area51novel Chapter 106 A New Beginning
Chapter 106 A New Beginning
When Zuo Cheng pushed open the meeting room door, the core team had already arrived.
Chen Hao sat at one end of the long table, several technical documents spread out in front of him, his brow slightly furrowed. Zhang Lei leaned back in his chair, twirling a pen in his hand. Fang Ze sat quietly in the corner, expressionless as always. Han Lu flipped through her notebook, while Liu Wei tidied up the projection equipment beside her.
"Is everyone here?" Zuo Cheng walked to the whiteboard, picked up a marker, and said, "Today's meeting is very important."
He wrote three words on the whiteboard: communication, Internet of Things, and AI.
"Since its inception, 402 has gone through three stages. The first stage was the Sky Dome project, where we established a firm foothold in the field of satellite communications. The second stage was smart cities and IoT platforms, where we transformed from a project-based company into a platform-based company."
Zuo Cheng paused, then touched the word AI with the tip of his pen.
"The third phase begins today. 402 officially enters the field of artificial intelligence."
The meeting room was silent for two seconds.
Chen Hao spoke first: "AI? Do we have that capability right now? The Sky Dome project has just completed its third phase of delivery, and the IoT platform is also in its expansion phase. Isn't it too hasty to divert our energy to AI at this time?"
"It's not a distraction, it's a natural extension of technology," Zuo Cheng said. "Think about it, what was the biggest bottleneck of the Sky Dome project? Spectrum scheduling efficiency. How did we solve it? With deep reinforcement learning. What is the core capability of an IoT platform? Edge AI inference. The core technology of both projects is essentially AI."
Chen Hao thought for a moment and nodded: "That's true. But I'm not talking about a technical problem, it's a talent problem. Our team only has Ma Hao and a few newly recruited algorithm engineers who understand AI; that's not enough to support a business unit."
"So the first thing is to recruit people." Zuo Cheng turned to Han Lu, "I need you to get three things done within two weeks: First, post job openings for AI-related positions, focusing on engineers specializing in deep learning, computer vision, and natural language processing; second, contact the AI labs at Huaxia University and Huaxia University of Technology to discuss joint training programs; and third, research the AI chip suppliers on the market."
Han Lu quickly jotted down notes, then looked up and asked, "What about the budget?"
"There's no upper limit to the recruitment budget; talent is the most worthwhile investment. Funding for joint training programs will go through research collaboration channels. For AI chips, we'll conduct research beforehand; there's no rush to buy them."
Han Lu nodded and marked an important point in her notebook.
Zuo Cheng returned to the whiteboard and began drawing the architecture diagram.
"The AI division is not positioned as a general AI company, but rather as a provider of AI solutions for vertical scenarios. Our existing communications and IoT businesses are the best examples. Sky spectrum scheduling, smart city sensor networks, and predictive maintenance for industrial IoT—these are all industry pain points where AI can deeply address their challenges."
He wrote two lines below the architecture diagram:
Core areas: Edge AI, intelligent scheduling, predictive analytics
Business Model: Technology Licensing + Solutions + Platform Services
Zhang Lei raised his hand: "Brother Cheng, I'll handle the chip issue. I've been in contact with several chip manufacturers in the IoT field, and I have some connections."
"Okay, you and Fang Ze will keep an eye on the chip issue together," Zuo Cheng said. "The intersection of IoT platforms and AI is edge computing, and we need low-power, high-computing-power AI chips. There aren't many options on the market right now. For domestic chips, we mainly look at Cambricon and Horizon Robotics, and for imported chips, we look at Nvidia and Xilinx. First, we need to figure out the performance parameters and prices of each company's products."
Fang Ze rarely spoke, but he said, "Power consumption is key. Our IoT terminals are all battery-powered, and if power consumption increases, deployment costs double."
"That's right. So, power efficiency should be the primary criterion for chip selection, followed by computing power." Zuo Cheng marked the priorities on the whiteboard.
The meeting lasted for more than an hour. When it adjourned, Zuo Cheng kept Chen Hao behind.
"There's something you need to prepare in advance," Zuo Cheng's tone turned serious. "For AI algorithms to be implemented, engineers alone aren't enough; we also need computing power. Our current server cluster is used for IoT platforms, and running AI training there would severely strain our resources."
"You mean you want to build a computing center?" Chen Hao frowned. "That's no small investment."
"Don't rush into construction. I have an idea: use federated learning to perform distributed training at the edge, distributing the computing power burden across the IoT terminals. This solves the problem of insufficient computing power while also leveraging the more than 100,000 terminal devices we've already deployed."
Chen Hao's eyes lit up: "Federated learning plus edge AI? This is an interesting direction. Data doesn't leave the local machine, and privacy issues are resolved. But the technical challenges are considerable; communication overhead and model aggregation are both daunting."
"That's why I need you to prepare a technical solution in advance." Zuo Cheng patted him on the shoulder. "Give me two weeks, and I'll figure out the core technical roadmap."
After Chen Hao left, Zuo Cheng returned to his office and closed the door.
He sat down in the chair and opened the system panel.
The tech tree grew quietly in the center of the panel. Four branches, thirty-eight leaves, two hundred and ninety-five points. The nine leaves on the AI branch emitted a soft glow, like nine neatly arranged stars.
Zuo Cheng clicked on the technology preview function.
The name of the fifth branch appeared on the panel: Unmanned System. The branch's outline was vaguely visible, but its leaves were all gray, showing no signs of activation. The sixth branch was New Energy, its outline even more blurred. The seventh branch remained a question mark, offering no information whatsoever.
Zuo Cheng stared at the words "unmanned system" for a long time.
AI is software, while unmanned systems are a combination of hardware and software. If 402 can establish a foothold in the field of AI, the next step is to use AI to empower unmanned systems, achieving a leap from software to integrated hardware and software.
But that's for the future. Right now, the most important thing is to commercialize AI.
He closed the technical preview and scrolled to the list of leaves on the AI branch. Of the nine leaves, he was most interested in federated learning and model compression optimization. Federated learning can solve the problems of computing power and privacy, while model compression optimization allows AI models to run on resource-constrained edge devices. These two directions happen to be the unique strengths of 402 in AI.
Zuo Cheng picked up his phone and sent Yu Ying a message: "Kongkong, want to have dinner together tonight?"
Yu Ying quickly replied: "Trying to make me work overtime again?"
"It's not that I'm working overtime, it's that I miss you."
"Hmph, I'd be a fool to believe you. But I happen to have something to tell you, see you at the usual place."
Zuo Cheng put away his phone, a slight smile playing on his lips.
402's AI journey has officially begun. Ahead lie challenges, opportunities, and areas he's never ventured into before. But he's not afraid, because the technology tree has already pointed him in the right direction.