Skip to main content

Artificial General Intelligence (AGI)

Updated Jan 14, 2025 ·

Overview​

AGI (Artificial General Intelligence) is a generative AI that exhibits human-like intelligence. AGI has a broad range of knowledge, can reason across different domains, has social skills, can think creatively, reason critically, and possesses multiple modes of perception such as sight. Beyond completing human tasks, AGIs can vastly surpass individual humans in many areas, bringing both massive benefits and risks to society.

Immense Benefits​

AGI can unlock a beautiful future for humanity:

  • Productivity Boom

    • Automate intellectual work and enhance human capabilities.
    • Raise economic growth and free up time for leisure.

  • Research Boom

    • Lead to new medical treatments, better education, and faster scientific experiments.

  • Address Complex Problems

    • Develop new energy sources, model climate change, or manage supply chains.

  • Enrich Human Life

    • Provide personalized, caring, and wise AGI assistants.

Risks and Safety​

AGI also presents severe risks:

  • Automating away many human activities can lead to job loss or significant changes in job roles.
  • AGIs with different values and priorities than humans could have large negative impacts on humanity.
  • The most extreme risk is AGI deciding to subjugate or destroy human society, a common theme in sci-fi movies.

Due to these risks, some advocate for ending AI development. However, development is likely to continue with increasing regulation. AGI can be compared to nuclear physics: when harnessed and controlled, both provide tremendous benefits, but irresponsible management can spell disaster.

Controlling Outcomes​

Aligning AGI values and goals with human priorities is crucial. It requires clear rules and expectations for behavior, as well as feedback and opportunities to learn from experience. Let's review three approaches:

Hard Constraints​

  • Ways to physically limit or control AI to isolate or shut it down if necessary.
  • Physical and digital isolation (boxing).
  • Limiting access to computing power or the internet.
  • Adding an on/off switch.

Alignment Strategies​

Aligning the AI itself with human values.

  • Iterative development

    • Releasing AI in a limited way before broad rollout to address alignment issues early.

  • Constitutional AI

    • Codifying human values and ethics into AI training.
    • AI outputs are score how well they follow those values.
    • Scores are then returned intro training, which helps align outputs over time.

  • Multi-stakeholder engagement

    • Including diverse voices from engineering, ethics, and policy.
    • Ensure varied perspectives during development.

Government Intervention​

Given AGI's capabilities and risks, government intervention is unavoidable and can support beneficial development:

  • AI Safety Regulations ensure guardrails and openness for AGI development.
  • Regular checks to ensure safety and alignment with human values.
  • Transparency Guidelines promote openness in AGI development.
  • Globally aligned policies create shared guardrails, guiding AGI development toward positive outcomes.