Sample: Vengeance No. 1
Wu Chen found himself momentarily distracted.
His sudden silence, however, displeased the other two.
Wang Hongfa rapped on the table. "Wu Chen, go on. But what?"
Lavrov shrugged. "Wang, researchers without creativity are always this dull."
Wang Hongfa agreed once again.
"Here's the thing, Dean Wang."
Wu Chen came back to his senses and ignored Lavrov's jab. He simply gave an offhand reply before considering the Sample Shop feature.
Originally, he had only planned to use the theory as bait to lure them into a quagmire.
But now, if he had a physical sample, it would be even more convincing.
Having made up his mind, Wu Chen spoke again.
"Although SNN theory still needs refinement, I've already asked a classmate's factory to make me a sample. I should have it in a few days."
Wang Hongfa immediately rebuked him. "You made a sample before even perfecting the theory? Is that how you conduct research?"
He thought Wu Chen was trying to get reimbursed.
Lavrov also gave a disdainful snort.
"Even if the theory were complete, what kind of sample could your little workshops here possibly produce?!"
He fired off several rhetorical questions in a row, even forgetting to hide his Russian accent.
Wu Chen, however, was quite confident.
"If you don't believe me, Professor Lavrov, you can wait a day or two and take a look when the sample arrives."
Rovrov shrugged without answering.
Wang Hongfa set his teacup heavily on the table.
He said to Wu Chen, "Wu Chen, you can go back for now. We'll continue the discussion when we have more ideas."
Wu Chen didn't argue.
He simply smiled and said goodbye to them, then walked straight out of Dean Wang Hongfa's office.
He didn't return to his own office either, instead boarding a bus straight home.
On this day in his previous life, the attention mechanism he had proposed had left Wang Hongfa and Lavrov exceptionally pleased. That evening, they had even dragged him to a business karaoke club and taken the opportunity to wheedle all his results out of him.
But this time.
Watching the scenery drift backward beyond the bus window, Wu Chen merely sneered.
Back in his rented room, Wu Chen immediately turned on his computer.
He planned to devise a sample with a sound theoretical basis but no practical value as bait, so that Wang Hongfa and Lavrov would commit themselves to SNN.
His computer, however, really wasn't up to the task.
It took nearly three minutes to finish booting up.
Yet a 360 pop-up in the bottom-right corner was still celebrating: Boot time: 2 minutes 54 seconds. Congratulations, you beat 1% of users nationwide!
When I have time, I'll definitely use the system to get myself a good computer first and make a 2025 graphics card and CPU ahead of time.
After that complaint, Wu Chen focused his mind again.
First, make a sample to chum the waters!
As everyone knew, one of SNN's greatest practical applications was real-time extraction of target features in dynamic scenes.
In layman's terms, that meant identifying fast-moving objects.
The finished product, then, should be a biomimetic vision camera.
The theory it required would therefore have to include, at a minimum:
A spiking neuron model, STDP learning rules, an event-driven encoding mechanism, and a sparse spike sequence decoding algorithm, among others.
After writing up these points of knowledge one by one, Wu Chen began writing the paper needed to produce the sample.
"Spiking neural network architecture for dynamic vision: theoretical exploration and prototype implementation."
Because he had studied this field in depth in his previous life, Wu Chen wrote quickly.
Before he knew it, a day had quietly passed.
Wu Chen finally finished the paper he would use as bait.
Not only did the paper head in the wrong direction, it also contained a great deal of unverifiable data. Wang Hongfa and Lavrov certainly wouldn't notice.
After checking it over again, he uploaded the finished paper to the system without hesitation.
[Paper received: "Spiking neural network architecture for dynamic vision: theoretical exploration and prototype implementation"] [Evaluating] [Evaluation complete] [The theoretical framework provided by the host has major flaws. An incomplete sample can be created. Continue?]
"Create it!"
[Sample created]
Soon, a rectangular camera icon appeared in the Sample Shop. Opening it revealed some specific details.
[Unnamed sample one] [Grade: white] [Note: grades from lowest to highest are white, green, blue, purple, and gold] [Price: 1 Skill Point] [Intelligent dynamic-vision processing camera] [Advantages: based on a spiking neural network (SNN) architecture, it can demonstrate excellent low power consumption and dynamic target tracking capabilities in an environment with constant lighting, a clean background, and target objects moving at a uniform speed] [Major flaws: extremely poor environmental adaptability; interference resistance approximately zero] [Name it now?]
After thinking for a moment, Wu Chen spoke decisively.
"Name it Vengeance No. 1!"
[Vengeance No. 1 named] [Spend 1 Skill Point to extract it into the real world now?]
"Yes!"
[Insufficient Skill Points. Extraction failed]
"Fuck."
Wu Chen was quite puzzled.
He had no choice but to check the ways to obtain Skill Points again.
Publishing a paper would take about a month, even through the fast-track channel of a core Chinese journal.
There were faster options, of course, but Wu Chen didn't want any academic stains on his record after getting a second life.
Completing regular teaching duties would have to wait until the semester started, at the earliest.
As for solving a practical engineering problem, he currently had no opportunity to work on a project, so that wouldn't work either.
He had plenty of ways to monetize technology in his head, but he couldn't do it in a short time.
Theoretical verification...
Wu Chen hurriedly checked the details.
[Propose and verify a technological theory unpublished at the current point in time]
That would work!
Wu Chen thought for a moment, his gaze falling on his old laptop.
It was a Lenovo Tianyi F41M.
The CPU was an Intel Core 2 Duo T5750, and the memory had originally been only 2 GB, which Wu Chen had upgraded to 4 GB. The graphics card was an 8400M GS!
In 2025, no one would want this thing even if it were free.
Even in 2008, it was a mid-to-low-end machine.
But it could still run the simplest theoretical model!
As for which model to choose, Wu Chen thought it over for a while and soon made a decision.
Since the pit he had dug for Wang Hongfa and Lavrov was an image recognition scheme that couldn't be put into practice, he would use one that could!
He quickly wrote down the title of his second paper.
"Image super-resolution reconstruction based on convolutional neural networks."
This theory hadn't been perfected until 2014, making it well suited to the system's requirements.
Once he had settled on a direction, Wu Chen immediately began writing and running the verification code while working on the corresponding paper.
With a brain more than a decade younger and his wealth of experience, Wu Chen found himself in excellent working form. It was as though he had returned to his peak in high school, with every point of knowledge and example in the paper coming readily to hand.
Time ticked by, second by second.
As the sun rose on the horizon and set again, Wu Chen finally completed the verification.
Although he was a reborn person, he had still overlooked one problem.
The 8400M GS didn't support CUDA.
He had actually waited three days for this convolutional neural network model to run!
During those three days, he had barely slept, apart from taking time to gnaw a few bites of bread.
Fortunately, the program finally ran successfully.
It output an image enlarged fourfold, with clarity and detail retention far superior to those of the traditional bicubic interpolation algorithm.
It worked!
This theoretical achievement, which should have appeared in 2014, had now emerged six years early!
After checking it over once more with bloodshot eyes, Wu Chen finally spoke in a low voice.
"System, submit the theoretical verification results!"
[Ding] [Theory of image super-resolution reconstruction based on convolutional neural networks verified] [Sample cannot be generated because the theory is incomplete] [Skill Points awarded: 5] [Note: Skill Points can be used to increase levels or purchase samples] [Currently available production blueprint: Vengeance No. 1] [Added to Sample Shop: Vengeance No. 1] [Congratulations, host, on obtaining Skill Points for the first time. You have received one chance to draw a prize!]