Skip to main content
  • Home
  • Library
  • Read
    • Popular
    • Charts
    • Browse
    • Categories
    • Wikis
    • Glossary
  • Request
  • Activity
  • Leaderboard
  • Help
Our Apps
  • DaoSearch
  • Narratorr
Social
  • Discord
  • Patreon
The AI Wave Tycoon's Rise
Chapter 466

Let me outline a direction first

1,892 words 6 min read 1 view 23d ago Source ​
Skip to the end of the chapter
Prev
Next

On Friday afternoon, Meng Yao, along with Zhan Xiang, the co-founder in charge of algorithms, and two researchers under his command, arrived at the office of Source Intelligence Beijing Research Institute —they would be working there starting next week.

As Meng Yao entered the office building, she took a closer look and noticed that the floor they were about to move to didn't yet have the name of the Beijing Research Institute listed.

Did they specifically rent a new floor to accommodate these twenty-odd people from Mengnao Technology?

With this question in mind, Meng Yao went upstairs and was led to a conference room by the receptionist. There, she met Han Luyi and Zhao Wenyuan for the second time.

The last time they met, the other party was a potential collaborator, and Meng Yao was on the verge of bankruptcy; this time, the other party has become the second largest shareholder of Mengnao Technology, and the two sides are about to begin joint research and development.

Whether it's a blessing or a curse, we should know soon.

Han Luyi greeted Meng Yao as soon as he saw her.

"President Meng, it's nice to see you again. It's a bit presumptuous of me to invite you to move so hastily, but I mainly wanted to make arrangements since confidentiality requirements might be higher in the future, and it would be more convenient for us to work together."

Meng Yao was taken aback, clearly not expecting Han Luyi to offer an explanation.

She thought the email was a show of force from the other party.

"Boss Han, President Zhao, let me introduce you. This is my co-founder, Dr. Zhan Xiang. He is responsible for the EEG intention algorithm in our team." Zhao Wenyuan glanced at Zhan Xiang upon hearing this.

According to the knowledge he obtained from Boss Han, the biggest room for improvement for Mengnao Technology lies in its algorithms.

It seems I'll be having quite a bit of contact with this Dr. Zhan in the future.

After everyone was seated, Han Luyi spoke first, saying, "I have to catch a flight in a bit, so let's hurry. The main purpose of today's meeting is to let the people from Origin understand what Mengnao Technology's current achievements and challenges are, so that we can explore the next direction together. President Meng, please go ahead."

Meng Yao nodded.

These things had naturally been discussed in advance, and she had already made preparations.

After connecting the computer to the large screen, the slides showed a series of head-mounted EEG acquisition devices that Mengnao Technology has developed over the years. The overall trend is towards smaller and smaller devices. The latest third generation is the one that Zhao Wenyuan tested in the lab last time.

Meng Yao began, "Our achievements so far include a patent we hold in the field of interaction. This head-mounted device can collect brainwave signals from the scalp without shaving the head, applying conductive fluid, or positioning electrodes. The device's tens of thousands of microelectrodes ensure that we can collect all brainwave signals that reach the scalp."

She spoke with a hint of pride in her voice. This technology greatly lowers the barrier to entry for using the device, allowing it to be worn and removed at any time. It has an overwhelming advantage over both Invasive methods that require pre-operative surgery and traditional Non-Invasive methods that require complex preparation.

Then, the slideshow changed, and a series of uneven waveforms appeared on the screen.

"Another aspect is the recognition and use of captured EEG signals. We can now reliably recognize several intentions under laboratory conditions, as you can see."

The slides labeled the waveforms with meanings like forward, backward, waving, and grabbing. When Meng Yao said this, her expression wasn't as proud.

Han Luyi also noticed that she specifically mentioned the laboratory environment. Lenovo their last meeting, when she mentioned that it would take her a month to develop her own hand-waving gesture, this was probably where the challenge lay.

Sure enough, Meng Yao continued, "But at this stage, we also face some challenges, some unique to our technological approach, and others common to the field of Brain-Computer Interface. I can summarize it in one sentence: --"

"Signal recognition is difficult, and individual differences are huge."

"The first point is that because we use a Non-Invasive approach, capturing brainwaves requires passing through the skull, which contains a lot of noise. Steadily identifying the signal we're looking for is a huge challenge. Often, it's not that the signal is missing, but that we can't separate it from the surrounding noise."

"The second point should be easy for everyone to understand. Although most people activate the same brain regions for the same action, the specific intentions they correspond to can be completely different from person to person."

"There's a lot to cover, so we'll stop here for now. Please feel free to ask any further questions."

No sooner had Meng Yao finished speaking than Zhao Wenyuan was the first to ask a question.

"President Meng, I saw the waveform on the slide just now—how stable is this waveform for the same individual? For example, if I want to do the same action at different times, how much deviation will there be in the waveform?"

Zhan Xiang answer the question, "President Zhao, that's a good question. Although there are great differences between individuals, in fact, the brainwaves of the same person for the exact same action are basically the same."

After listening, Zhao Wenyuan glanced at Han Luyi and said, "If that's the case, could we also use the Transformer approach, instead of using strict rule definitions to extract signals, put all the feature values into a high-dimensional latent space, and use an attention mechanism to infer their meaning?"

These words weren't directed at the people at Mengnao Technology, but rather at Han Luyi and the team of scientists at Source Intelligence. From the beginning, they never intended to use Mengnao Technology's current matching algorithm, but rather to leverage their experience in Model Training to build a new pipeline for intent recognition.

Zhan Xiang frowned and couldn't help but say, "President Zhao, brainwaves aren't text. It's not that no one in the industry has thought of this approach, but the results so far haven't necessarily been as good as traditional manual feature extraction. The idea of self-supervised training is good, but where will the data come from?"

Zhao Wenyuan immediately said, "That's exactly what I wanted to ask. How much data has Mengnao Technology accumulated so far? As a partner, can't Source Intelligence legally conduct training?"

Brainwave data is not like publicly available documents online, where hundreds of terabytes can be downloaded at will. Its collection and processing are subject to strict restrictions. In many cases, it cannot even be traded as an asset when a company is sold; instead, it is destroyed as private data.

Zhan Xiang stunned again; many questions he hadn't been able to figure out before seemed to have been answered.

It turns out that what they valued was the data accumulated by Mengnao Technology.

This time, Meng Yao provided the answer, "For the past few years, we've had in-depth collaborations with the rehabilitation departments of several top-tier hospitals in Beijing. Because our approach has low risk—or almost no risk—many patients are willing to cooperate. Combined with the volunteers we've recruited, we've accumulated over 30,000 hours of continuous EEG recordings to date. Each recording includes details of our interactions with the patients at the time, what they were thinking, and their physical state. The only problem is that we can't translate all this data into a unified translator."

"Regarding the scope of data use, please rest assured that we conducted rigorous anonymization procedures during the initial data collection, and each participant signed an agreement. We are ready to share this data with our partners for research purposes."

After hearing Meng Yao's words, Zhao Wenyuan couldn't help but gasp.

Even with over 30,000 hours of data, and despite its complexity, if it can be effectively cleaned before training...

Thinking of this, he couldn't help but glance at Han Luyi again.

With Boss Han here, effective cleaning shouldn't be a problem, right? Han Luyi didn't pay attention to Zhao Wenyuan's thoughts at this moment, and instead asked another question.

"President Meng, President Zhan, I previously reviewed Neralink's Invasive solution, which takes less than 100 milliseconds from EEG signal reception to device operation. I wonder how long our current solution takes?"

Upon hearing this question, Zhao Wenyuan also became interested.

That day, Zhao Wenyuan and Han Luyi watched Meng Yao operate the robotic arm to perform actions, but the question was, how much delay was there between the thought and the action? They couldn't tell at the time.

Upon hearing this question, Meng Yao gave another wry smile.

"The outlook is not optimistic. Even if the intent is successfully recognized, the operation delay is approximately 800 milliseconds to one second." This can be considered another major flaw in Mengnao Technology's current products.

"One second?" Zhao Wenyuan exclaimed in surprise.

Han Luyi wasn't surprised; it was consistent with the information he had seen in his Sight.

He glanced at his watch and suddenly said, "Thank you, President Meng and President Zhan, for your explanations. We have benefited greatly. I have to catch a flight, so I must leave early. Before I go, I'll briefly share my thoughts. President Zhao will be in charge of the specific details of the cooperation afterward."

"The first step is to train a general basic model using offline data. President Zhao an expert in this area. Combining the capabilities of the Tangyuan Model, we can first create a general basic model. For each different user, we can calibrate it based on a few basic movements during the first use. Then, we can perform offline fine-tuning training for each user to generate a personalized EEG interpretation model."

"Once the customized model is completed, it needs to be used as quickly as possible during the actual invocation process. Regarding this, a new model architecture that President Zhao has recently been researching should be helpful. Our goal at this stage is to achieve a speed of less than 200 milliseconds."

"That's the general direction. There will definitely be many technical challenges, but I believe everyone can overcome them. If you need any resources or support during the process, feel free to contact me anytime." After saying this, Han Luyi stood up, smiled at Meng Yao and Zhan Xiang, and said, "I wish you all the best in your research. See you next time."

After saying that, he got up and left the meeting room.

Meng Yao and Zhan Xiang exchanged a glance, then looked at Zhao Wenyuan at the same time.

Zhao Wenyuan understood the meaning behind their glances and said, "Our Boss Han had strong ideas about the overall direction of technology. We are planning to try this approach. I wonder what your thoughts are?"

Meng Yao first shook her head, then stopped halfway through, realizing something was amiss. She quickly nodded, then stopped again, saying, "Then let's try this approach, as your company suggests."

This Boss Han is quite unexpected—he has ideas about research and development.

End of Chapter
Prev
NovelSummaries
Next

Explore the wiki

Han Luyi
Zhao Wenyuan

Comments

No comments yet.
AboutStatsAPIChangelogStatusContactPrivacyTermsDMCA
© 2026 RandomTranslator·v3.26·RSS