Read this AI Index 2025: State of AI in 10 Charts
BLOT: This is the must-read AI report of the year. AI is becoming more efficient, accessible, and globally competitive, but its rapid adoption also brings escalating risks and ethical challenges.
This is it. The 2025 AI Index from Stanford HAI has arrived and reveals a rapidly evolving landscape where smaller models are achieving impressive performance gains [1][2]. Microsoft’s Phi-3-mini, with just 3.8 billion parameters, reached 60% accuracy on the MMLU benchmark, something that previously was only seen in models over 500 billion parameters. At the same time, the cost of querying AI models is dropping, enhancing accessibility at scale. China is also closing the gap with the U.S. in AI development, producing 15 notable models in 2024. Performance differences on benchmarks like MMLU and HumanEval have narrowed, signaling a more globally competitive environment.
Yet, risks are rising in parallel: the number of AI-related incidents surged to 233 in 2024, a 56.4% increase from the previous year. These include deep fakes, bias, and misuse, highlighting the growing need for governance frameworks and safety systems. But political and institutional will remains elusive. As AI integrates into critical sectors, the hard work ahead is ensuring that innovation is aligned with public interest.
Footnotes:
[1] 10 Chart Version: https://hai.stanford.edu/news/ai-index-2025-state-of-ai-in-10-charts
[2] Full Report: https://hai.stanford.edu/ai-index/2025-ai-index-report