I Became a Genius Engineer - I Became a Genius Engineer chapter 86
Artificial Intelligence Developer Cheolwoo Lee (10)
To find out what the situation was, Changsu and I returned to the NF Foundation.
“Hello. CEO Lee. It’s been a while.”
When we entered the NF Foundation, team leader Hyuk Cho greeted us.
“I heard that there is work. They say that the utilization rate is close to 100%?”
“Yes. That’s right. Almost all resources are drawn from the contract at a specific address.”
“You don’t know who the owner of that address is?”
“Yes. We had no way to confirm.”
hmm… … . Who the hell is using NF networks and why?
“What’s the address? I’ll have to check it myself.”
“Yes. Wait a minute. The address is 0x21FA……F2974.”
I searched for the address that Team Manager Jo called and looked for various materials.
As expected, it was not possible to confirm which programs were uploaded because they were private. However, the NF they had could be confirmed.
「610,212.17808172」
600,000 NF. At current market prices, he was holding about 6 billion won. 6 billion. Actually, it wasn’t that big, but it was enough to borrow the computing power of the NF network for a few hours.
After looking into it a little more, it’s been running for about an hour.
I used from 750,000 NF to 600,000 NF in one hour. I was using about 50,000 NF more than I thought.
“Um, I think it’ll be back to normal in a few hours, but I’m very uncomfortable. I need to find out a little more. Manager Joe, please contact Bitsome. I need to check what kind of calculations I’m doing.”
“I don’t know if Bitsome has any data, but I’ll check it out.”
“yes.”
“Ah, but you know what kind of calculations are you doing?”
“yes.”
“……? Is that possible?”
“Um, isn’t it possible? Even if you can’t see the source, after all, the calculation is done on the computer that lends NF.”
“Isn’t that the realm of complete reversing? Isn’t it to find out which program is running by analyzing assembly language?”
Computer languages fall into two categories. high-level language and machine language.
High-level languages are programming languages that are easy for people to understand, such as C++ and JAVA. When people develop in that language and compile it, it is translated into machine language.
Is machine language supposed to be a language that computers can understand? Anyway, it’s hard for people to understand.
“Well, in a way, it must be similar. Flyer language is a high-level language, and if you compile it and upload it to the NF network, it will be converted into machine language that runs on the blockchain virtual machine.”
“Uh… is that what you mean by analyzing the machine language?”
From noble mtl dot com
“yes that’s right.”
“……? Is that possible?”
“I made it, nothing is impossible.”
“It doesn’t make sense, then does the person who made assembly language fully understand machine language? It’s nonsense that the person who made soccer is good at soccer.”
“Ha ha ha… I see. Even if you’re a developer who made a game, being good at that game is another thing.”
“I see.”
“Well, I guess I’m a freak.”
“but.”
“Why! I agree.”
“Oh, no. Then I’ll go to Bitsome to find out!”
Team leader Jo left my room in an instant.
‘Well, if the utilization rate is almost 100%, it will be borrowing computing power from our lab’s server.’
The server that is mining NF in labs is probably also lending computing power to the address in question. I went right into that server and dumped the memory.
「call 009A1316
mov dword ptr [ebp-8],1
mov dword ptr [ebp-14h],2
… … .」
When I dumped the memory in use and opened it with a reversing tool, the assembly code was output.
‘Well, I did a mov at ebp-8… … .’
Although there was a lot of data to analyze, that is, assembly code, the analysis continued without difficulty.
‘uh? Is this a bit strange? Why does it keep repeating itself?’
Continuing the analysis, a few oddities were discovered. First, the point that OUTPUT goes into INPUT.
This code continued to run in a recursive function and produced some result.
‘what… … ? Isn’t this a function of the deep learning algorithm?’
As I continued to analyze the function, I knew what it was doing.
deep learning algorithm.
In other words, it was an operation to learn artificial intelligence. with reinforcement learning.
And one event that comes to mind.
‘… … . uh, maybe? AlphaGo?’
Today’s AlphaGo, it seemed that the artificial intelligence was doing reinforcement learning through the NF network.
To be a little clearer, I analyzed the INPUT data.
‘If my prediction is correct, this must be notational data.’
A notation that records the game of Go. With that, AlphaGo would have learned. And even now, you will be building notational data on your own, learning, and repeating.
The data was converted to conform to the rules of notation.
OK.
“P4, D16, P16, D3… … .」
The values of the data estimated as coordinates were written.
‘hahahahaha, isn’t it interesting?’
I don’t know how it is in English, but AlphaGo, which played against 9th Dan Lee Soo-dol today, was conducting reinforcement learning through the NF network.
In the meantime, the door to my room opened, and Team Leader Joe entered.
“CEO Lee. I contacted Bitsome, but they said there is no user information deposited to that address.”
“Um, that’s right.”
“Has the representative found out anything?”
“I just found out a little bit of fun.”
“What facts?”
“It could be Google’s AlphaGo that is using the computational power right now.”
“Yes? Google’s AlphaGo is using it?”
“Yes. I don’t know how far you know artificial intelligence, but Google’s AlphaGo is currently trying reinforcement learning with the computational power of the NF network. No, it’s doing it.”
“……What do you mean by that?”
“It is literally.”
“No, didn’t Google say they wouldn’t use our NF network? But what is that…”
That’s why I thought it was funny. Google had declared that it would not participate in our NF network.
But just because they are in a hurry, they use the NF network?
It was a complete contradiction.
“So, isn’t it an interesting fact? I’ve declared that I won’t use it, but I’m using it.”
“Heh heh…… So tomorrow’s AlphaGo can show a slightly different appearance?”
“Hmm, I think that’s probably the case. Lee Soo-dol, 9th Dan, might be at a disadvantage.”
“Tomorrow will be fun.”
“Let’s watch. Oh, and just in case, please organize the data of this case.”
It could be a good weapon if Google rebels against the NF network.
“Okay. I’ll try to organize the data.”
* * *
The next day, the second station of the Google Deep Mind Challenge began.
This time, the four of us watched the live broadcast: Cheol-Woo Lee, Min-Seok Cha, Team Leader Cho, and Researcher Jeong.
Team leader Jo thought before the match started.
‘Hmm, Google’s AlphaGo really tried reinforcement learning?’
Team leader Jo couldn’t believe what Lee Chul-woo said yesterday.
‘Google is not a hole-in-the-wall shop, but a global company. It is also a company in the top 10 by market capitalization, but behind the scenes, they are using NF Network?’
Google has already nailed down its decision not to participate in the NF Network. But I couldn’t believe that he had said that and used it behind his back.
‘Couldn’t Cheol-woo have misunderstood something?’
Joe couldn’t help but think so.
No matter how talented Cheol-Woo Lee was, analyzing assembly language was not an easy task.
Assembly language is not a human language, it is a machine language.
If it was deciphered in such a short amount of time… … .
‘Is it a machine, not a person?’
He couldn’t help but think so. I say machine language because it is difficult for humans to understand, but machine language is also easy to understand.
Isn’t that like a machine?
‘Well, I’ll have to watch this game for now. If the real AlphaGo had reinforcement learning, as Cheol-woo said, it would show something different in today’s match.’
He focused on broadcasting.
-hello. This is Im So-yeon, the host of the Deep Mind Challenge. The previous day, AlphaGo showed a somewhat disappointing appearance due to Lee Soo-dol, 9th dan. Indeed, I am looking forward to what the results will be today.
-hahahahaha, then, does caster Soyeon want Lee Soo-dol, 9th Dan, to lose?
– Could that be? Still, I’m rooting for Lee Soo-dol’s 9th Dan.
– The match begins. Let’s meet right away.
After a while, the match began.
Number 112. Lee Soo-dol’s 9th dan’s expression was stained in a bad way.
-Soodol Lee’s 9th Dan expression has gotten worse. Commentator, did the move just change the flow of the game?
-ah… … . Just number 112. I think it might be AlphaGo’s mistake.
-A mistake? But why doesn’t Lee Soo-dol’s 9th-dan expression look good?
-According to my guess, I think it’s because there are numbers that are not in the calculation.
– Well, I’ll keep an eye on it a little longer.
The great power continued.
165 numbers.
-Ah, the countdown of Lee Soo-dol’s 9th dan has begun. I don’t have much time now. Commentator: Why did the situation turn out this way?
-112 numbers. That number, which we knew as AlphaGo’s mistake, was the key. Because of that number, Lee Soo-dol is now at a disadvantage.
-Are you talking about that mistake?
-yes. Number 165 right now, I can’t believe I calculated the number after seeing this.
-Is it very different from the previous day?
-That’s right. The previous day, I analyzed it by looking ahead at no more than 20 moves, but today is a little different. 50 count? 100 count? AlphaGo seems to be looking ahead.
-Uh, what do you think about AlphaGo’s sudden improvement?
-well. Maybe Google’s AlphaGo was hiding its skills?
210 numbers. Lee Soo-dol, 9th Dan, threw a stone.
– The game is over. The defeat of Lee Soo-dol, 9th Dan.
– This is shocking news.
-yes. That’s right. Google Deep Mind Challenge 1 Win 1 Lose. Where will the outcome of the match go? In the midst of a considerable shock, people’s attention is focused.
The second round of the Google Deep Mind Challenge has come to an end.
‘What, the real AlphaGo has been strengthened? Then, can Cheol-woo even analyze the real machine language?’
In fact, Team Leader Cho thought after seeing how AlphaGo had been greatly strengthened.
“Huh, the real AlphaGo wins.”
Minseok Cha said.
“What did I say, did I say we won today? Yesterday’s network usage was used by Google’s AlphaGo.”
Cheolwoo Lee said.
“No, they don’t have any faith? You said you wouldn’t participate in the NF network, but then you used it?”
“There’s a difference between using it and participating.”
“That…… CEO Lee, did AlphaGo really use it? The computing power of the NF network?”
“Yes. You said yes yesterday.”
“Oh, I can’t believe it.”
“What?”
“No, you deciphered the machine language in that short time. That’s nonsense. CEO Lee isn’t even a machine…”
While Cha Min-seok and Lee Chul-woo were talking, team leader Jo interjected.
“hahahaha, well. You must be talented.”
Hearing Cheol-woo’s last words, Team Leader Jo thought.
‘Is that going to be the realm of talent? Is it just a machine?’
Team leader Joe got goosebumps at his talent for deciphering machine language.
‘Really, definitely a genius is a genius.’
For a while, he forgot what kind of person Lee Chul-woo was.