Intelligence level 0 -> level 1

"You mean it killed your research career, right?"

Jiang Nan apparently didn't have much of a sense of humor.

"Right."

Of course, Wu Chen couldn't tell anyone about his rebirth.

After a while, Jiang Nan asked again, "Then do you have a better solution for the artificial intelligence pill inspection system my father wants?"

Wu Chen had been staring absentmindedly at the scenery outside the window. When he heard her question, he answered without thinking, "Yes, dear, I do."

Jiang Nan frowned.

Wu Chen realized what he'd said. He'd been taking so many orders lately that he couldn't shake the habit of talking that way.

"Sorry, I was distracted."

Wu Chen sat up again before answering seriously.

"I already said this afternoon that SNN isn't currently suitable for commercial use, but a HOG feature extraction algorithm combined with an SVM classifier would work. It can not only extract pill information accurately, but also use CLAHE adaptive illumination compensation to suppress reflections from the coating. Theoretically, the recognition accuracy can reach over 99.5%... blah, blah, blah."

He briefly explained the two sets of theories.

After listening, Jiang Nan looked momentarily bewildered. After all, changing fields was like crossing mountains and seas.

Finally, she said, "My best friend works in machine vision at Peking University's computer science school. I'll consult her. If it works, I hope we can reach an agreement to cooperate."

Wu Chen nodded, unconcerned about whether she was merely painting him a pie in the sky.

"Actually, I've already submitted a related paper to PAMI, but their review cycle is three to six months. Once it's accepted, I can show it to her."

Jiang Nan glanced at him in surprise. "A top journal?"

Wu Chen nodded.

Only then did Jiang Nan understand why he wasn't afraid of Wang Hongfa retaliating.

If he could produce results like that, then at Xiaoheshan University of Technology, Wu Chen really didn't have to fear a dean.

But was Wu Chen actually capable of it?

Jiang Nan didn't feel it was her place to ask more, so she simply said, "Do Dean Wang and the others know about this?"

"No."

Under Wu Chen's original plan, Wang Hongfa would have been mired in the SNN quagmire by the time the paper was published. Then he would have been happy that Wu Chen had researched another direction instead.

But their falling out today had thrown things off course, and he might have to lay his plans anew.

Jiang Nan turned to look at Wu Chen again, at a loss for words for a moment.

If she and Wu Chen hadn't been there today, her father might really have been swindled out of a huge sum of money.

And the man beside her was the one who had started it all.

It gave her a sense that the world was rather fantastical.

Of course, if she knew that Wu Chen was also the one who had restored her grandmother's photo, that feeling would probably have grown stronger.

But for now, all she felt was relief.

"Then I'll wait for your good news, Teacher Wu. Also, if possible, I hope you can make a sample for testing."

"No problem."

Wu Chen said.

What the Jiangdong Pharmaceutical Factory needed happened to be the industrial-grade intelligent visual inspection camera he had previously drawn from the system.

A sample was in the system shop, and he could redeem it whenever he wanted.

If he had enough skill points, he could even redeem the information for an entire production line, solve the process issues, and go straight into mass production.

For now, though, he had neither the energy nor the money for that.

His most pressing task was to secure a teaching position, then obtain some startup capital and take the initiative entirely into his own hands.

After driving all the way to the entrance of the Xiaohe Mountain Apartments, Jiang Nan showed no intention of going upstairs for a cup of coffee. She said goodbye and left.

Back home, he gave the kitten that came to greet him with its tail raised a little kick.

"Wu Shufen, you can't even learn a backflip. You're really useless!"

"Meow?"

Wu Shufen was a Napoleon dwarf cat he kept, and a bit of an airhead.

Wu Chen cooked himself a bowl of instant noodles and topped up Wu Shufen's water and cat food. The big one and the little one then tucked in, slurping and crunching away.

After eating, Wu Shufen lay beside him licking her fur, while Wu Chen opened his computer to handle his Taobao orders and considered which theories he should research next.

The end goal of the first stage was definitely the Transformer architecture, which wouldn't appear until 2017. It was the foundation of future artificial intelligence theory.

But he couldn't throw out that trump card all at once. He had to introduce it step by step while gradually building his academic reputation in the process.

So he opened Word and made himself a research list.

Word2Vec, to let artificial intelligence understand semantic relationships in language.

Seq2Seq and LSTM/GRU, to handle sequential information such as sentences and paragraphs.

The attention mechanism, to resolve the Seq2Seq bottleneck.

A few-shot transfer learning strategy to address the shortage of labeled data.

More than an hour later, Wu Chen finally had the general framework sorted out.

He knew something about all these theories, but many details had been lost, and he needed to reproduce them one by one.

Moreover, his personal attributes could also improve during the process.

Stretching, Wu Chen glanced at the time.

Eight in the evening.

The perfect time for research.

With a cup of coffee and a pack of Liqun cigarettes ready, Wu Chen opened his computer and began writing his first paper.

"Optimization of Word2Vec-based semantic vector representations and its application to text semantic matching"

"This paper proposes a Word2Vec semantic vector representation method based on adaptive window adjustment and negative sampling optimization, with the aim of addressing the core problems of semantic isolation and sensitivity to contextual dependencies in traditional word vectors. For example, traditional methods struggle to distinguish between the multiple meanings of 'apple' (the fruit) and 'Apple' (the company), and cannot capture semantic associations such as those between doctors and patients or teachers and students..."

But as he wrote, Wu Chen discovered that things weren't so simple.

Word2Vec technology hadn't matured until 2013, and there was still a great deal of prerequisite knowledge behind it.

For example, the basic theory of distributed word representations and methods for building neural network language models (NNLMs).

So halfway through, Wu Chen had no choice but to open another document and begin writing a paper on the theoretical foundations of distributed word representations.

But that wasn't the end. It was the beginning of yet another undertaking.

Like nesting dolls, as Wu Chen advanced toward Word2Vec technology, one theory after another that had yet to be solved in 2008 kept cropping up.

It was hard to imagine how much knowledge he would need to accumulate when he finally set out to complete the Transformer architecture.

But Wu Chen didn't back down.

It was as though time had returned to a few days earlier, and he once again entered that state of deep research.

Coffee, instant noodles, and cigarettes—the three treasures of research-were his sails on the ocean of knowledge, pushing him ever forward.

He couldn't stop. He simply couldn't stop!

After the sun had risen for who knew how many times, Wu Chen finally achieved his original goal.

Word2Vec.

There were also as many as five short papers to go with it, covering key prerequisite theories such as distributed word representations, NNLM construction, and semantic association modeling.

Looking at the entire set of papers, Wu Chen could hardly believe he had actually done it.

At the same time, he heard a crisp, pleasant notification in his mind.

[Congratulations, host! Your intelligence level has increased!]

The sound was like celestial music, instantly sweeping away Wu Chen's fatigue from the past several days.

He felt as though his brain had taken a cold shower, washing all the dust from its folds. His thinking speed was ridiculously fast.

Many theories that had been hazy before now became crystal clear, and he could even instantly think of ways to link different pieces of knowledge.

He hurriedly opened the system panel.

[Host: Wu Chen]

[Current basic information:]

[Intelligence: level 0 (100%) -> level 1 (0%)]

[Stamina: level 0 (10%)]

[Skill levels:]

[Artificial intelligence: level 0 (25% -> 70%)]

Finally, they weren't all level 0 anymore!

And as a foundation for generative artificial intelligence large models, Word2Vec had actually raised his skill level by a full 45%!

At this rate, as long as he completed one or two more key theories, his artificial intelligence level could also increase!

Before Wu Chen had time to celebrate, he heard the system issue another notification.

[Congratulations, host, on entering the Intelligent Life-Form Sequence. Reward: a random draw]