Ethics and Governance of Artificial Intelligence: Evidence from a Survey of Machine Learning Researchers

AI researchers strongly support safety research and institutional oversight: A strong majority of surveyed ML researchers believe AI safety research should be prioritized and that machine learning institutions should conduct pre-publication reviews to assess potential harms before releasing new work.

Significant trust gaps exist between different actors in AI governance: Researchers place high trust in international and scientific organizations to guide AI development in the public interest, but show low trust in national militaries, Chinese tech companies, and Facebook—suggesting concerns about how different entities might deploy AI technology.

Researchers draw clear ethical lines on military AI applications: While overwhelmingly opposed to lethal autonomous weapons development, researchers show less opposition to other military AI applications like logistics algorithms, indicating nuanced rather than absolute positions on defense-related AI uses.

Moderate skepticism toward Western tech companies: Despite being closer to moderate trust levels, researchers' views on most Western technology companies suggest concerns about whether profit-driven entities will prioritize ethical considerations in AI development and deployment.

Researchers recognize their unique responsibility in AI governance: The survey highlights that ML and AI researchers are "well placed" to influence ethical AI development due to their proximity to the technology, positioning them as key stakeholders whose views and actions matter significantly for responsible AI advancement.