I Became a Genius Engineer - I Became a Genius Engineer chapter 83
Artificial Intelligence Developer Cheolwoo Lee (7)
「Learning Completed.」
A green window indicating that learning was completed was floating brightly.
“……This is what happened, right?”
“Yes. It seems so.”
Researcher Jeong and I were at a loss. What is artificial intelligence really? I made it according to the calculation and as expected, but it was the first time I actually made it.
“Now, how do I do this?”
“In fact, artificial intelligence is not a big deal right now.”
“which?”
“Put in a picture, and the data will be extracted. Anybody working on that LFH?”
“Are you talking about putting a box on an image?”
What computers can’t do. It was to look at the picture and hit the box on the harvest.
“yes.”
“Are you saying that a computer can do that now?”
“I will. Please wait a moment.”
I found an artificial intelligence core on the NF network.
「Test_Model.dlp」
The name is test model, and the extension is dlp. It was the core of artificial intelligence that had completed learning.
“What file is this?”
Researcher Jeong looked at the file and said with a curious expression.
“This is the core file of artificial intelligence. It contains the results of training data. Hmm… To be more precise, it is a formula.”
“formula?”
“Yes. It contains learned formulas. Let’s see… The capacity is about 2㎆.”
“The formula is 2㎆? That’s the formula I think, right? It’s used in mathematics.”
“yes.”
“…… Isn’t that a very large capacity? 2㎆ by formula alone.”
It was a larger capacity than I thought.
Formulas are simply text data. If the capacity is 2㎆, would it be about the size of a book?
“Yes. That’s right. If you write down the resulting formula of the learned data into a book, it will be about one book long.”
“Wow, crazy… just one book of formulas is huge, isn’t it? Doesn’t that increase computer usage?”
“Well, does a book’s worth of multiplication and division make my computer heavy?”
“Well, that’s what you think, so it’s like that again. I thought the computer’s throughput was too easy.”
“I’ll try it. It’s clearer to try it than to explain it in words.”
“yes.”
I did a few simple things right away with the test core.
It was a program that automatically recognizes paprika by inserting an image with paprika, and the formula of the artificial intelligence core works.
“I’m ready.”
From noble mtl dot com
“It’s over soon. What should I do now?”
“Do you have this image right now? I’m going to put this in the program over there and run it.”
He said while showing an image of paprika.
Of course, it was the same as the original image. The image has a green box or not.
“Oh oh…….”
“Then I’ll put it right away.”
「Completed.」
When I put the original image in, a message popped up saying it was done in an instant, and the program created a new file.
「Test_img.jpg」
“It’s done. I’ll open it right away.”
“I’m looking forward to seeing what kind of images artificial intelligence has created.”
When I opened the file, an image of paprika appeared, with a green line drawn along the perimeter of the paprika.
“…?”
“oh.”
Researcher Jeong didn’t know what kind of English it was, and I was a little surprised at the more accurate performance than expected.
“No, it’s not that the box is lined with paprika, but there’s a line drawn along the outside of the paprika?”
“The performance is better than expected.”
“No, obviously, the training data is a box, but why does this result come out? Shouldn’t an image with a box drawn in the same way come out…? No, it’s not like that…. Even if you just draw a box on the image now, it’s amazing.”
“hahahahaha, do you know how many pieces of data were trained?”
“10 million copies?”
“Yes. That’s about it. Each image will have a different angle, size, shape, etc.”
“Yes. Isn’t it?”
“Artificial intelligence has learned. The result of aggregating variables such as various angles, sizes, and shapes is directed toward paprika, so it is recognizing paprika.”
“…… Crazy. So now the computer can accurately recognize paprika?”
“Okay. Shall we test one?”
I picked up a paprika next to the desk.
“When did you get that paprika?”
“I think you’ll need it. Wait a minute. Let’s make it real-time, not an image this time.”
“…?”
Researcher Jung was surprised or not, but I did the work.
It was a program that recognizes paprika in real time from a smartphone image. Of course, it was a program based on the NF network.
There will be some NF consumption, but… … . What do you want to do? It’s not like you eat that much.
“I’m ready. I’ll do it right away.”
“…”
Leaving behind Researcher Jeong, who had become mute after eating honey, I ran the program I just made on my smartphone.
Then, the camera function was activated on the screen of the smartphone.
“Is it really going to be good…”
While worrying about that, I took a picture of the paprika I was holding in my hand on the screen of my smartphone.
OK.
“and…….”
On the smartphone screen, a green line was drawn along the outline of the paprika in my hand.
“Hey, that’s fine.”
I turned the angle around and reported the distance between the smartphone and the paprika, but there was still a green line drawn on the outer line of the paprika on the smartphone screen.
“…… Crazy. Isn’t this a real computer recognizing paprika?”
“That’s right. That’s why we’re making artificial intelligence.”
“No, I only heard about it. Seeing it in action like this… I’m sure, until recently, I couldn’t do this work, so I hired someone and they drew paprika in a green box, didn’t I?”
“Yes. It was.”
“But… now that computers can recognize it so precisely, no… If it’s paprika, doesn’t it recognize it better than grandma or grandpa?”
“hahahahaha, it could be.”
In fact, the computer was clearly recognizing paprika.
“It’s amazing. That a computer can recognize objects…”
“This is just the beginning. There are many things to learn. In the future, you need to learn the tools to be harvested, and you have to learn how to fly according to the height. It’s not only that. It has to be trained so that it can be done automatically. It’s not just one or two things to do.”
“Heh heh…… Then aren’t drones more intelligent than animals?”
“hahahahaha, compared to animals, it’s a bit like that. Because that’s the realm of strong artificial intelligence.”
“Strong artificial intelligence?”
An intelligent dog can understand human speech. Learning is possible through words and actions. But now artificial intelligence is not.
Even if I did, it would be to the point of covering my eyes and screaming.
Only special-purpose learning was possible.
Currently, artificial intelligence has not been able to learn what it sees or feels on its own.
“There is such a thing. Look for it later if you are interested.”
“yes.”
“I’ll be busier in the future. I’ll have a lot of data to request from the LFH.”
“Wasn’t this the end?”
“I just heard something. It’s not just one or two things to learn, and you need data to learn that. You have to make a carrier and there’s a lot of work to do.”
“hahahahaha, I’m sorry. I was so surprised.”
In fact, Researcher Jeong was still staring at the paprika recognized by my smartphone, wondering what was so strange.
“Go to work now! I’m busy.”
“all right.”
* * *
Time continued to pass quickly. It seems like 2015 has just begun, but it has already passed March and is now April.
Successfully created paprika recognition artificial intelligence, and made many things here and there.
First, make harvesting tools for drones to use.
The job was easy. Through the free vector gear, I made a gadget arm that moves freely, and made it possible to use scissors depending on the crop.
What I did after that was to make a carrier.
It was a basket containing the harvested crops. There were no special features. It was just a matter of attaching a sensor and letting the drone recognize that it is enough to put the harvested crops here.
The problem of charging the drone was not difficult either. It was not learned by artificial intelligence, and it was designed to go to a charging station and charge if there is no battery by the internal battery sensor.
And the last thing left was to learn to fly the drone.
It was difficult in many ways. First of all, there is no normalized data like paprika. The user could not directly learn the signal to fly the drone.
Because every farm environment is different.
It was not possible to learn by carrier signal strength or the like. In the end, I had to learn all the overall images, and I had to judge whether I could pass or not in those images.
‘Well, on the ground it might be more troublesome.’
I thought about it while making it. If you make it on the ground, you will have to think about how to avoid obstacles again, but the drone will be able to solve it a little easier by adding coordinates called height.
Of course, it is true that complexity increased with the addition of a variable called height.
‘Well, forget about the past… … . Now how do I learn this?’
The trouble deepened.
First, there were too many types of obstacles. In the case of a facility house, there were many steel structures around it. I had to avoid all the structures, and there were many obstacles made of plastic as well as steel.
Not only that, but you also have to avoid crops.
“CEO Lee. CEO Cha is here.”
Meanwhile, Minseok arrived.
“Yes. Please come in.”
After a while, Minseok came into my room.
“How are you? Lee Cheol-woo?”
“hahahaha, what am I doing?”
“Aren’t you always doing research? I made a studio for you, but it didn’t come out once.”
“What’s the need to go out? We eat and sleep here, but there’s no such thing as heaven.”
“Let’s get some fresh air.”
“I still go to LFH often.”
“…… What kind of air is that? You’re going for work.”
“It’s quiet and nice though, over there.”
“I’m not going to tell you, I don’t know where the old Lee Cheol-woo went. They always said it was annoying and that it wasn’t fun.”
“Because I have work to do now.”
“…”
“So why did you come? I’m sure you didn’t come to see me.”
If I had asked to see her face, I would have called her outside. Perhaps, the fact that he came here to the lab was because he had other ulterior motives.
“Google is starting to move.”
Minseok got right to the point.
“What about Google?”
“uh.”
“How? You started moving?”
“August of this year. They said they were planning an event.”
“event?”
“Yes. It’s called DeepMind Challenge. It seems like they developed Go artificial intelligence.”
Go artificial intelligence?
“Is this correct information? Where did you hear it?”
“Didn’t I say there was a teacher I used to attend? That’s reliable information.”
“I need to see that master, too.”
“hahahaha, I’ll come with you sometime.”
“Uh, anyway, making a Go artificial intelligence?”
“Uh. I guess they’re going to play against a professional Go player with that Go artificial intelligence. And they’re trying to show off their technological prowess. The corporate value will go up as well.”
“… sure. That’s right.”
“But is that possible? A computer beating a professional Go player at Go?”
“It’s not impossible. I can make it now too?”
It was easy. If you give MI a huge amount of notational data and learn it, you can easily make it.
“Wow, that’s great too.”
“no.”
“Only Cheol-woo believes in that. Anyway, I came here because I wanted to know what to do about it. I wonder if we should take some action as well.”
Harvesting robots need to be trained to fly, but there is plenty of time in August. You can make it until then.
But is it okay to show off our technological prowess against Google?
“Uhm… it’s a problem I’m worried about.”