Skip to content

History

This section traces the development of AI safety as a field, from early theoretical concerns to the current mainstream recognition of AI risks. Understanding this history helps contextualize current debates and institutional structures.

The field’s founding period, dominated by the Machine Intelligence Research Institute:

  • Eliezer Yudkowsky’s early writings on AI risk
  • Founding of SIAI (later MIRI) in 2000
  • Development of foundational concepts (orthogonality thesis, instrumental convergence)
  • Superintelligence (2014) brings ideas to academic attention

Deep learning breakthroughs reshape the landscape:

  • AlphaGo (2016) demonstrates superhuman capability
  • GPT-2 (2019) shows language model potential
  • Anthropic founded (2021) by former OpenAI safety team
  • Growing recognition in ML community

AI safety enters public consciousness:

  • ChatGPT (Nov 2022) captures public attention
  • Pause letter (March 2023) signed by prominent researchers
  • Geoffrey Hinton leaves Google to speak freely about risks
  • Congressional hearings on AI safety

AI safety becomes a policy priority:

  • Biden Executive Order on AI (Oct 2023)
  • Bletchley Park AI Safety Summit (Nov 2023)
  • AI Safety Institutes established globally
  • Major labs adopt responsible scaling policies
YearEventSignificance
2000SIAI foundedFirst AI safety organization
2014Superintelligence publishedBrought ideas to academia
2017Asilomar PrinciplesEarly multi-stakeholder agreement
2022ChatGPT releasedPublic awareness breakthrough
2023UK AI Safety SummitFirst major government summit
2024EU AI Act enactedFirst comprehensive AI regulation