OpenAI公布超级智能路线图：模型明年有望重大飞跃，2028年实现全自动AI研究员，基建投资承诺1.4万亿

Core Insights - OpenAI has set two key milestones: achieving AI research interns by September 2026 and fully autonomous AI researchers by March 2028, with the expectation that scientific research will be fully automated by 2028 [1][3][6] - The company has committed to building over 30 GW of infrastructure, corresponding to an investment of approximately $1.4 trillion, which is seen as just the starting point for future expansions [1][9][10] - OpenAI is transitioning from an AI assistant to a platform service provider, aiming for external developers to create more value on its platform than the company itself [3][11] Infrastructure Expansion - OpenAI plans to invest approximately $1.4 trillion in infrastructure, with a goal of creating 1 GW of computing power weekly [9][10] - The company is collaborating with major partners in the chip, data center, energy, and manufacturing sectors, including AMD, Google, and Microsoft [9][10] - The aim is to reduce the cost of each GW to about $20 billion over a five-year lifecycle, requiring significant innovation and collaboration [10] Research and Development - OpenAI is optimistic about achieving significant advancements in model capabilities by September 2026, with expectations of major progress in the coming months and years [4][5][6] - The company is focusing on scaling deep learning training and believes that the timeline to achieve superintelligence may be less than ten years [5][6] - A structured five-layer safety framework has been proposed to address safety concerns as AI capabilities advance [7] Organizational Structure - OpenAI has restructured its organization, with a non-profit foundation at the top controlling the public benefit corporation, OpenAI Group, which holds about 26% of the shares initially [14] - The foundation aims to maximize societal welfare through AI, with initial focuses on using AI to help cure diseases and building resilience against AI-related risks [14][15] Product Strategy - OpenAI is evolving its product strategy from AI assistants to an "AI cloud" platform where others can build applications and services [11][12] - The company emphasizes user freedom and privacy, aiming to provide users with significant control and customization capabilities [11][12] - A hierarchical architecture is being developed, with a focus on hardware, model training, and an ecosystem for applications and services [12][13]