Reborn in 2010, the AI ​​in my mind is too powerful.

In the afternoon, Chen Yiran took a nap, something she hadn't done in a long time. It was really great not having to go to class.

After getting up, I first checked GitHub, specifically the web scraping projects under the chen1ran account. Last time I checked, it had over 600 stars, but now it has over 1200. I found quite a few repositories; someone modified a Reddit version, and a Japanese person added multilingual support, including Japanese and Korean.

The issue section is full of bug reports, feature requests, and questions about why proxies aren't supported—you name it.

I scrolled through my private messages again and found a few from strangers. One was from a Brit who wanted to do data analysis and discuss the commercialization of web scraping. Another email address was from someone named "Y Combinator," asking for coffee.

I called out Jasmine's name in my mind.

【exist. 】

"Let's begin with the framework you described."

[First, catch up on your studies. The CNN section in CS229 only covered the concepts, not delved into them. To build a framework, you need to understand the architectural details of CNNs first.]

"Didn't you finish learning CS229?"

[The course is finished, but CNNs aren't the focus. SVMs, logistic regression, and clustering are the core concepts. CNNs were only briefly mentioned at the end, and twenty minutes of that went by. To build a framework capable of image recognition, simply knowing what convolution is isn't enough.]

"Alright, tell me what else you need to learn."

[In CS223A, Khatib's lectures on feature extraction and image filtering are directly related to CNNs, so finish that part first. Then read a few papers, including LeNet-5, and how existing CNN architectures are designed. Also, study Theano's source code to see how it implements automatic differentiation.]

"How long will this take? It seems like there's a lot of stuff."

[Now that there are no classes, I can catch up in two or three days and then start writing.]

"OK."

The CS223A lecture notes opened on the screen, and Chen Yiran skillfully began to turn the pages.

[No need for manual page turning; I can write a script to enable this function in programs that cannot automatically turn pages.]

"Oh, this is great news! Let's hurry!"

Chen Yiran's fingers began to move up and down, and following Jasmine's request, he wrote a program.

In just over twenty minutes, the script for a small function was written.

[It's finished. Open the PDF and try it out.]

Chen Yiran switched back to the CS223A lecture notes and ran a script in the terminal. The page started scrolling down automatically, at a moderate speed.

[That's it. You just need to keep your eyes on the screen.]

Chen Yiran leaned back in her chair, her eyes fixed on the screen. Great, now her hands are free. She just needs a brain-computer interface to free her brain.

Around 3 PM, the visual part of CS223A was finished.

Now let's look at LeNet-5. Try searching for it.

Chen Yiran started searching, and after a long time, finally chose to read Yann LeCun's paper.

It has to be said that the CS/ML field has a strong culture of sharing; the author simply puts the paper's PDF on his personal homepage, and you can download it directly without paying.

If this were in fields like medicine or law, it would easily cost tens of dollars.

I found a classic paper from 1998 and downloaded the PDF. Then I opened the automatic page-turning script to take over.

Twenty minutes later, Jasmine had memorized it all.

"Is that enough? What else do you need to see?"

[Search for Theano's source code.]

This one isn't hard to find, and it's much denser than academic papers; you can finish reading it in just over ten minutes.

Okay, now you can prepare some food.

Chen Yiran sighed, went to the refrigerator and took out a can of chocolate ice cream, cooling down her body in one go.

With a voice that seemed to be bravely facing certain death, he whispered, "Jasmine, start! Go ahead and do it without worry. You don't need to go to class, and you don't need to worry about whether you can get up tomorrow or whether you're in good shape."

【OK. 】

At first it was fine, but after twenty minutes, Chen Yiran felt like she was going to die.

In order to focus on teaching, he has basically not let Jasmine do any large-scale reasoning exercises lately. Jasmine can now conduct reasoning exercises at the lowest level according to Chen Yiran's physical condition, ensuring that it will not affect Chen Yiran's health.

I took a few big bites of ice cream, and the icy stimulation helped to alleviate some of the discomfort.

Perhaps Jasmine's previous minimal learning had some effect; Chen Yiran felt that her resistance had improved a lot, and now she knew she was very uncomfortable, but she could still persevere.

Chen Yiran lay there for three hours before Jasmine finally gave a response.

The architecture of LeNet-5 is clear. It consists of five convolutional pooling layers and two fully connected layers, taking a 32x32 grayscale image as input and outputting ten categories. Its automatic differentiation uses symbolic differentiation. It first builds a computation graph, then performs symbolic differentiation on the graph, and finally compiles the differentiated graph into C code for execution. I don't think it's very practical. The disadvantage of symbolic differentiation is the difficulty in debugging and the incomprehensible error messages. I have another idea: using numerical differentiation with caching.

Chen Yiran waved his hand: "Okay, that's good enough. You definitely need to rest today. If you're in good form tomorrow, then play it out."

【OK. 】

After watching all day, Chen Yiran rubbed her neck and stood up to stretch.

It was already dark outside. I glanced at my phone; it was 7 PM.

He declined Xu Li's invitation to go out to dinner via text message. He was too lazy to move at the moment. After ordering two pizzas, he started browsing the web.

I opened BitcoinTalk again. It's a forum Satoshi Nakamoto created last November, hosted on bitcoin.org. It has a little over a hundred registered users, but never more than ten are ever online. Most posts are technical updates from Nakamoto, with only a few replies.

I entered the "Economy" section.

The pinned post, published by NewLibertyStandard, is titled "1,309.03 BTC = $1.00 USD". The post states that it calculated an exchange rate based on the electricity cost of mining and offers buying and selling services. Most of the replies are along the lines of "How could this possibly have any value?"

There's a post on the second page, posted two days ago by an ID named "dust":

What can you buy with 1 coins?

"I mined about 10,000 coins, and my graphics card ran it for over a month. Can this stuff really be exchanged for money? Has anyone actually used this to exchange for anything?"

The following are eight replies.

"No. This thing is worthless."

"You can exchange dollars at NewLibertyStandard, but the exchange rate is too low; 10,000 coins only get you a few dollars."

"Isn't this the same as using an idle graphics card for gaming and supporting scientific research?"

"At least you're helping NASA find aliens, who are you helping with this?"

"The person upstairs is right, this thing has no practical use. At least it's an aid to protein folding research, but what is this, a hash value?"

"Whatever it is, the graphics card is just sitting idle anyway."

"Has anyone tried trading on the forum? I'm willing to accept USD with PayPal and sell BTC."

"PayPal transactions are risky. The other party can request a refund after receiving the funds, so please be careful."

Chen Yiran stared at the post for a while.

He registered on this forum several days ago but hasn't posted anything yet.

A reply was sent below dust:

"DM me your address, and I'll order two pizzas for you."

One minute later, dust replied:

"??? Seriously?"

Chen Yiran sent a private message.

"Seriously. Send me your delivery address, and I'll have two Papa John's deliver to your house. All you have to do is give me 10,000 coins."

Five minutes later.

"Hahaha, are you serious? Ten thousand coins for two pizzas?"

"Yeah. Anyway, you have nowhere to spend those coins, so why not buy yourself a meal?"

"Alright, you actually agreed. Here's the address. The wallet address is 1Dust1a7xxxxxxxxxxxxxxxxxxxxx. Order your pizza first, and I'll transfer the coins to you once I receive it."

Chen Yiran opened Papa John's. She ordered two large sausage pizzas, plus delivery and tip, for $27.

The order confirmation email arrived quickly. I sent a screenshot to dust.

"It's ordered. It should arrive in about 40 to 50 minutes."

Three minutes later.

"Holy crap, you actually yelled."

"Okay. Let me know when you arrive."

A while later, dust sent me a private message.

"The pizza arrived, just signed for, and it's still hot. Awesome. My roommate was stunned."

Chen Yiran opened the Bitcoin client. He pasted the wallet address provided by dust. Transfer amount: 10,000.

I clicked "Confirm" and entered my password.

A progress bar pops up. The transaction is broadcast to the network.

Ten minutes later, a block was confirmed.

Refresh your wallet. Your balance has increased from 0 to 10,000.

Open BitcoinTalk and see that dust updated a post:

"Update: Someone actually used 10,000 coins to buy two pizzas. Just finished eating them. This stuff actually works! My roommate said our graphics card wasn't wasted after all."

People started responding enthusiastically.

That's good. No matter how well I do in this life, at least I have a safety net.

I also need to take some time to back up my wallet, physically isolate it, and find a safe place to store it.

It would cost a lot of money to do AI or brain-computer interfaces, especially since he is Chinese.

It's hard to say whether he will be restricted or not. If he displays too much strength, it might lead to a series of restrictions imposed on China ahead of time.

Staying in the United States indefinitely is not very realistic.

It's hard to say about others, but his parents are definitely beneficiaries of China's existing system. Their education was funded by the government, and later they worked for state-owned enterprises, even profiting from real estate.

He can't get over the thought of just walking away, so he needs to be prepared beforehand.

Now at least I have a safety net. Even if I never achieve anything in this life, I can always become a billionaire.