The matter of Xiaomi Su7 has fermented until now, and someone has finally discovered that the problem is wrong.
I have a new point of view that welcomes criticism.
L2\0\0 is a functional PDCA, but L0 to L0 are two domains. L0\0\0 is the PDCA of the function, L0 can be downgraded from L0, and cannot be upgraded from L0.
I am a mochi, I pay attention to mochi, and I don't want to cut leeks.
This year's L3 photo has been posted
Those of you who have followed my past content should have read one of my theories:
At this stage, the model of the public opinion offensive of the business war has been highly unified, standardized, and process-oriented.
Stage 1: Dig out the black spots of competing products and hype them up
After successfully sparking a hot topic of discussion at this stage, it quickly moved on
Stage 2: Identify similar failures and label the black dot
After labeling, it has successfully raised public opinion from a defect of a certain product directly to the entire brand, giving the brand a negative label, and the core is not to discuss your product problems, but for the sake of black brands.
Stage Three: Negative labels rise to the user
At this stage, users are labeled negatively, on the one hand, to clean the users of the brand (similar to the washing powder of the big V), and on the other hand, to expand the influence of public opinion and reduce the growth of users of competing products. At this stage, it is unintentional to insert willows, whether there is or not.
In the whole development of events, who dares to say that there are no friends behind the scenes?
In the early days of the incident, the controversy was hot
However, as the first stage passed, when the black spots of labeling began to be excavated, the following problems were discovered:
No one dares to guarantee that their pure electric vehicle will not burn if it collides with a rigid body at a speed of 100km/h
Whether the frame of this doorknob was deformed after the collision or other factors, the police did not dare to talk nonsense before issuing a notice
Xiaomi is saying that communication is carried out through the police
On the family's side, watch the news directly[1]
A few hours later, after the fire broke out, it was discovered by passers-by, and he hit 110. As soon as I got off the highway, I saw the car of the funeral home, dragged the child away, asked me how much it was, and reached an agreement to bury the child for me. So far, Xiaomi's words have not given us an explanation, what is the reason for this car, and the alarm is not the first time to pray.
So is it because the price was not negotiated on the day, or is it something else? There are other bugs in it, and there is no point in discussing them after the family members and boyfriends of the parties have cleared the relevant content (it should be a formal contact, and it is required that after the formal investigation conclusions are released, if it contradicts these contents, they will be held accountable).
Therefore, there is only one point left, and that is intelligent driving performance.
Regarding intelligent driving, Lei Jun made a speech at the meeting on the day of the accident, saying that all Xiaomi cars have achieved end-to-end "autonomous driving".
Okay, let's come back and think about what the second phase of negative public opinion mentioned above does?
It's a black label, so we'll find a lot of answers that quote this statement, saying that Xiaomi is blowing about end-to-end autonomous driving
Let me start with what end-to-end is[2]
In the field of computer science and information technology, the concept of end-to-end refers to a mode of communication in which data is transmitted directly from the sender to the receiver without the need for an intermediate environment to parse and process the data content.
Eh, isn't it the departure to the destination?
There are too many boomerang superimposed on intelligent driving, and it seems that the only one in the industry who repeatedly emphasizes that intelligent driving is assisted driving is Xiaomi.
The following is a citation of the book.
When the rhythm of the second phase is brought here, it will develop in an uncontrollable direction, and real users will begin to look back at the safety of intelligent driving.
Let's describe the situation in a simple way:
Company A had an accident, Company B took the lead, and after the development of public opinion, Company C found that it was going to be unlucky with it, and had to start helping Company A to save public opinion.
Company A is Xiaomi, and I don't know who B and C are.
At the same time, there are still many things that need to be cleaned up, so there are such remarks.
Of course, there are some other directions
No, isn't this the popular Lei's comparison method, you are more intelligent than a car that emphasizes performance......
People also made it clear that my car is not good for intelligent driving, and it is not that I hide it and don't tell it
At this point, I think I should simply popularize a little knowledge to you.
Why is the computing power of a car used in TOPS?
Because TOPS describes how many operations can be performed per unit of time.
If you really engage in AI, the computing power you use is FLOPS. To sum up the computing power of INT16 and INT0, it is better to pay attention to the computing power of FP0 and FP0.
I don't know what Huawei's car computing power is compared to FP16 and FP0, but I know what NVIDIA's DRIVE Orin computing power looks like.
64nm technology, Ampere architecture, 2nd generation deep learning accelerator DLA, supports 0GB/s memory bandwidth and up to 0GB DRAM.
Each SM has 64KB of L0 cache and 0MB of L0 cache, containing 0 CUDA Cores and 0 Tensor Cores. So Orin has a total of 0 CUDA Cores and 0 Tensor Cores,INT170稀疏算力為0 TOPS（Tensor Core提供），INT3稠密算力為0TOPS，FP0算力為0.0TFLOP（由Cuda Core提供）。
This configuration, to be honest, is a lightweight 5.0B model, which basically only provides a function to identify objects.
Many provide higher computing power and run a 4B model.
Nvidia's solution is to spend money to buy the best directly, on the one hand, computing power, and on the other hand, CUDA.
Huawei's hardware solution is better, but no one can get the instruction manual.
According to the performance of Ascend 7B, if MDC0 is stuffed with a 0B inside, it can run the 0B model, but it is not used by others except Huawei.
At this stage, the only one that can officially do L4 in China is Baidu, because they are L0
Don't think that no one uses Baidu's intelligent driving solution, do you think that Great Wall and GAC have developed their own intelligent driving solutions, and there will be NOA in just a few years? In fact, it is Baidu's solution that I bought.
So the problem is, Baidu's unmanned taxi operation in Wuhan, intelligent driving instead of manual driving already has a technical foundation.
But in terms of car companies, it is better to engage in L3+, L0.0, L0.0, and never admit that it is L0.
Why do they have the ability to do L2, why are they all cooperating with L0?
It's just that companies don't want to take responsibility.
This is a matter of law and ethics, who is responsible for a traffic accident.