DeepSeek Prioritizes AGI Development Over Competition with Tech Giants
AI firm DeepSeek is not aiming to become the next ByteDance or Tencent. Instead, the company is focusing its strategy on achieving Artificial General Intelligence (AGI) through principles of restraint, open-source development, and cost efficiency.
AI company DeepSeek is charting a course for Artificial General Intelligence (AGI) that deliberately avoids seeking market dominance or direct competition with established tech giants like ByteDance or Tencent. This strategy, outlined by founder Liang Wenfeng during a recent investor meeting, emphasizes restraint, open-source contributions, and cost-effectiveness as core pillars for achieving AGI.
Liang stated that DeepSeek is not prioritizing maximum commercial profit. The company aims to reduce investment in non-core business areas, focusing resources on advancing AGI. Products are viewed as intermediate milestones on the path to AGI, rather than primary objectives. This approach allows the company to maintain a high technological position for developing more foundational AI capabilities.
DeepSeek will not heavily invest in areas such as 3D generation, video generation, or world models. While multimodal capabilities are important for user-facing products, Liang believes they are secondary components, not the core of AI intelligence. For the Chinese market, the company's focus is on developing general-purpose AI agents, sidelining more specialized agents for sectors like finance or healthcare.
A critical gap in current AI, according to Liang, is the lack of continuous learning. He described the path to AGI as a staircase, where previous steps involved "Chain-of-Thought" reasoning and the current step is focused on agents. The next major hurdle is enabling continuous learning in AI models. This advancement is expected to lead to a phase where "AI accelerates AI research," with AI systems aiding in the development of even more sophisticated AI.
Open-source strategy is considered crucial by DeepSeek, even with the potential trade-off of market advantages. Liang believes it enhances team morale and fosters the broader AI ecosystem. The company plans to release its open-source models with performance parity to their internal versions, not fearing competition from other companies, as sustained large-model development requires significant long-term commitment.