I Became a Genius Engineer - I Became a Genius Engineer chapter 82

Artificial Intelligence Developer Cheolwoo Lee (6)

Google made the internal decision to move forward the AI event, which was initially expected in early 2016.

“Mr. Martin. Time is running out in the middle of 2015.”

Martin relayed the decision to director Pale Johnson, who heads the AI development lab, and Pale Johnson responded.

“I know. It’s half a year ahead of schedule.”

“… you know, did you make that decision?”

“It was not made by me. Director Nelson of the Strategic Planning Headquarters made the decision, and the CTO approved it.”

“Ha…. Less than a year left.”

“Yes. Yes. Is that possible?”

“If the decision is made at the top, and we ask if it’s possible, what will we do?”

“I’m sorry again for that. But it was an inevitable decision. The NF network created by Chul-Woo Lee from Korea. It’s a really innovative network. Before that network dominated the market, we had to find a way somehow. .”

The NF network was playing a bigger role than I thought.

NF Network as a platform. As a platform business Google, there was no greater competitor than that.

“Is it because of the NF network?”

“Yes. You know, Director Fail, too? How great is that network?”

“I know. It’s a really great technology. It solved the copyright problem, which no one else did, and showed that royalties can be applied to IT technology.”

The NF network was not just playing the role of a platform. It was difficult to collect royalties for IT technology.

In particular, algorithms could be used by anyone if they were made public.

But the NF network has solved that problem.

“Yes. So, LF, who owns the NF Blockchain Foundation…. They are our competitors. We have to show that we have more technological power than them.”

“I know… I know… Artificial intelligence is not as easy as it sounds.”

“What you’re making right now is Go artificial intelligence, right?”

What they are preparing is an artificial intelligence robot that plays Go.

Until now, a computer has never beaten a professional Go player. They were trying to show through it.

Go, which was also called the unique domain of humans… … . If the computer wins that far, and if it is artificial intelligence created by Google, the corporate value will rise vertically.

“Yes, that’s right.”

“Hmm… isn’t there anything difficult? All the notation data is ready.”

“Ugh, I don’t know. Even if you have notational data, it’s not an easy task to create a model that trains it. First, you have to create a deep learning algorithm, and then you have to design a model that is a collection of those algorithms. That’s not the end. Then you have to train the data…”

In fact, in order to create artificial intelligence, it requires a lot of effort, time, money, and even technology.

“It’s not going to be an easy job either.”

“Yes. That’s right. It’s not easy.”

“Then is it impossible?”

“…”

“It could be a matter of life and death for us Google.”

When Martin asked straight-forward, Pale Johnson, director of research at the Artificial Intelligence Development Lab, was deeply troubled.

‘Is it possible?’

The data is ready, and the algorithm has just been created. The remaining task is to design an artificial intelligence model.

‘Well, half a year. To develop TensorFlow and create AlphaGo… … . It’ll be tight, but it won’t be impossible. If you solve just one problem.’

It was a conclusion reached after much thought.

“There is one problem.”

“Yes. What is the problem?”

“It’s study time.”

“Study time?”

“Yes. Artificial intelligence needs to learn, but that takes a long time. With the computer we are using now, it will take about 3 months.”

In order to learn artificial intelligence, a huge amount of computing power was required.

“Three months?”

And Martin, unaware of the fact, asked back as if surprised.

From noble mtl dot com

“Yes. It takes half a year to complete the model and three months to train it.”

“Ummm… Then it will be beyond the scheduled time. It takes time to test.”

Martin didn’t get it right, but he got the gist.

“Yes, that’s right.”

“Then is there a solution?”

“The fastest solution is to use NF networks.”

“……Is that the solution?”

In order to block the NF network, the development of artificial intelligence is in a hurry, but the NF network was needed to hurry.

In this absurd situation, Martin was dumbfounded.

“Yes. NF networks can instantly gain computational power explosively. It is an optimal environment for learning artificial intelligence in a short period of time.”

“Heh heh… is there any other way?”

“Well, speaking of another method, there is a method using a supercomputer… Is there anything left for our company?”

“I’ll think about which method would be better. There will be support for that part in any form.”

“Yes. I see. I’ll have to return my vacation.”

“thank you.”

“No. We also want to show that we, Google, no, we, the United States, have the best technology in the world. At least, shouldn’t we be left behind in IT technology?”

“That’s right. This is not just a problem for Google. How was Lee Cheol-woo when he made the ITE-X engine? At that time, Toyo Motors, which was known to make the best automobile engine, quickly passed over to Korea’s Hyunseong. And Japan’s The economy sank… there’s no guarantee that it won’t be our job.”

Of course, America, which boasts superpowers, would not collapse so easily, but Martin had many worries ahead of him.

“Okay.

“Yes. Thank you. We will spare no support whatsoever.”

“yes.”

“Then please suffer. Thank you.”

An emergency has fallen in Google’s artificial intelligence development lab.

* * *

「Cheongyang red pepper: 6.31 million.」

「Onion: 4.21 million.」

「Paprika: 2.21 million.」

I saw data uploaded to LF-LAB’s NF sharding network.

The data of Cheongyang pepper, onion, and paprika to be used as test models were 6.31 million, 4.21 million, and 2.21 million, respectively. A huge amount of data was being accumulated continuously in real time.

“CEO Lee. How is the development going? Is there anything I can do to help?”

Researcher Jung said.

“Well, I developed a learning algorithm, but now the model is a problem.”

“Model?”

“Yes. Just because there is an algorithm, it is not created. You have to write a learning model.”

“Well, how should I write that.”

“I have something on my mind.”

“thought?”

“I was wondering what it would be like to combine NF network and artificial intelligence.”

“Well, I don’t understand. Is there anything good about combining the two?”

“That’s amazing. First of all, you know the concept that in order to create artificial intelligence, you have to train it, right?”

“Yes. I know.”

“It takes a lot of computational power to do that learning.”

“Is that so?”

“Yes. Even if you think about it with common sense, there are more than 6 million pictures of Cheongyang peppers right now. When will the computer analyze all of this? You have to learn by accumulating data there, so the computing power increases exponentially.”

To put it a little more accurately, creating artificial intelligence was a process of analyzing photo data and accumulating the analyzed data.

Analyzing and scaling photo data required a lot of computer power.

“Um… if it increases exponentially, how much do you really need?”

“I’ll have to try to be precise, but is it about the amount of electricity a small town uses for a month?”

“Heh heh, even a small town would use it tremendously.”

“That’s why I thought of making a system that AI learns into an NF network program.”

“Then, inevitably, a lot of NF is not consumed.”

“Well, that’s right. NF is consumed as much as computer usage.”

“Ohhh. Then how are you going to make it?”

“Hmm, it must feel like an artificial intelligence development tool. The name is…… Let’s just call it Make Intelligence, MI.”

“It’s a development tool that can create artificial intelligence… so are you planning to release it?”

“I’m not sure about that. First of all, I’m going to make it private and decide.”

“Yes. I understand. So, is there anything I can do to help you?”

“hahahaha, yes. Unfortunately, no.”

“Yes. Have a hard time.”

* * *

While making MI, time passed in an instant.

It was mid-January, when 2014 had passed and 2015 had begun.

In the meantime, I worked hard on making MI, and after various trials and errors, I was able to complete it.

AI Development Tool MI.

There is no change in the need to insert data, but it was a tool that could be used if you studied to some extent without a very difficult theory.

「Paprika: 10 million.」

Papkira was chosen as the first test model.

There were many reasons for choosing it, but the important thing was that it was easy to distinguish between paprika and paprika trees.

“Are you finally ready?”

“Yes. It’s finally time to create the first artificial intelligence.”

“How prepared is NF?”

“There are so many NFs. Don’t worry. Learning will probably be completed in an instant.”

“I’m looking forward to it. What the hell is artificial intelligence?”

「NF Network MI deep learning Ready.」

A message popped up on my computer saying that MI deep learning on the NF network was ready.

“hahahaha, let me show you. Let’s start right away.”

“yes.”

Just hit the start learning button.

「Learning — 10%」

The learning system hit 10% in an instant and went up.

「Learning —– 21%」

「Learning —– 32%」

「Learning —– 39%」

「Learning —– 48%」

The gauge continues to rise.

「Learning —– 91%」

Maybe 5 or 10 minutes have passed? The gauge went up one by one and was already over 90%.

“… Um, is it because of my mood that I’m getting slower?”

“hahahaha, that’s normal. The larger the amount of data, the slower it goes.”

“Something is bothering me.”

“Wait.”

「Learning —– 97%」

I kept waiting until it was 97% complete.

“Aren’t you doing this wrong? It doesn’t go up any more than 97%?”

Certainly, the further back you go, the slower the gauge fills up.

“No, it’s not counting. How long has it been since you started studying?”

“It’s almost 20 minutes now.”

“30 minutes. Probably, it will be up to 30 minutes.”

“If it’s an NF network, isn’t there a problem with computing power?”

“Oh, I put a limit per minute just in case.”

It was a device that was set up in case an error occurred and an excessive fee was paid.

「Learning —– 99%」

After waiting another 5 minutes, it went up to 99%.

“99%…”

“How much time has passed?”

“Twenty-eight minutes.”

“Let’s wait two more minutes.”

“yes.”

Researcher Jeong and I anxiously waited for two minutes.

「Learning —– 100%

Learning is 100% complete.

「Learning Completed.」

Immediately after, a green window appeared indicating that learning was complete.