Algorithm system

I’m going to cheat”Ten minutes for you,” Dean Li said to Lin Yuan, holding up one finger.

”This laptop isn’t even his; it’ll take time to get familiar with it,” Professor Zhang immediately retorted. ”At least twenty minutes.”

So Lin Yuan sat down in a corner of the interview room, focused on tinkering with the laptop.

The interviews here were still ongoing.

However, every subsequent interviewee who came in curiously observed Lin Yuan, then whispered to others after leaving, ”It seems that guy really has connections; he can even change code right before the exam,” and ”I can’t believe even a prestigious Zijin Aviation School is so corrupt.”

After nearly twenty minutes, Lin Yuan had finished importing the code and deepening the neural network.

This wasn’t a very complex task, simply because OCR recognition was already quite a mature project, and the assessment time was only three hours.

Therefore, Lin Yuan didn’t need to build an overly complex model, nor did he have time to build one now.

The solution, naturally, was only one—brute force.

And in the AI field, ”brute force” refers to compensating for an insufficient model with more computational power.

As long as one wields the whip of computational power and lashes it fiercely at a model that hasn’t fully reached its limit, better results can always be achieved.

But all of this depends on whether the strength is sufficient.

Relying solely on Professor Liu’s laptop with an RTX 2060 graphics card was clearly not enough power. Normally, one would need to connect an external 4090.

Thus, Lin Yuan had no choice but to summon his computational power system.

’I’m going to cheat!’

—

Cheating also has its principles.

The computational power system is not some incredible power from a wish-fulfillment novel—the kind of invincible golden finger that produces results with a single click.

Lin Yuan couldn’t transfer the computational power of the system to the computer via wired or wireless transmission; the computational power system only provided a calculation result based on the inputted data.

In short, his computational power system still adhered to scientific principles. Lin Yuan had to input the data himself, and he had to read the output results himself.

So Lin Yuan began the data input process.

The training dataset for OCR character recognition, naturally, consisted of various images containing numbers and English letters.

Lin Yuan had discovered early on that the input process could be completed through his own eyes, meaning: after he looked at it once with his eyes, the training data was considered inputted into the computational power system. This process was actually imperceptible to him.

But that was good. At least he didn’t have to insert a data cable into a hole in his body or open a hole in his head.

Lin Yuan quickly switched through images containing characters, and the computational power system’s panel continuously displayed prompt messages: [Inputted!]

As long as a floating prompt message of ”Inputted” appeared on an image, it meant that the image had been collected.

Soon, all datasets were collected. Lin Yuan was secretly relieved; thankfully, these images weren’t very complex, and he quickly completed the input in batches by using thumbnails.

After the training data collection was completed, the next step was the input of the model formulas.

This process was incredibly fast because the model formulas were only a few pages of code. The input method was still through Lin Yuan’s eyes.

And after the formulas were inputted, the computational power system would even display those formulas vividly and flexibly in VR form.

In others’ eyes, they were merely lifeless formulas, but in Lin Yuan’s eyes, they were living, moving spirits.

Take, for example, the process of convolution.

In Lin Yuan’s eyes, it was as if an X-ray machine was scanning data, and a rectangular light spot representing the convolution kernel would rapidly move back and forth, up, down, left, and right over the data content, thereby extracting individual calculated numbers from the data content.

The extracted numbers would then fly up as if they had grown wings, subsequently forming an advanced data content.

[Training data collected!]

[Model loaded!]

[Estimated computational power consumption per 10,000 training iterations: 300T.]

[Current accumulated computational power: 1024P.]

[Please set computational power: __]

[Please set operation time: __]

The relationship between ”computational power” and ”computational power output” displayed by the system is similar to the relationship between energy and power.

Energy = Power * Time.

Thus, Computational Power = Computational Power Output * Time.

The cursor for setting computational power output blinked, and Lin Yuan could input numbers by simply drawing in the air with his finger.

At this moment, Lin Yuan didn’t have much time left for training; the twenty minutes given were almost up.

1P equals 1000T, 1T equals 1000G, 1G equals 1000M, 1M equals 1000K, and 1K equals 1000 operations.

In other words, 1P of computational power equals 10 to the power of 15 calculations.

Of course, Lin Yuan also knew that any method of evaluating computational power that disregarded the complexity of the operation was a sham.

In fact, he wasn’t sure at this moment what 1T of computational power in the computational power system equated to 1T of computational power under which real-world computational power evaluation method. However, Lin Yuan was clear: the computational power system ran fiercely!

When using the computational power system in the past, Lin Yuan generally set it below 50T/S. That is, 50T calculations per second.

Below this value, Lin Yuan’s body would experience almost no side effects, at most he would just need to eat frequently.

Gasoline cars burn gas, electric cars consume electricity. Lin Yuan’s computational power system relies on ”eating.”

It was precisely because of this that Lin Yuan worked so hard delivering food. Almost all the money he earned from delivering food was spent on the consumption of the computational power system.

The accumulated computational power value on the computational power system’s panel merely represented the computational power value Lin Yuan could use, and did not mean that there would be no consumption when using computational power.

It’s like being lent a laptop; that doesn’t mean you can use it without plugging it in.

The accumulated computational power value represents the right to use, while the system’s extraction from the body during operation is the consumption of use.

Previously, Lin Yuan had set the computational power output at most to 1P/S. At this level of power consumption, Lin Yuan would crash instantly. Literally instantly.

Lin Yuan clearly remembered that out of curiosity, he had once set it to 1P/S, and then he instantly lost consciousness. When he woke up again, there were two lines of undried bloodstains beside his nostrils, and his eyes were bloodshot.

In other words, setting the computational power output too high would ”blue screen” himself.

Lin Yuan pulled back his thoughts, gritting his teeth; time was truly pressing.

He drew in the air with his finger.

The system panel displayed characters as he drew.

[Computational power: 300T/S.]

[Operation time: 10S.]

[Please note: The greater the computational power, the greater the burden on your body. Based on past evaluations, under these parameters, you will receive the following debuff: Moderate dizziness lasting 10 minutes.]

Lin Yuan had no choice but to modify the operation time to: [5S.]

The system immediately updated the prompt: [Under these parameters, you will receive the following debuff: Moderate dizziness lasting 5 minutes.]

Lin Yuan was afraid of making a fool of himself on the spot, so he modified the parameters again.

[Computational power: 150T/S.]

[Operation time: 10S.]

[Under these parameters, you will receive the following debuff: Mild dizziness lasting 10 minutes.]

Mild, that’s fine. It should be acceptable.

’Calculate for me!’