Ye Chen said: "It has been fully delivered. It still needs to be slowly explored in use. The next step is how to maximize the strength of this set of models."

After that, Marven asked him: "Brother, do you think this thing is of great use to your professional field?"

Lang Hongjun thought for a while and said: "If it is just communication, it is not very useful, but if it involves encrypted communication, it is still very useful. In terms of encryption strategy, AI should be more complex, which greatly increases the possibility of cracking." Difficulty, and if we do satellite monitoring in the future, AI should still be very helpful.�

Ye Chen asked him: "Which type of satellite surveillance is this brother talking about?"

Lang Hongjun said: "It's just satellite photography. Today's civilian satellite photography is very accurate. I can't say that Google's satellite pictures can clearly capture license plate numbers on the ground. Letters as big as the hood can definitely be photographed very clearly." This kind of satellite image is actually very practical, and ordinary military targets can also be identified with the naked eye.�

With that said, Lang Hongjun added: "Actually, I was thinking on the way here that your current AI model feels like a nuclear power plant that specializes in producing computing power. Its computing power is extremely abundant, and it can not only supply certain A factory produces a certain product, and it can supply an industrial city. Many factories of various kinds produce a variety of products, so I hope that in the future we will prepare to build our own satellite like Google. Picture library, so that AI can be used to achieve image quality enhancement and intelligent retrieval, which will be of great help to Wanlong Palace�s security and interests in the Middle East, and even its interests around the world in the future. It is also my time I especially want to come and see this data center and meet with you about its core power.�

Ye Chen nodded and said: "If civilian-level people can clearly capture the large words on the engine cover from satellite orbit, then the clarity is really amazing. Whoever dug an underground bunker somewhere can probably also see it. If the data is updated quickly enough, the timeliness and practicality will be further enhanced.�

"That's right." Lang Hongjun said: "But the key to the problem is that the higher the accuracy of the image, the higher the requirements for the equipment, the greater the workload of collection, and the larger the image database will be;"

"If we compare the earth to a football field, and there are 50,000 people in the football field, and we take a group photo of 50,000 people, each person in the photo will probably only have a few pixels, and there is no way to tell who is who. , if you use this kind of photo as a basis, it is completely impossible to find someone from it. Even if you know that there is the person you want to find, there is absolutely nothing you can do with such a photo;"

"If we place the camera equipment in the center of the football field, divide the football field into ten areas, take ten photos in each area and combine them, then the accuracy of the picture will be increased by ten times, but due to the large number of people , even if the accuracy is increased by ten times, it is estimated that you still can�t tell who is who, you can only see the vast sea full of people;�

"If we divide 50,000 people into a hundred square arrays, take a photo for each square array, and then combine the one hundred photos together, the accuracy will be improved to a certain extent. With a little effort, we can probably see each other. It�s a boy or a girl, but at this time, the number of pictures in the database has reached one hundred;�

"If we continue to improve the accuracy and want to ensure that the facial features of 50,000 people can be identified in one photo, then we will probably need to increase the material library to a thousand;"

"If we apply the most precise satellite image ratio, it would probably mean that we need to use a high-resolution telephoto lens to take N close-up photos of each person among 50,000 people to ensure that each person's facial features and every pore can be clearly seen. Presented, each person's facial close-up photos are included in a database. In this way, if we want to find a certain mole on a person's face among 50,000 people, his corresponding database must be there. As long as we find For his photos, just find the one covering the mole from N photos.�

"But finding a mole is relatively simple. After all, the mole is on the person's face. As long as you find the person, you can find the mole on his face. But if we want to find a specific person in a stadium with 50,000 people, Even if we know that mosquito must have been captured in a certain photo, it is very difficult to find one mosquito from hundreds of thousands or even millions of photos.�

"If you scale up 50,000 people to the entire earth, the database would be unimaginably huge. In this case, it would be as difficult for us to rely on human power to find and identify a moving target. It would be like finding a needle in a haystack. We know that the needle is in the sea, but it may not be fished out in a lifetime, but if we feed all the data to the Ai model and use it to train it so that it can act according to our ideas, then Ai can be used in the shortest time. , help us find whatever we want to find.�

Speaking of which, Lang Hongjun added: "In addition, there is another area that AI is very good at, which is image quality enhancement. If we take a photo from a distance, the photo will be a little blurry when enlarged, but it is not blurry enough to be completely recognized. If it is not clear, Ai will automatically optimize the blurred image based on its own understanding of the image and the pixels to make the blurred image clear. However, when facing an extremely large database, the computing power will If it�s not enough, if you want to optimize them all, you probably won�t be able to do it in a year or two;�

"But if there is enough computing power, it will probably take a few days of work. Private companies can take satellite pictures, and private companies can also launch or commission satellites used to take pictures, but ultra-high-precision satellites are not private companies. The company can do it. The best we can do in terms of hardware is only about the same as Google's satellite images. In this case, if there is an Ai model and enough computing power to support it, we can use Ai To make up for the lack of hardware accuracy, the image quality can be ahead of Google;"

"The pictures are more accurate and can be automatically retrieved by AI. In that case, the strategic use will be great."

Ye Chen said thoughtfully: "If this step can really be achieved, the strategic significance is indeed great, not only for Wanlong Palace, but also for me personally. Brother, is it difficult to achieve this step?"

"It's difficult." Lang Hongjun nodded and said: "To tell you the truth, Mr. Ye, anything related to satellites is difficult. It's technically difficult, it's difficult to produce, it's difficult to launch. It's even more difficult to launch successfully and get into orbit smoothly. We're making preparations now." Yes, it is a communication satellite. It is one of the lowest-tech among all satellites, but if you want to launch a photographic reconnaissance satellite, it will be more than ten times more difficult in all aspects. " JrNovels.com