Skip to main content

AI Decision Framework

What criteria should be considered to allow an ordinary person to give an informed decision on how to proceed with AI?

Perspectives​

Consider the views of leading AI researchers, computer scientists, and ethicists.

  • Look at the credentials and track records of those making arguments on both sides.
  • Pay attention to consensus statements from reputable scientific organizations.

Risks vs Rewards​

Evaluate the potential positive and negative impacts of AI development, including:

  • Economic effects (job displacement vs. productivity gains)
  • Social impacts
  • Safety and security concerns
  • Potential for misuse
  • Long-term existential risks

Timelines and Urgency​

  • Assess different projections for when transformative AI capabilities may be achieved.
  • Consider whether proposed actions are time-sensitive or if there is flexibility to wait and gather more information.

Feasibility​

Feasibility of proposed actions

  • Examine how realistic and implementable suggested policies or interventions are, both technically and politically.
  • Consider potential unintended consequences.

Evidence​

Evidence and reasoning quality

  • Evaluate the strength of evidence presented, distinguishing between speculation, reasoned arguments, and empirical data.
  • Look for logical consistency and consider counterarguments.

Ethical Frameworks​

  • Consider different ethical perspectives on AI development, such as utilitarianism, human rights-based approaches, or virtue ethics.
  • Reflect on your own values and how they align with different positions.

Predictions​

Global and long-term perspectives:

  • Think beyond short-term national interests to consider global impacts and effects on future generations.
  • Consider existential risks and opportunities for humanity as a whole.

Accountability​

Transparency and accountability:

  • Assess proposals for their provisions on AI governance, oversight, and public engagement.
  • Consider how different approaches might affect democratic control and corporate responsibility.

Control Systems​

Adaptability and course correction: Look for approaches that allow for ongoing assessment and adjustment as AI capabilities evolve and new information becomes available.

History​

Historical analogies: Consider how the development of other transformative technologies (e.g. nuclear power, biotechnology) has been managed, and what lessons might apply to AI.