The world's richest man started by developing gutter oil

Because JD.com has not disclosed anything about this technology, ordinary people have no idea about it. Some people who understand technology think it is not simple, and some technical groups have started talking about JD.com's technology.

"How do you think JD.com achieves this technology? Does it identify data by searching for fragments of data?"

"Use data segment search technology to make an image recognition technology for me. The angle of the user's photo is fixed. The photos taken by the user and the photos of the product are completely different. This technology is probably modified by Deep Blue Group using facial recognition technology."

"Deep Blue must have used some special algorithm to make it, but I think the algorithm alone is not enough. It will not work if the algorithm alone cannot keep up with the computing power. Deep Blue must have developed a dedicated acceleration hardware. Unfortunately, Deep Blue has not disclosed any technology to the public."

"If you were in this situation, would you share your core technology?"

"Technology is borderless, okay? Deep Blue just wants to monopolize it."

"The person above has probably been brainwashed by foreigners. He has forgotten about Deep Blue's chips being cut off. If technology really has no borders, why would they cut off Deep Blue's chips?"

"Please ask the newbie above to learn about the Wassenaar Arrangement. Technology without borders is just a scam for ordinary people."

"I still remember that Deep Blue's disk array cabinet in the United States was stolen a few months ago. Can you guess why these thieves stole Deep Blue's disk array cabinet? If Deep Blue's security technology was not the best in the world, it is estimated that some of Deep Blue's technologies would have been stolen by foreigners."

"I suspect this technology is AI technology. Ordinary technology can't do it. Our company has also studied image recognition, but our technology is far inferior to JD.com. Our technology can only recognize frontal photos, and it can't recognize photos regardless of angle like JD.com does."

As JD.com’s biggest competitor, Ali.

After learning about this matter, Lao Ma immediately called the head of the technical department to his office for a meeting.

"Yongming, did you watch the press conference held by JD.com just now?" asked Lao Ma.

"I have seen it, Mr. Ma," Wu Yongming replied.

"I need you to develop a similar function." said Lao Ma.

"Mr. Ma, this thing is more complicated than we thought. I suspect it is an artificial intelligence technology, not an ordinary search technology at all. Ordinary search technology can't do it at all. In addition to the algorithm, it also needs hardware support. Even if we invest a lot of money in research, it will take at least 3 to 5 years to develop results." Wu Yongming said.

Lao Ma was a little uneasy when he heard Wu Yongming's words. He didn't think that it would take three to five years to develop a technology. However, the reality was more difficult than they imagined.

Not to mention whether they have the ability to develop AI, even if they can develop the image recognition algorithm, they do not have enough computing power to support it unless they can buy an AI acceleration card from Deep Blue.

But it is impossible for Deep Blue to sell the AI ​​accelerator card to them if Deep Blue is willing to make the AI ​​accelerator card public.

Then Deep Blue wouldn't have kept it so secret and not released any information.

"No matter it takes three or five years, you have to research it out for me. If you don't have enough people, you can ask the human resources department to recruit externally. We must develop this technology in the shortest time possible." Lao Ma said.

"Okay, Mr. Ma." Wu Yongming said with a bitter expression.

Lao Ma gave him a death order. Although he was also a shareholder of Ali, he had no way to refuse Lao Ma's order unless he completely withdrew from Ali.

Deep Blue Industrial Park Deep Blue Group.

Just now, Han Chen received an email from Andrew Ng. Their AI department needs to recruit more mathematics experts. If they only recruit through DeepBlue Direct Hire, there is no way to meet their current needs.

Andrew Ng hopes to discover more mathematics experts by hosting a global mathematics competition.

After reading Andrew Ng's email, Han Chen picked up his phone and called Tong Jingjing.

Twenty minutes later.

Tong Jingjing pushed open the door of Han Chen's office and sat down opposite Han Chen.

"Mr. Han, do you have any arrangements with me?" Tong Jingjing asked.

"As the company grows, it needs more and more talents. It is very difficult to recruit some talents alone, so I plan to develop a talent reserve plan. This talent reserve plan has two modules. The first module is to recruit talents through a competition mode." Han Chen briefly explained his idea.

Han Chen's competition model is somewhat similar to the Ali Global Mathematics Competition in later generations, except that the Ali Global Mathematics Competition only competes in mathematics, while Han Chen plans to add subjects such as physics, biology, materials, and radio.

"For each competition, the company will give out RMB as prize money. As for how to distribute the prize money, you can decide. The second module is the talent school module. A private university will be established in the name of Deep Blue Group. This university will only recruit the elite of the elite as students. Anyone who graduates from the school and obtains internship certification in the group can receive a job offer with an annual salary of one million yuan from the group after graduation.

Our school does not charge any tuition fees. All expenses are covered by the Deep Blue Group. Admitted students can also receive a scholarship of 50 yuan.

In terms of teaching staff, please invite the best teachers to come back and teach for me. We can offer international-level salaries, the same as what Ivy League schools offer.

Our school offers all the majors required by the group. Each professional class only accepts 100 students each year. We do not admit any students who are not Chinese nationals. The assessment of students must not only assess their professional knowledge, but also their emotional intelligence and imagination.

As for the application and construction of the school, I will ask the secretariat to assist you." Han Chen said.

"Mr. Han, I will prepare the plan within a week and set up the entire operation team within a month." Tong Jingjing turned off her phone and returned to the meeting function.

In the past, Tong Jingjing would have taken a pen to write down every word Han Chen said. However, the Deep Blue MAX3 has real-time recording and text conversion functions, so she only needs to take out her phone and turn on the meeting mode.

"Okay, then go back and do your thing." Han Chen said. "Okay, Mr. Han." Tong Jingjing replied.

Conference Room No. 2, Building No. 4, Phase of Shenlan Group.

Yan Huiwu, the person in charge of Pangu system R&D, called all R&D team leaders and management of Pangu system to the conference room for a meeting.

"I believe everyone is curious about why I called you here. Just now, President Han gave me a new task. This task is a new challenge for us, and it is a very severe challenge." Yan Huiwu said and opened the PPT sent to him by Han Chen.

The first page of the PPT had the words "Spatial Computing System" written on it in large letters.

"That's right. What President Han wants us to develop is the spatial computing system. So what exactly is a spatial computing system? I believe everyone here has seen Iron Man. When Iron Man used holographic projection to design his mecha, the technology he used contained the spatial computing system.

It's just that the screenwriter who wrote Iron Man didn't know what to call this thing. Our task is to realize the space computing system. In the words of President Han, Hollywood is responsible for imagination and we are responsible for realization. Do you have the confidence to make the space computing system?" Yan Huiwu asked the crowd in the audience.

"We have confidence." The audience responded in unison.

For these scientific researchers, the more challenging and interesting it is, and this is how they can fully display their talents.

"Boss Han has already created the entire spatial computing framework. To be honest, I was a little skeptical at first, until I saw the PPT sent by Boss Han. Then I realized that our boss is so amazing. Everyone here considers themselves geniuses, but they are still far behind our boss." Yan Huiwu exclaimed.

Everyone was surprised when they heard Yan Huiwu say that Han Chen had already created the framework of the space system.

"What we have to do now is to fill the content into the framework. I am also very confused about spatial computing. If it weren't for the framework designed by Mr. Han, I wouldn't know how to develop it well." Yan Huiwu added.

In fact, spatial computing is nothing new, and some people present have heard of the concept of spatial computing.

The term was first introduced in a paper by Simon Greenwold of MIT in 2003.

However, the hardware level at that time was very backward, and even if someone studied it, they would not be able to make it. Therefore, only a very small number of people knew about this thing, and only a very small number of people were researching this project.

“Spatial computing is a computing model that integrates virtual reality (VR), augmented reality (AR), mixed reality (MR) and other technologies, aiming to merge digital information with the real world.

This fusion creates a new computing environment that can perceive and understand the space around us and interweave digital content with real-world scenarios. Spatial computing has a wide range of applications, such as geographic information systems (GIS), intelligent transportation systems, urban planning, and other fields.

Spatial computing originally refers to the technology of calculating and analyzing maps and other geographic location data to achieve positioning and measurement. With the development of technical fields such as XR, virtual humans, and digital twins, the computing demand for microscopic space is also gradually increasing.

Other more natural input methods such as voice, vision, gestures, etc. will further enrich traditional interaction modes such as mouse, keyboard, and touch screen, allowing people to access and interact in the way that best suits their current scenarios and business processes.

From the perspective of core technology, spatial computing is supported by a series of technologies including AI technology, 3D reconstruction, spatial perception, user perception, and spatial data management. It is the key technology for building the metaverse space and achieving free switching and mutual integration with the real world.

From the perspective of computing carriers, spatial computing can be divided into edge computing, cloud computing, and cloud, edge, and edge collaborative computing with 5G as an important medium. Through spatial computing, people, objects, machines, and virtual space can be seamlessly connected, digital twins can be built, a new economic form of virtual-real integration can be created, and ultimately a new round of industrial transformation can be driven.

Spatial computing reflects how we interact with objects, people, animals, and targets in the real world. Humans convert 2D images from our eyes into a 3D model of the world, understand objects in the world, and then direct our hands to act.

For example, when we pour a cup of tea, we look at the cup as we pour it, determine when the cup is full, and then stop when the cup is full. The same is true for spatial computing, but using sensors, computers, and actuators.

Spatial computing involves multiple steps.

First, technologies such as photogrammetry, lidar, and radar can capture 3D models of the world. Lidar or radar capture 3D models by measuring the reflection of laser or radio signals from objects around the scanner to automatically capture a point cloud representing the distance to each point.

Photogrammetry is described as the art and science of creating 3D models from photographs by combining images from multiple images or cameras, and newer AI techniques such as Neural Radiance Fields (NeRF) can capture richer representations using just a small number of images.

Second, technologies such as machine vision analyze this data to understand the images, and AI techniques help identify individual objects in a scene, find defects, understand gait patterns, or analyze how different workers perform a process.

In construction, for example, these technologies can help monitor work progress, pinpoint issues such as doors not closing properly, and identify problems such as contractors forgetting to install electrical outlets.

The third aspect of spatial computing involves taking action. For example: a self-driving car detects a pedestrian ahead and decides to stop in real time, a building control system decides to adjust the heat or light when someone walks into a room based on their preferences stored in a database, and a construction management system schedules a team to install a missing plug.

The operations in these examples are possible due to the analysis of digital 3D images captured from the physical process," Yan Huiwu explained to the audience according to the content on the PPT.

At this moment, everyone present was listening with great interest and absorbing the knowledge in it crazily.

In fact, Sora, which is being developed by the AI ​​department, also involves spatial computing. Many people think that Sora is just a video generation AI. Sora is a diffusion model that inherits the ability to understand text. The so-called diffusion model follows the logic of reversely restoring a noisy image to the original target image.

Open AI fully leverages the advantages of its large language model, allowing video data to be used directly to train the model.

It is on this basis that Sora is able to understand the various physical laws of the real world, and then convert them into video works based on the prompt words, creating an effect that is realistic enough to be mistaken for real. This ability to understand the world is not possessed by other AI video generation models, and is also an important reason for the difference in their video quality.

Sora achieves spatial computing by understanding the various physical laws of the real world.

It has been mentioned before that Sora can also be applied to autonomous driving. It is not a very simple task to let an AI that can understand various physical laws of the real world to interpret autonomous driving.

It's just that Sora and XR glasses are not used in the same field, but they can be mutually beneficial. (End of this chapter)