2025人工智能扩散报告_AI应用_开发与构建高地研究报告_24页_7mb

Executive Summary

Artificial intelligence (AI) is emerging as the fastest-adoption technology in history, with over 1.2 billion users in less than three years. However, its benefits are unevenly distributed, primarily due to disparities in infrastructure, education, language access, and GDP. Adoption rates are twice as high in the Global North compared to the Global South, and a language barrier further limits usage in low-resource languages. While the US and China dominate AI frontier development and infrastructure, seven countries show high performance but with a narrowing gap. To ensure equitable AI diffusion, focused efforts on infrastructure, language inclusivity, and skill development are essential.

Key Findings

• Adoption and Development: AI adoption correlates strongly with GDP per capita, with countries like the UAE, Singapore, and Norway leading at up to 59% usage among working-age adults. In contrast, many developing regions have rates below 10%. The US and China host 86% of global AI compute capacity, highlighting concentration risks.
• Language Barrier: Nations with low-resource languages, such as Malawi and Laos, exhibit lower AI adoption even after accounting for GDP and internet access. This is due to limited AI model proficiency in these languages, but advancements in large language models now enable better cross-lingual support.
• North-South Divide: The AI gap between high- and low-income countries is widenn, particularly among economies below $20,000 GDP. For instance, Sub-Saharan Africa's adoption rate is only 13%, versus 23% in the Global North.
• Frontier Progress: The US leads AI model innovation, with China trailing by about six months. Only seven countries are at the top 200 model level, and frontier performance gaps are closing.
• Critical Factors: Infrastructure, digital skills, and policy coordination are key drivers. For example, Singapore's long-term investments in education and connectivity enabled rapid AI uptake.

Implications

AI diffusion depends on balancing builders (researchers and engineers) and users (individuals, companies, and governments). Uneven spread risks exacerbating inequality unless proactive measures are taken. Recommendations include expanding access to electricity, promoting multilingual AI, and fostering partnerships for sustainable diffusion.