Chapter 1257: Fancy or Big Gram?

Everyone looked at Liu Yang, Wu Bo frowned and said:

"The new architecture design is not mature yet. It is difficult to quickly improve the performance of artificial intelligence wafers and make the system run stably under the existing conditions."

Liu Yang did not refute Wu Bo's words, but nodded in agreement.

"Well, yes, Mr. Wu is right. Our current generation of artificial intelligence wafer design is not yet mature and may not be used at this stage.

But we can create other conditions.

We all know that there is another way to improve the performance of artificial intelligence, which is that more data is needed for training and tuning.

And this kind of tuning and training does not just need to be tested in simulated scenes and closed venues.

The situation on the actual road is more complex and changeable, so we need to collect parking data of vehicles in the real environment.

These data include the layout and management regulations of different parking lots in different cities, the parking habits of drivers in different regions, etc.

We currently only have the data of the last hundreds of thousands of car owners connected to the backend data center. To be honest, it is too little.

Therefore, I suggest that all the collected data of all old models be added to the background database, so that there will be more data for tuning and training.

In this way, the performance of this artificial intelligence wafer can definitely be improved better and faster. "

Everyone sitting here is not a novice, of course they know whether the method he said is useful.

The improvement of artificial intelligence wafers does require a large amount of data for training and tuning.

Massive data is the key foundation for artificial intelligence wafer training models and performance improvement.

Through a large amount of data input, Jingyuan can learn various patterns, characteristics and rules.

Moreover, the rich data can help artificial intelligence wafers continuously adjust and optimize the parameters of the model and improve the accuracy of the model.

At the same time, a large amount of data can also enhance the generalization ability of the model, allowing Jingyuan to make accurate predictions and judgments when facing new and unseen data.

Take autonomous parking as an example!

During the automatic parking process, the vehicle needs to accurately perceive the surrounding environment, including the location, size, and shape of the parking space, as well as surrounding obstacles such as other vehicles, pillars, walls, etc.

Different parking lot environments vary greatly, including open-air parking lots and underground parking lots, with different lighting and spatial layouts;

There are also various types of parking spaces, including vertical parking spaces, parallel parking spaces, diagonal parking spaces, etc.

Only through a large amount of data training can artificial intelligence wafers learn to accurately identify this information in various scenarios.

For example, in a dark underground parking lot, Jingyuan needs to accurately determine the parking space lines and obstacles based on the data collected by the sensors. This requires data training based on a large number of similar scenes, so that Jingyuan can master the parking spaces under different light conditions. Recognition ability.

Moreover, the parking lot is not a static environment, and there may be dynamic situations such as pedestrians walking and vehicles driving.

Artificial intelligence wafers must be able to sense these dynamic changes in real time and adjust parking strategies in a timely manner according to the changes.

This requires a large amount of dynamic scene data for training, so that Jingyuan can learn to identify and predict the movement trajectories of dynamic objects in order to avoid collisions during automatic parking.

Moreover, automatic parking needs to plan a parking path that is both safe and efficient so that the vehicle can park smoothly into the parking space.

This involves precise control of the vehicle's steering angle, driving speed, braking timing and other factors.

Through a large amount of data training, artificial intelligence wafers can learn the best parking paths under different vehicle sizes and parking space conditions, thereby improving the accuracy and efficiency of automatic parking.

During the actual parking process, the vehicle's sensors may have certain errors, such as radar measurement errors, camera visual deviations, etc.

Artificial intelligence chips need to learn to correct and optimize these errors through a large amount of data training to ensure the accuracy of parking paths.

For example: Based on past data experience, Jingyuan can determine that in a specific parking lot, the radar measurement data at a certain location may have certain deviations, and thus make corresponding adjustments during path planning.

Finally, during the automatic parking process, some special situations may occur, such as the parking space being occupied, the parking space line being unclear, vehicle failure, etc.

Artificial intelligence wafers need to learn to identify these special situations and take corresponding countermeasures through a large amount of data training.

For example, if Jingyuan recognizes that a parking space is occupied, it needs to re-search for other available parking spaces; if the parking space lines are not clear, Jingyuan needs to make inferences and judgments based on the surrounding environment information.

In short, a large amount of data from car owners is collected and used to tune and train this Jingyuan, which can definitely improve this Jingyuan.

There is no shortcut in this regard. Unlike a new force in later generations, which only has about 100,000 cars on the road, it can be said that the assisted driving functions tuned based on these data can enter the first echelon.

Anyway, Chen Changliu definitely doesn't believe it now.

It's just an automatic parking function. The flower shop already has hundreds of thousands of cars with reversing images. The data has been transmitted to the background, but it still hasn't been adjusted to stability.

Not to mention the more difficult smart driving.

...........

Chen Yanfeng took over Liu Yang's words and said in a deep voice:

"Mr. Liu, if the data of nearly 300,000 cars are connected to the backend database in real time, it will not only improve their car computer problems, but also add some accessories.

Then the cost is not 30 or 20 million, at least more than 300 million."

At this point, everyone was silent, because they didn't know how to choose now, and could only look at Chen Changliu!

Chen Changliu didn't think too long and made the decision directly.

"Mr. Liu, you and Chen Yanfeng make a plan, I approve it.

As I said, we need to race against time. As long as we can solve it with money, we can solve the technical problems faster, then I will support it........."

Chen Changliu made the final decision, and the incident became a foregone conclusion.

At the same time, some passers-by uploaded the live video of Chen Changliu's test to the Internet.

However, these are the videos that were taken after Chen Changliu appeared later, basically focusing on the scene of Chen Changliu interacting with the car computer system after getting on the car.

Although the platform has not yet pushed the stream, the public relations department of the flower planting family has not yet left.

However, the traffic brought by the name Chen Changliu is considerable, so although the number of views of these videos has not reached the point of explosion, it is actually not low.

And the heated discussion is inevitable.

"What function is this? I haven't seen it before. It seems a bit cool to control these functions with voice!"

"It's really interesting. It's very practical on the highway."

"Don't you think it's fancy? Things that can be solved by pressing two buttons have to be said with your mouth, and it doesn't seem to be very efficient?"

"Hey, the upstairs, I saw your homepage. You brag about your Kamoula all day long. It's a car suitable for the general public.

If these functions appear on your Kamoula, you are definitely kneeling and licking it now, instead of saying here that these are fancy functions!"

Now the flower grower and Chen Changliu have become the focus of everyone in the domestic business community, so when this matter has just become a little hot, many people in the industry have seen these videos.

Then every car manufacturer frowned, because they can't do these functions at all now.

Although they can see from the comments that some people say these functions are fancy, their vision is different from others.

Let alone other things, this is definitely the best.

And all the current models of Zhonghuajia are mainly young and fashionable, and such functions are definitely popular among people of this age group.

They can already imagine that when the new Qin and Song are launched, the competition in the A-class sedan and compact urban SUV track will be more intense.

The sales of Qin and Song, which are already ranked at the top, are expected to be even better.

This video originally showed the performance of the car, which should be a concern for the executives of these car companies.

But in fact, when a person in the track saw these videos, his face changed more than those executives of the car companies.

PS: The typos will be updated first and corrected later.