Top of the big era

Although Zhou Buqi is a technical layman, before his rebirth, he was often brainwashed by concepts such as cloud computing, big data, and artificial intelligence.

After being reborn, I paid special attention to these aspects and gradually gained some understanding.

He was quite smart. When he heard what Professor Tang said, he immediately reacted and blurted out: "What is the change in algorithmic thinking from logic to statistics?"

Professor Tang was slightly startled, as if surprised, "Mr. Zhou's understanding is amazing!"

Pei Yao brought the four T7 technical experts and gained a new understanding of Boss Zhou's level.

Awesome!

From artificial intelligence to big data, this is a cutting-edge scientific and technological topic in the world. Even insiders are confused and can't figure out the direction.

Even one of the T7 experts couldn't react immediately.

The big boss, a layman, figured it out so quickly?

Of course, this is related to Zhou Buqi's level. He usually interacts with scientists at the level of Dr. Shen Xiangyang, Dr. Wang Jian, and Dr. Lu Qi, or communicates with scientists from Tsinghua and Oxford.

With the best scientists in the world as teachers, can we not learn things quickly?

At the same time, this scene also made many Ziweixing staff present feel deeply, and they deeply understood the reason why the company requires younger people in the talent structure.

It’s good to be young!

Learn things so fast.

Pei Yao admired her from the bottom of her heart and said with a smile: "Our boss is a layman with a pure computer major. He may not have even read an introductory book on "Computer Basics". However, in his usual work, he will deal with many scientists. , I will also participate in some technical meetings. After a long time... I may understand it."

Superboy seconded, "Mr. Zhou is a genius."

Professor Tang nodded in agreement, "Mr. Zhou's comprehension and understanding are even better than those of an expert."

Zhou Buqi has always been very humble and said with a smile: "I'm not even close to being a genius! I met an entrepreneur in Silicon Valley named Elon Musk. That's the real genius. He studied computer science." Yes, but he wants to make rockets. He doesn’t know anything, so he just studies and makes products at the same time. It is said that his current understanding and application of aerospace engineering technology has reached the level of a first-class scientist."

Professor Tang said: "Mr. Zhou is not bad either."

At this time, there were already many people in the multimedia laboratory, and many other professors and doctors in the laboratory appeared to listen.

Professor Tang is the director of the Multimedia Laboratory and the leader in the field of artificial intelligence research in this laboratory. All professors, doctors and masters must follow his general research direction.

If he makes the right choice, everyone will step onto the big stage in the field of artificial intelligence and become senior experts in the industry. He made the wrong choice, and everyone was doomed.

However, Professor Tang is a Ph.D. graduate from MIT and is very authoritative, and everyone trusts him.

Facing so many colleagues and students in the laboratory, Professor Tang was just like attending class, speaking in detail. First, let’s briefly explain the change of thinking from logic to statistics.

The image is composed of countless points. The higher the resolution, the more points; the lower the resolution, the fewer points.

Doubling the image size actually doubles the number of points.

Algorithms in the past were based on mathematics and logic. That is, after having the previous node, and then deducing it to the next point according to mathematical logic, one point becomes two points, and the expansion is completed. Process and expand all the points in the image, and the enlargement of the image is completed.

But this is an exact process.

The selected node must be certain before the second point can be derived.

If the first node is vague and uncertain, no matter how correct the mathematical formulas and logical ideas are, it will be useless.

The new image processing ideas under big data apply statistical algorithms.

For example, there are two roads, both of which can reach the finish line. Which road does the athlete choose to take less time?

The traditional way of thinking is through mathematical and logical calculations.

Calculate the length of the distance, the athlete's speed, and the probability of falling affected by the athlete's condition, endurance, acceleration, deceleration, friction, etc.

Come up with a very complicated formula.

However, what if the athlete coming over is a disabled person with no legs, or a blind person? The formulas of the past are no longer useful.

The idea of ​​statistics is very simple.

There are two roads, it doesn't matter whether it's a normal person, a disabled person, a bicycle or a car, just statistics. The road on the left is measured 10,000 times, and the road on the right is measured 10,000 times.

After testing, the left side won 8,000 times and the right side won 2,000 times, and the results came out.

It takes less time to choose the path on the left!

Applied to the clarity processing of blurry photos, it is statistics under big data.

For example, a blurred nose, what does the original image look like?

Just statistics.

Select 10,000 different types of noses and blur them. Whichever nose point after blurring has the highest overlap with the nose point in the blurred photo will be the one.

However, achieving this is very complex.

As long as it is big data, it will definitely involve the issue of computing power. If you want to process and count so much data, an ordinary computer is useless and will crash as soon as it is started.

Professor Tang said: “In the past, image processing has never been able to surpass the scope of the human eye. If we want to rely on computers to break through the limits of the human eye, in addition to the understanding of big data and artificial intelligence related technologies, we also need powerful computing power. force."

Zhou Buqi smiled and said: "Computing power is not a problem. Ziweixing's cloud computing business is approaching maturity. If it is applied in the laboratory, I can help you contact Lenovo and use their HPC."

HPC is a High-type PC, that is, a high-performance computer or a supercomputer, which has very powerful computing power.

The technical principles used are similar to those of cloud computing, both of which are parallel clustering projects. Dr. Yang Zhenkun of Ziweixing single-handedly developed the Feitian database and provided technical reserves and support for the cloud computing business. Before he was poached by Ziweixing, he mainly did HPC research at Lenovo Research Institute.

Today's Lenovo supercomputing is not as powerful as it was ten years ago, and it can go all over the world to build supercomputing platforms for people. However, in the field of market-oriented applications, it has now reached the first level in the country. Ziweixing has two supercomputer platforms, one large and one small, the large one is HP and the small one is Lenovo.

The size of this multimedia laboratory is limited, and it is enough to help them build a small supercomputing platform.

Professor Tang couldn't help but be stunned, "The price is not low."

Zhou Buqi smiled and said, "It doesn't matter, I'll give it to you."

Pei Yao said with a smile: "We, Mr. Zhou, are the most supportive of the research of innovative technologies, especially in the fields of artificial intelligence, big data and cloud computing. They are mentioned in almost every meeting."

The professors and doctors in the laboratory were very happy.

This is great!

Their office purchased a batch of HP computers with excellent performance, but these were just ordinary computers with insufficient computing power. In the demonstration just now, it took 3 minutes to enlarge a very small picture. If it were a larger project, it would be stuck immediately.

With HPC, it will be much more convenient for their future research projects.

Professor Tang was very grateful, but the class was not over yet. He wanted to continue what he had just said: "What the eyes can see is processed in a precise way. As I just said, from 1 to 10 to 100 to 1000. The changes are very big. Large enough to be distinguished by the naked eye. But it is different if it is reduced, from 1 to 0.1 to 0.01, and when it is reduced to 0.0000...001 and 0.0000...002, the difference will be very small and invisible to the naked eye. But computers can distinguish. This is the key for computers to surpass the human eye in image processing. Powerful computing power is the foundation."

It would be difficult for Zhou Buqi to get involved in such a complicated matter.

Several experts from Ziweixing discussed it with them.

The general meaning is to label each clear point and then correspond to a fuzzy point. A clear point is 1, the corresponding fuzzy point is 0.0000...001, a clear point is 2, the corresponding fuzzy point is 0000...002, count them first.

In this way, when the fuzzy points 0.000...00ε appear, you can move closer to the clear point ε, completing the transformation from uncertainty to certainty.

Of course, this process may not be one-to-one, but may be many-to-many, and requires very complex and detailed statistics and analysis.

Zhou Buqi thought of something and asked, "Is this the principle behind using a mobile phone to take pictures of the moon?"

"What do you mean by photographing the moon?" Professor Tang didn't quite understand.

Zhou Buqi said: "The pixels of the mobile phone are too low, and the lighting of the night scene is not good, so it is impossible to take pictures clearly."

Professor Tang waved his hand, "If we just optimize the moon picture, it would be very simple. We can do it now. The moon is fixed and known, and the composition is simple. Just take a rough model and use artificial intelligence to enrich it. . But the original appearance of the existing blurred photos is unknown, and the difficulty will be greatly increased. If it is to take pictures of the moon, the difficulty for mobile phones is not to clarify the blurred pictures, but whether the computing power of the mobile phones can keep up. , the chip is better. There is no such mobile phone chip in the world now."

Another white professor said in Chinese: "Our laboratory is currently mainly engaged in portrait processing, mainly human faces. The complexity of human faces far exceeds that of the moon, and it requires face detection, high resolution, and facial recognition." To study all aspects such as feature distribution, you also need to have enough face images.”

Zhou Buqi immediately understood, "Yes, big data, you have to have data! The most important thing Ziweixing lacks is pictures, which can support it!"

There must be enough exact faces to obtain enough blurred face databases for point-to-point analysis and docking. Then, after a blurred face appears, a clearer face can be deduced through point-to-point analysis and docking.

This project must be done.

On the one hand, it is the accumulation and precipitation of technology.

On the one hand, it does make sense!

This is an unsolvable problem all over the world. If Ziweixing can solve it, it will make a lot of money and can be sold to police stations around the world.

Police stations all over the world are departments within the system, and cooperation with such departments brings great benefits. Moreover, this is a technology that is urgently needed for the police and will greatly increase the detection rate of crimes. As long as they appear, police from all over the world will rush to buy them.

Then make a fortune!

However, foreign faces are different from Chinese faces, and it is necessary to obtain enough samples of foreign faces for analysis and statistics. With the popularity of mobile phones and social networks, this is not an obstacle. Just write a crawler algorithm and go to the Internet to collect pictures.