director of Science, Naver Labs Europe
To make Naver bots versatile assistants, NLE is creating AI-powered laptop imaginative and prescient, machine studying for bots, and pure language processes.
“It’s true that we rely quite a bit on AI-powered laptop imaginative and prescient to raised perceive the world (semantics and geometry), reinforcement studying to show robots to carry out duties, and pure language processing and knowledge retrieval to make data (textual content, photos, and so forth.) accessible to our robots.” ,” Humenberger tells TechCrunch.
offers an instance of a concrete analysis exercise associated to real-world fundamental fashions, croco. CroCo is mainly a 3D mannequin in a single picture. “It’s designed to grasp the 3D world utilizing photos,” Homenberger stated, including that it’s not supposed for generative AI or giant language fashions however for robots and digital twins. “At Naver Labs, we use CroCo as a complicated robotic cognition unit for duties comparable to localization, navigation, and human-robot interplay.”
“We’ve got an enormous testbed the place we are able to develop robots and function them in the true world, extremely expert engineers and researchers, and a strong cloud infrastructure,” Homenberger explains. “This can be very tough to safe even considered one of these circumstances, and it’s protected to say that Naver is the one firm outfitted with all of them.”
NLE has greater than 100 researchers and scientists in areas of synthetic intelligence, starting from fundamental machine studying to consumer expertise, in keeping with Homenberger. Naver Labs didn’t disclose the variety of staff in Korea on account of confidentiality points.
South Korean expertise giants, comparable to Samsung, LG, Hyundai and Doosan, are within the technique of creating robots, from industrial robots to cooking, service and supply robots, and Injecting capital into the sector of robotics. earlier this yr, Samsung stated It plans to launch an EX1 robotic to assist the aged train, and its analysis division is creating Samsung’s robotic platform. Doosan Robotics, the robotics unit of Korean producer Doosan, has filed for an preliminary public providing, in keeping with its submitting.
Need to get high robotics information in your inbox each week? Join as an Actuator right here.
