Nearly thirty years old, established in science and technology

Chapter 69 If you want to move, you must be still first, and then take action when you are sure.
“Didn’t I tell you if it’s true or not?”

These things may not have been mentioned to others, but with Zhang Jian, Ji Hong confirmed that he had definitely mentioned it to him before.

"That's true," Zhang Jian said with a smile: "How can there be direct performance data that can be photographed on the face and feel!"

Ji Hong knew. The guy might have believed it when he told him last time, but he still didn't quite understand it, or he didn't have an intuitive concept.

"It's not that I don't believe you, I believe you very much." Although Ji Hong didn't say it, Zhang Jian probably sensed it and immediately explained:
"However, the impact is different. It's like transferring money to my bank card. I may not feel anything when I see a series of numbers, but when a large bag of cash is slapped on my face, it must be extremely shocking. Now, I feel like I got hit in the face with a hundred dollar bill!"

"Stop explaining. Explanation is just a cover-up. You just don't believe me!"

Of course Ji Hong understood Zhang Jian's mood. Even if he believed it in the past, before he saw the real effect, a big stone was always hanging in his heart. Now it has fallen to the ground, but he still deliberately teased He asked Zhang Jian for a while.

"Trust, trust, trust!" Zhang Jianli answered immediately.

"You can take your time and write the letter yourself. I'm busy. Hang up!" Ji Hong was too lazy to talk to him.

"Wait a minute." Zhang Jian hurriedly stopped him.

"Is there anything else?" Ji Hong wondered.

"Seeing this data, I have an idea, but I don't know if it is feasible or not."

Zhang Jian said: "Can we make a batch of small-scale graphics card cluster host equipment like NAS, such as a set of two or four graphics cards, etc., and directly equip the whole machine.

"And we have done all the preliminary work well, providing access interfaces, or directly providing an interface-based software tool, or even directly providing fixed training templates and parameters for some common types of AI models, and users can adjust them on the interface and configuration.

"This will lower the configuration and setting threshold for AI training, allowing non-professionals, even people who don't know how to calculate, to get started immediately and train their own AI models. What do you think?"

Ji Hong thought for a while: "But what do ordinary people use it for? If it is not equipped with corresponding application scenarios, isn't computing power just a false demand?

"Just like us, the previous personal work assistant, this one thing can prompt users to choose our graphics card. This idea is indeed feasible, but the computing power needs to adapt to the application scenario. Does this need to be designed?
"For example, if you tell an old man and woman in the countryside: Hey, uncle and aunt, our computing power has surpassed that of NVIDIA, do you want one?
"The aunt has to poke you with a cane. It's useless to her! For ordinary users, you have to first consider what the use of computing power is. It's better to start a survey first and see what the real needs of ordinary users are."

Ji Hong has always been like this. If you want to make a move, you have to be quiet first, and then take action after seeing it clearly. However, Zhang Jian's idea is still very good, and it can give full play to the capabilities of Juaner Intelligent Technology's AI-based on-chip interconnection system:

"We can start with individual developers, small teams, etc. who already have computing power needs, and make a batch to try it out!" Ji Hong suggested:

"Isn't Sora from the previous Vincent video very popular? Recruit a group of technology bloggers, provide them with hardware, and let them reproduce Sora's capabilities through open source. It doesn't matter whether it's strong or weak, just make it, and then directly promote it overwhelmingly.

"Moreover, there are many institutions in China that are preparing to reproduce Sora's capabilities. You can also contact one to cooperate and reproduce the full version or even a stronger version. Then a big news will be broken: [Just do this, Sora's capabilities will be 100% Reproduce at half the cost!】

"Isn't the breaking point you want for individuals or small-scale AI training companies to abandon Nvidia and embrace Moore Technology? Let's speak directly with facts! Our computing power and algorithm level are not just bragging!

"You act first. This data will not be disclosed to the public for the time being. We have to wait for the first public results of the growth pilot model. It would be best if the two sides can form a linkage."

"Okay!" Zhang Jian was also excited. Once upon a time, he suffered from the pain of not having any skills, but now he has the wisdom of curling ears, and he can speak so forcefully. Although he is almost twice Ji Hong's age, he has now developed the habit of consulting Ji Hong first on important matters——

For example, the launch of a small-scale cluster host should have been discussed with the company's senior management first, but as soon as the phone call came, I couldn't help but want to talk to him.

This is probably the suppression brought about by technology, and he is still happy with it - of course, it is also because he can see that Ji Hong really has no desire to get involved in traditional hardware design and manufacturing.

"As for chips with higher-level architecture..."

After hanging up the phone, Zhang Jian was also thinking. He had communicated so much with Ji Hong, and he knew some of Ji Hong's ambitions: "If he really gets involved, the impact will be huge, and it won't just affect the Moore Technology family." .”

Zhang Jian shook his head, no longer thinking about these things, but immediately started to take action - collecting information to see which companies and institutions were doing the Wensheng video to reproduce Sora.

……

Ji Hong, on his side, made detailed arrangements for the next step in training the growth pilot model.

Model training, especially initial training, is inseparable from human intervention, including data cleaning and calibration, judgment of expected results, and feedback.

The quasi-thinking model has certain key decision-making capabilities, but after all, it is not real thinking. Pure self-training is not enough.

But the early personnel problem is a problem. After communicating with Professor Zhang Mingzhen, it was decided to mobilize 40,000 college students to do these basic tasks.

"Let me tell you, you are right to think so. College students are serious and cheap, especially freshmen. You don't need to pay them. If you add comprehensive test scores to them, they will scream and rush forward. It is easy to mobilize. "Yes." Professor Zhang Mingzhen said pointedly.

The corner of Ji Hong's mouth twitched and he thought to himself: Who do you want to order? Did I say I won't pay?
He also deliberately said: "So good? Then let Professor Zhang discuss this matter. Juaner Intelligent Technology has just started nationwide recruitment and is arranging stations in Beijing, Shanghai, Guangzhou and Shenzhen. The funds are not enough."

"Huh? Recruiting? Why didn't you come to our school to recruit people?" Zhang Mingzhen has forgotten about money. The most important thing is to increase the school's employment rate. To be honest, he doesn't forget about current affairs. School.

"It's the end of May. Those graduates should have already found jobs, right? The rest..." Ji Hong's meaning is obvious. For those who haven't found a job yet, it probably won't be enough to go to Juer Smart Technology:

"And if you really have an idea, you can directly submit your resume for an interview. We recruit fresh graduates. And the company is not far next door. With our reputation in Hezhou University, we don't need to preach or anything. It doesn't make any difference whether you come or not. .”

"Okay, but you have to pay for data calibration!" Professor Zhang Mingzhen emphasized.

"Am I that stingy? Really!" Ji Hong said speechlessly:

"We will also have a relevant budget. Every time we calibrate or give feedback, we can get a certain amount of income based on the quality. If nothing else, we can still earn hundreds or thousands of dollars by just painting a few pictures a month. If we add a The meal problem is not a big deal.

"In addition, I will add another special benefit to them. If they do well, we will evaluate them and they will be given a quota to directly enter Juer Intelligent Technology. Either part-time or full-time positions are available."

"The prerequisite is that your project payment must be made on time." After being teased for a long time, Ji Hong also teased him back.

(End of this chapter)