Summary of Key Points
This news article discusses the industry of embodied intelligence (intelligent robots that can perceive and act like humans) at the "Future Stars of Chinese Entrepreneurs" annual conference. The guests explored four main topics:
1. Industry bubbles are normal and beneficial; the focus should be on products, not valuations.
2. When implementing technologies in real scenarios, start with semi-structured applications that generate revenue (such as in industry and research) before moving onto consumer use cases.
3. In a rapidly appreciating market, organizational management must be driven by a clear mission, solid technical foundations, and a focus on tangible results.
4. The most common mistakes companies make include relying solely on financing without pursuing commercialization, expanding recklessly, and neglecting quality control.
Detailed Analysis
1. Industry bubbles are not a threat but a catalyst for growth
The guests agreed that new industries cannot thrive without some form of bubble.
- The value of bubbles: Similar to the internet bubble, they attract significant investment and talent, eliminating companies that only talk the talk and leaving those that actually develop useful products. For example, after the internet bubble burst, companies like Amazon and Google grew rapidly.
- Current situation: The potential of embodied intelligence is enormous (similar to how large language models have evolved from chat tools to profitable solutions), but current valuations and investments are still far from the levels seen during the internet boom.
- Don’t let valuations dominate decisions: Valuations should be a result of successful products, not the starting point. Companies should focus on delivering quality products and finding practical applications.
2. Start with scalable, profitable use cases
There was debate about whether data or specific use cases should come first; the consensus was to start with revenue-generating scenarios.
- Learning from computers: Computers were initially seen as useless but became widely adopted after being used in research and education before reaching households. Embodied intelligence should follow the same path, starting with semi-structured applications:
- Industrial applications: Self-producing robots in factories can generate revenue to fund further development.
- Research applications: Robots can be used in laboratories to collect data for technological improvement.
- Consumer applications: Household use cases are more complex and will take time to emerge, possibly starting with simple tasks like controlling household appliances.
3. Stabilize your organization during rapid valuation growth
Three companies founded in 2023 saw their valuations soar to tens of billions within three years, with teams expanding from dozens to hundreds of people. They faced challenges such as talent loss and communication issues. Here are some strategies:
- Mission-driven approach: Clearly define the company’s purpose (e.g., “to create a pocket-sized version of Doraemon that can make anything”). This helps employees understand their value and reduces internal conflicts.
- Technological expertise: Hire talented individuals from top universities (Tsinghua, Peking University, Stanford) who are motivated to work late into the night on research and development.
- Result-oriented management: Use OKR (Objectives and Key Results) to align goals frequently and adapt quickly to industry changes.
4. Avoid common pitfalls
The guests highlighted several critical mistakes:
- Don’t rely solely on financing: Cheng Hao’s first startup had millions of users but no revenue, nearly leading to bankruptcy. He realized the need to establish a sustainable business model (technology → product → profit).
- Don’t expand recklessly: Focus on applications that generate recurring revenue and can be scaled.
- Maintain quality control: Poor quality products can damage reputation; prioritize quality during mass production.
- Don’t stick to one technical approach: Be flexible, offering multiple solutions (e.g., direct drive, linkage mechanisms, tendon-like actuators) to meet different customer needs (industrial durability, research flexibility).
The core message is that embodied intelligence must shift from being a showcase of technology to a practical tool that enhances productivity. Companies should focus on solving real problems, expand gradually, and manage their growth effectively, avoiding common pitfalls. For consumers, this means robots may first appear in factories and laboratories before becoming more widely available in households.