AI Alignment Benchmark Debuts

AI Alignment Benchmark Debuts

10 July 2025

Pat Gelsinger, former Intel CEO, has launched the Flourishing AI (FAI) benchmark to assess AI models' alignment with human values. The FAI benchmark measures how well AI models support human flourishing across seven dimensions: character, relationships, happiness, meaning, health, finances, and faith. These dimensions are based on research from the Global Flourishing Study, Harvard's Human Flourishing Program, and other faith-based organisations.

The benchmark uses over 1,200 questions and employs both objective and subjective assessments. Multiple judge LLMs with specialised expert personas evaluate model responses based on their primary dimension's rubric and other relevant dimensions. The FAI benchmark aims to promote the development of AI that not only performs well but also aligns with ethical considerations and advances collective human flourishing.

The methodology is being publicly released to encourage contributions from researchers, developers, and interdisciplinary experts. Initial FAI benchmark findings revealed that no current model has achieved a 90-point threshold for excellence in supporting human flourishing across all seven dimensions.

AI generated content may differ from the original.

Published on 10 July 2025
aiintelethicsbenchmarkgelsingerllm
  • AI Models' Safety Concerns

    AI Models' Safety Concerns

    Read more about AI Models' Safety Concerns
  • AI Child Abuse Surge

    AI Child Abuse Surge

    Read more about AI Child Abuse Surge
  • Grok AI Praises Hitler

    Grok AI Praises Hitler

    Read more about Grok AI Praises Hitler
  • Moonvalley's Marey AI Goes Public

    Moonvalley's Marey AI Goes Public

    Read more about Moonvalley's Marey AI Goes Public