"In the Financial Report telephone conference in February 2018, I said that LiDAR is Autonomous Driving's 'crutch'.
Many people don't understand and think I'm crazy.
But you are on to the point - LiDAR is using expensive hardware to make up for the shortcomings of the algorithm.
The real way to solve the problem is not to add more sensors, but to make the algorithm good enough so that the camera can solve the problem."
Zhao Chen said:
"What a person relies on to drive is two eyes and a brain.
Without LiDAR, without Millimeter-Wave Radar, people can still drive a car very well.
If computer vision can reach the level of the human eye, then LiDAR is not a necessity.
The key to the question is not 'should we use LiDAR', but 'whether computer vision can reach or even surpass the human eye'."
Elon Musk looked at him and suddenly smiled. "You have read my 2017 TED talk."
Zhao Chen nodded. "I read it. You said - 'Once you solve the camera problem for vision, you can solve Autonomous Driving's problem;
If you don't solve the visual problem, there will be no solution to Autonomous Driving's problem.'"
Elon Musk picked up the cup of cold coffee, took a sip, and put it down.
"In May 2016, after the Florida accident, I broke up with Mobileye.
Their solution is too conservative, and the EyeQ3 chip does not have enough computing power to handle complex visual scenes.
Moreover, Mobileye does not open up the underlying code, so I can't change it even if I want to."
Zhao Chen said:
"So you developed Autopilot 2.0 yourself, replaced it with the Drive PX 2 chip of Nvidia, and the computing power increased directly from 0.256TOPS to 10TOPS.
The number of cameras has been increased from 1 to 8, covering 360 degrees. Millimeter-Wave Radar has changed from an auxiliary sensor to a main sensor.
There were huge changes in that year, from hardware to algorithms, it was almost a complete overhaul."
Elon Musk leaned on the back of the chair and looked at the ceiling.
His expression became a little distant.
"It was a painful year. All of Mobileye's code was removed and we started from scratch.
Everything must be built by ourselves - visual recognition, sensor fusion, decision-making and planning, control execution. The entire link is all self-developed."
He turned his head and looked at Zhao Chen. "Do you know why I must develop my own chip?"
Zhao Chen thought for a while.
"Because the Nvidia chip is not enough. Drive PX 2 has a computing power of 10TOPS, which sounds very high, but it is not enough to actually run a neural network.
After Autopilot 2.0 enables 8 cameras, the amount of image data generated per second is huge. 10TOPS is barely enough to run L2. It is far from enough to run L4 and L5.
So you need to develop your own FSD Chip, whose computing power is at least 20 times that of Drive PX 2."
Elon Musk looked at him with a hint of surprise in his eyes. "You are very familiar with these numbers."
Zhao Chen smiled.
"I have studied it. Not only FSD Chip, but also your self-developed Dojo Supercomputer.
The essence of Autonomous Driving is AI, and the essence of AI is computing power and data.
Self-research on computing power and self-collection of data is the most difficult path, but it also has the highest barriers. If others want to copy it, it will take at least five years."
Elon Musk was silent for a while, and then changed the topic.
"What do you think of the commercialization time of Autonomous Driving?"
Zhao Chen thought for a while.
Of course he knows the timeline of Previous Life--
Tesla only really started to push NOA in 2019, launched FSD Beta in 2020, and released Robotaxi in 2024.
But that's the result.
In the spring of 2018, Elon Musk had just reduced the timetable on South by Southwest Conference from "two years" to "18 months", and confidence was booming.
He didn't want to pour cold water on it, but he also didn't want to blindly agree.
He said something ambiguous but true.
"The time for commercialization depends on two things - when the technology matures and when regulations are relaxed.
In terms of technology, I think it may take a little longer than you expect. It's not that the technology isn't good, it's that there are too many long-tail problems.
99% of the scenarios are solved, and the remaining 1% of Corner Cases may take 99% of the time.
But this is the right path, we are just walking on it."
Elon Musk looked at him and suddenly asked, "How long do you think it will take?"
Zhao Chen thought for a while and spoke the truth.
"The full-scenario L4 may take more than five years.
But functions such as highway Autonomous Driving, automatic parking, and summoning will be very good in the next two to three years."
Elon Musk nodded without refuting.
He stood up, walked to the window, and looked at the Hudson River. The sun shone on the side of his face, and his expression became a little softer.
"Zhao," he said without looking back, "Do you know why I am willing to talk to you about this?"
Zhao Chen said nothing.
Elon Musk said:
"Because you don't just invest money. You really understand what we are doing. It's not about Financial Report, it's not about the K-line, it's about the technology itself.
You understand chips, sensors, and algorithms. You even know the computing power requirements of FSD Chip.
Tesla doesn't need investors who can only invest money, it needs people who can understand why we do what we do."
Zhao Chen stood up, walked next to him, and looked at the river outside the window.
"Elon, I believe in Autonomous Driving. Not because I believe in technology, but because I believe in data.
You now have billions of miles of road test data, Waymo only has a few million miles.
This gap cannot be made up by LiDAR. Data is the fuel of AI. Whoever has more fuel can run further.
Tesla There is no opponent on this road."
Elon Musk turned around, looked at him, and was silent for a few seconds. Then he smiled and stretched out his hand.
"Zhao, welcome to SpaceX. Thank you too, I believe in Tesla."
Zhao Chen holds his hand.
"Elon, come on. I'm waiting to see your car drive from California to New York without touching the steering wheel."
Elon Musk smiled. "It will happen. Maybe sooner than you think."
He glanced at his watch, said he had a meeting, turned around and left.
Zhao Chen stood in front of the window, watching Elon Musk's car drive out of the parking lot and disappear into the traffic of Manhattan.
He leaned back in his chair and took a long breath.
The conversation just now about Autonomous Driving was the first time since his rebirth that he had such an in-depth conversation about technology with someone.
It's not Financial Report, it's not valuation, it's not ROI - it's the technology itself.
This feeling is different from talking about business.
He picked up his phone and sent a message to Zhang Qingxue:
"The talks are over. Elon Musk has confirmed the financing. 100 million US Dollar, proceed as originally planned."
Zhang Qingxue replied instantly, "Received. Xiao Chen, what other arrangements do you have over there?"
Zhao Chen thought for a while and replied, "We will start talking about Existing Shares tomorrow. Are you ready?"
Zhang Qingxue replied:
"Lawyer Lin has made an appointment for Wednesday and Thursday. The price bottom line is estimated at 24 billion, and the lock-in period will not exceed three years. We will not pursue Board of Directors Seat."
Zhao Chen took a look, put away the phone and walked out of the club without saying anything else.
The sun shone warmly on his face. It was New York in March, the wind was still cold, but the sunshine already had the flavor of spring.
He got in the car, leaned on the seat, and looked at the Hudson River shining in the sun outside the window.
There are several more sailboats on the river than yesterday.
He closed his eyes, still thinking about the conversation just now - the schedule of pure vision and LiDAR, FSD Chip, Autonomous Driving.
These topics are far away from investment and very close to the future.
He chose to believe in Tesla, not because Financial Report is beautiful, not because the K-line is beautiful, but because he believes that Elon Musk's technical judgment is correct.
Previous Life's experience is the best proof.