What's wrong with me being a rich man?

Chapter 453 Insights
Yu Xing was astonished by the performance of Yu Kai, a senior executive at Baidu.

Are you still using Baidu?
No, no, you've decided to start your own business, haven't you?

Yu Xing dealt with the sudden surge of enthusiasm, his mind racing. Seeing the unusual interest the man had in LiDAR, he called Li Yifan of Linggang Company and asked him to come and meet the vice president of Baidu Deep Academy IDL.

Although he is called the vice president, the position of president, Li Yanhong, is clearly symbolic. Therefore, he is the de facto president of IDL. However, even as a de facto president, such a department faces considerable constraints in mobilizing resources within Baidu.

When Li Yifan arrived, he saw Yu Kai marveling at the prototype vehicle.

Yu Xing's introduction was only halfway through when he saw the person in charge arrive, so he smoothly wrapped it up: "President Li said they are confident that they can make a product with performance close to the Velodyne 64-line radar, and can also reduce a lot of costs. If Baidu needs it, we can cooperate at that time."

Yu Kai already knew that the machine being loaded onto the truck was Linggang's 32-line prototype. Upon hearing this, he asked, "Is Mr. Li already working on a 64-line machine in his lab?"

Li Yifan shook his head and replied, "No, we are working on a 40-line radar."

Yu Kaiqi said, "40 lines? Everyone's using 64 lines now."

Google, Baidu, and some similar projects in the United States are all using Velodyne 64-line radar, while its predecessor was the 32-line radar on the "Jiuzhou" prototype vehicle in front of us. If Linggang Company wants to develop the lidar business, creating a 40-line radar will undoubtedly increase the compatibility work for customers and directly increase the difficulty of sales.

Li Yifan explained, “We conducted a test comparison and found that the current 32 and 64 line distribution is not reasonable enough. The 40 line is more balanced. When we used Velodyne 64 line to scan a 1.7-meter-tall pedestrian 50 meters away, we could only get a maximum of 5 lines, but the 40 line in the laboratory can get 6 lines.”

This Linggang 32-line LiDAR unit is not being tested on a vehicle for the first time. It has also been tested and compared with the Velodyne 64-line LiDAR unit multiple times. According to current research, a major requirement for LiDAR is that the density in the central area of ​​the imaging range must be high enough during scanning.

Upon hearing this, Yu Kai was even more astonished: "How did you do that?"

Baidu has an autonomous driving R&D team and has also purchased Velodyne 64-line radar. Although it is not their main focus, they are aware of the general situation.

Li Yifan glanced at President Yu and, seeing that he had no objection, said, "Velodyne's wiring harness is evenly distributed, but after our actual vehicle testing, we tried using different resolutions for the areas in front of and on both sides of the vehicle, and the effect was better, which could significantly improve the point cloud density in key areas."

He paused, then smiled with a touch of self-satisfaction: "With this non-uniform distribution design, a 40-line radar can almost be equivalent to an 83-line radar in key areas."

This is obviously not easy.

Yu Kai sincerely praised, "Indeed, everyone has their own expertise!"

Li Yifan sighed at this point: "But there are still many problems, and the road ahead is long and arduous."

Yu Kai fully understood the emotion in those words, but the unexpected surprise he encountered at the Silicon Carbon Group today made him feel quite invigorated. He smiled and said, "Being on the road is good enough. The road may be long, but we will reach our destination if we keep going."

He didn't dwell on the sentimentality, but went straight to a more professional question: "Are you working with silicon carbide companies on mobility solutions? How do you process the point cloud data from your LiDAR?"

LiDAR generates more than a million point cloud data points per second, and real-time processing of this data is a critical issue.

Li Yifan nodded in response to the first question, and then said, "Carbon Silicon and Linggang have established a joint laboratory. We are now preprocessing point cloud data. First, we convert the 3D point cloud into a sparse voxel mesh by voxelization to reduce the data dimensionality."

"This process uses 3D CNN for processing, followed by multi-view projection to project the 3D point cloud onto multiple 2D viewpoints, and then uses 2D CNN to extract features."

"Mr. Yu should know that LiDAR has high requirements for computing resources. We are currently considering a two-stage processing strategy: high-frequency, low-precision real-time pose estimation and low-frequency, high-precision global map optimization. This will balance the requirements for computing resources and accuracy."

Yu Kai's mind raced quickly, and he immediately said, "The information fragmentation between modules and the lack of adaptability to dynamic scenes under such a multi-stage processing framework is a very tricky problem. In a dynamic scene, an error in the point cloud segmentation of the previous frame will cause the target tracking trajectory to deviate in subsequent frames."

“If the point cloud is sparse, voxelization will further reduce the resolution and affect the accuracy of target detection. If we process it in modules, we will repeat the calculation. There are structural problems in the utilization of computing resources.”

He looked at the focused Li Yifan and discussed, "I think we still need a new deep learning model. I heard that Stanford is working on a path to directly process raw point clouds, which is to solve the problem of point cloud disorder by considering symmetric function modeling."

Li Yifan subconsciously raised an eyebrow: "This will directly change the mainstream technical path of point cloud processing!"

Yu Kai nodded: "Yes, I just don't know if it can be done. If it can be done, it will be a great milestone."

The two spoke quickly as they discussed the industry's top-level developments.

Yu Xing couldn't get a word in edgewise, so he just leaned against the front of the car, smoking silently, pondering the significance that Vice President Yu's trip might reveal.

He knew that Baidu was the "Whampoa Military Academy" of AI and had considered actively recruiting them, but he didn't feel an urgent need at this stage. However, judging from Yu Kai's performance, Baidu's "early bird" mentality might be about to kick in again.

After a while, Yu Kai's interest in talking was finally satisfied. He looked at President Yu again and couldn't help but ask, "President Yu, why are you doing this kind of thing now?"

Yu Xing countered, "Shouldn't we do it? If President Yu is interested, we can take some time at noon to hold a small meeting and call the lab staff over to discuss it together."

Without the slightest hesitation, Yu Kai said, "Okay, then I'll take my leave."

He was gradually becoming depressed on Baidu. When he finally met a fellow traveler, it was like meeting a fellow traveler in the desert. Even though he didn't see an oasis, he was still excited and thrilled.

Moreover, unlike Baidu, Silicon Carbon Group is an automaker, and its technological layout at this stage indicates a grand ambition.

Yu Kai was itching to go, so he simply called and asked for two days off to postpone his return to Beijing.

During the lunch break, Yu Kai had a great time discussing various issues, including those he had encountered at Baidu and those he hadn't, as well as directions Baidu had tried and angles he hadn't considered.

After all, Baidu is just an internet company, and it can't always start from reality in many areas.

The meeting ended at 1 p.m., and we ate lunch at the cafeteria, which we hadn't had yet.

With Yu Xing and Li Yifan accompanying him, Yu Kai reflected on what he had seen and heard and what he had discussed over the past half day, and he repeated his sentiments: "President Yu, intelligent driving is a huge challenge for large companies. They all have to tolerate and accept that a group of the smartest minds are doing some work that is difficult to quantify and whose ROI is difficult to calculate."

This was still a question that came from the bottom of his heart.

What kind of large companies are conducting research on intelligent driving?
Google, Apple, and Baidu are among the companies that are testing the waters.

These companies face the very serious problems he mentioned: they can't figure out the direction, they can't understand the processes, and they can't figure out the expectations.

Therefore, even for someone like Li Yanhong, his actual level of investment and interest was not as high as it appeared on the surface.

Yu Xing looked at Yu Kai's affectionate expression and wondered what he was babbling about... He thought for a moment and pointed outside: "Mr. Yu, can you see that sign over there?"

Looking out the window, Yu Kai could just see the sign for "Carbon Silicon Data Company" from this angle.

“Electrification and intelligentization,” Yu Xing said. “Carbon silicon data, carbon silicon data, we want to mine the value of data in the new era, this is our goal in the electric vehicle era.”

Yu Kai bluntly stated, "But that's very difficult. Does Silicon Carbon Group have that kind of capability?"

"Baidu certainly has the capability, right? Has Baidu invested according to your wishes?" Yu Xing asked two questions in a row. "Mr. Yu previously mentioned wanting to start a business. Do you think you have the capability to get involved in that field if you want to run a startup?"

He looked at Yu Kai, who was deep in thought, and laughed, "Carbon Silicon Group can't compare to Baidu, but it can compare to your company that you haven't even registered yet. If you can start your own business, why can't we?"

Yu Kai let out a sigh and said, "Mr. Yu is right. I don't think Baidu can pull this off."

Yu Xing then reminded him, "You haven't started your own business yet; you're still the vice president of Baidu."

Yu Kai shook his head, then looked at Li Yifan, who had returned from abroad, and asked curiously, "How did you decide to return to Lingang to start a business, Mr. Li? The market for lidar is clearly more promising overseas."

"Mr. Yu is willing to invest money, to facilitate investment from institutions, and to help find people," Li Yifan said succinctly. "We believe that the domestic industrial chain has a great opportunity to develop."

After thinking for a while, Yu Kai talked about his considerations regarding the chip aspect of intelligent driving.

Simply put, he sees chips specifically designed for intelligent driving as a trend.

The meaning behind his words is quite clear.

At the end of his eloquent speech, Yu Kai asked, "Mr. Yu, do you think we can do this?"

“It’s definitely possible, I understand what you mean.” Yu Xing understood Yu Kai’s dilemma and could find a comparable company. “You want to create a company similar to Mobileye. Well, if you’re really determined, I can help you contact friends like IDG, Today Capital, and Red Falcon Capital. Carbon Silicon Group can also invest in the angel round.”

Mobileye is an Israeli company that developed the EyeQ series of dedicated chips because traditional general-purpose processors could not meet the real-time requirements of visual algorithms. It has collaborated with Mercedes-Benz, BMW and Tesla, and its technology covers functions such as automatic emergency braking and pedestrian detection.

However, it currently focuses on visual algorithms for cameras and has not yet been heard of involving multi-sensor fusion solutions such as LiDAR. It is said that it is expected to go public in the US in the second half of the year, with a market value of several billion US dollars.

Yu Kai was not surprised to hear President Yu mention Mobileye's name, and said seriously, "Domestic chip design is relatively weak, and the environment is different from that abroad."

Yu Xing laughed and said, "That's still better than chip manufacturing."

Yu Kai nodded: "Indeed, um, Mr. Yu really thinks it's feasible?"

“If you’re short on money, then keep using Baidu. If you’re not short on money…” Yu Xing looked at Li Yifan and said, “Lingang welcomes you.”

Yu Kai is indeed not short of money.

He sighed: "Mr. Yu, Lingang is a good place, with Linggang Company and Silicon Carbon Group. However, I promised my friend that I would stay at Baidu for a while longer. Mr. Yu, I will come to Lingang. I hope you can introduce me to those investor friends then."

Yu Xing smiled and said, "Then come anytime, and I'll introduce you."

Yu Kai thanked President Yu for his promise and, although he would remain at Baidu, decided to shorten the timeframe.

Yu Xing handed him a cigarette and made a small request: "I asked around at noon and heard that Baidu Deep Academy has amazing technology in speech recognition. We at Carbon Silicon are really lacking in this area. I wonder if you could help me make an introduction?"

Silicon Carbon Group's electric vehicle project requires voice recognition technology, and it is currently conducting research and development in conjunction with Silicon Carbon Data, but progress is slow.

Baidu is quite good at this. Its voice team was one of the best teams in the internal evaluation in 2012, and Deep Academy is said to have made great strides in deep learning technology.

Yu Kai nodded slightly and agreed: "No problem. I'll talk to them when I get back to Beijing and see if we can make a joint effort. Baidu is indeed a good company. If anyone is willing to come to Hong Kong, then it depends on their own personal wishes."

He then added with a touch of self-deprecation, "What's even better is the application of deep learning in advertising systems and webpage ranking."

Yu Kai wasn't against Baidu using it this way, but he felt that the company was prioritizing it far too much.

Yu Xing remained restrained and did not take the opportunity to make sarcastic remarks.

In the afternoon, Yu Kai, with a changed mindset, visited the neighboring Silicon Carbon Data Center. After the visit, he suddenly felt like he was in a mini-Baidu, since both Zhai Yifei and Yang Zhenyuan were originally from Baidu's technical team.

Yu Kai joked with Yu Xing: "No wonder Lingang welcomed me; it turns out they've welcomed me here many times already."

"This is Lingang's welcome. If you'd like to join Silicon Carbon Group, Silicon Carbon Group would also welcome you," Yu Xing said objectively. "However, Silicon Carbon Group may not be able to do things exactly as you envision."

Yu Kai's mind was made up: "Yes, thank you, President Yu. Indeed, I still plan to do this part myself."

He believes this prospect is definitely achievable, although he may not be the one to realize it.

Yu Kai stayed in Lingang for another day to witness the joint research and development between Silicon Carbon Group, Linggang Company, and even Zhibo Company. He felt a strong urge to join such a "circle of friends".

Two days later, Yu Xing from Lingang received an angry phone call from Li Yanhong.

"President Yu, what did you say to Yu Kai?" Li Yanhong asked, her tone extremely dissatisfied.

"He said he wanted to start his own business, and I told him I supported him." Yu Xing guessed what was going on and answered truthfully.

"You...you..." Li Yanhong said, "What makes you think you can support him?"

Yu Xing retorted, "If you don't support him, does that mean you can't let others support him? He came here on his own and asked the question himself. I think he's a good person, so I'll support him. If you can persuade him to stay, then go ahead and persuade him."

Li Yanhong hung up the phone in annoyance. She was extremely dissatisfied with Yu Kai's request to resign after three months and Yu Xing's instigation. She also couldn't understand why Yu Xing's support could have such a big impact, according to Yu Kai.

What is Yu Xing's status? What is the status of Silicon Carbon Group? And why?
At the same time, he noticed the flow of talent from Baidu and decided to curb this trend.

(End of this chapter)