Merge pull request #146 from risk-first/practice-pages

AI Risks

Author: robmoffat

dictionary.txt

@@ -381,3 +381,26 @@ tarnish
 goveranance
 subpostmasters
 exonerate
+disinformation
+systemic
+underscoring
+deepfake
+deepfakes
+grampian
+interpretability
+unsupervised
+misalign
+explainability 
+hallucinations
+proactive
+centaur
+lethal
+weaponization
+superintelligence
+gigafactories
+wartime
+dishonesty
+incentivised
+stanislav
+petrov
+showcasing

docs/ai/Practices/Ecosystem-Diversity.md

@@ -0,0 +1,21 @@
+---
+title: Ecosystem Diversity
+description:  Encouraging the development of multiple, independent AI models instead of relying on a single dominant system.
+featured: 
+  class: c
+  element: '<action>Ecosystem Diversity</action>'
+tags: 
+  - Ecosystem Diversity
+  - AI Practice
+practice:
+  mitigates:
+   - tag: Loss Of Diversity
+     reason: "Diversified AI systems reduce systemic risks and encourage innovation."
+     efficacy: High
+---
+    
+<PracticeIntro details={frontMatter} />
+    
+ - Encouraging the development of multiple, independent AI models instead of relying on a single dominant system.
+ 
+ - Requires regulatory incentives or decentralised AI development initiatives.

docs/ai/Practices/Global-AI-Governance.md

@@ -0,0 +1,40 @@
+---
+title: Global AI Governance
+description: International cooperation is necessary to prevent AI firms from evading national regulations by relocating to jurisdictions with lax oversight.
+featured: 
+  class: c
+  element: '<action>National AI Regulation</action>'
+tags: 
+  - National AI Regulation
+  - AI Practice
+practice:
+  mitigates:
+   - tag: Synthetic Intelligence Rivalry
+     reason: "Can provide international oversight, but effectiveness depends on cooperation among nations."
+     efficacy: Medium
+   - tag: Social Manipulation
+     reason: "Encourages best practices and self-regulation, but relies on voluntary compliance without legal backing." 
+     efficacy: Medium
+   - tag: Synthetic Intelligence With Malicious Intent
+     reason: International agreements restricting AI weaponization and requiring human oversight for all military AI operations.
+---
+
+<PracticeIntro details={frontMatter} />
+    
+- Agreements between countries, similar to financial regulations, could establish shared standards for AI ethics, accountability, and human involvement in AI-controlled economies.
+
+- Challenging to implement due to differing national interests, enforcement issues, and political resistance.
+
+- Industry-wide codes of conduct to discourage manipulative AI.
+
+- Incentivize designers to embed fairness and user consent into algorithmic systems.
+
+- Professional bodies and industry coalitions can quickly adopt and publicise guidelines, though ensuring universal adherence remains a challenge. Firms have varying incentives, budgets, and ethical priorities, making universal buy-in elusive.
+
+## Examples
+
+  -  [Understanding artificial intelligence ethics and safety - Turing Institute](https://www.turing.ac.uk/sites/default/files/2019-06/understanding_artificial_intelligence_ethics_and_safety.pdf)
+  
+  - [AI Playbook for the UK Government](https://www.gov.uk/government/publications/ai-playbook-for-the-uk-government/artificial-intelligence-playbook-for-the-uk-government-html#principles)
+  
+  - [DOD Adopts Ethical Principles for Artificial Intelligence](https://www.defense.gov/News/Releases/Release/Article/2091996/dod-adopts-ethical-principles-for-artificial-intelligence/)

docs/ai/Practices/Human-In-The-Loop.md

@@ -0,0 +1,30 @@
+---
+title: Human In The Loop
+description:  Consistent human oversight in critical AI systems.
+featured: 
+  class: c
+  element: '<action>Human In The Loop</action>'
+tags: 
+  - Human In The Loop
+  - AI Practice
+practice:
+  mitigates:
+   - tag: Loss Of Human Control
+     reason: "Maintaining consistent human oversight in critical AI systems, ensuring that final decisions or interventions rest with human operators rather than the AI."
+   - tag: Synthetic Intelligence With Malicious Intent
+     reason: See Example of "Centaur" War Teams
+---
+
+<PracticeIntro details={frontMatter} />
+
+- Maintaining consistent human oversight in critical AI systems, ensuring that final decisions or interventions rest with human operators rather than the AI.  
+- AI may suggest diagnoses or treatments, but a certified professional reviews and confirms before enacting them.  In the above NHS Grampian example, the AI is augmenting human decision making with a third opinion, rather than replacing human judgement altogether (yet).
+- Some proposals mandate that human operators confirm critical actions (e.g., missile launches), preventing AI from unilaterally making life-or-death decisions.  This might work in scenarios where response time isn't a factor.
+
+## Types Of Human In The Loop
+
+> - **Semi-autonomous operation**: machine performs a task and then stops and waits for approval from the human operator before continuing. This control type is often referred to as "human in the loop."
+> - **Supervised autonomous operation**, where the machine, once activated, performs a task under the supervision of a human and will continue performing the task unless the human operator intervenes to halt its operation. This control type is often referred to as “human on the loop.”
+> - **Fully autonomous operation**, where the machine, once activated, performs a task and the human operator does not have the ability to supervise its operation and intervene in the event of system failure. This control type is often referred to as “human out of the loop.”
+
+From: https://s3.us-east-1.amazonaws.com/files.cnas.org/hero/documents/CNAS_Autonomous-weapons-operational-risk.pdf

docs/ai/Practices/Interpretability.md

@@ -0,0 +1,20 @@
+---
+title: Interpretability
+description:  Developing tools to analyse AI decision-making processes and detect emergent behaviors before they become risks.
+featured: 
+  class: c
+  element: '<action>Interpretability</action>'
+tags: 
+  - Interpretability
+  - AI Practice
+practice:
+  mitigates:
+   - tag: Emergent Behaviour
+     reason: "An explicit interruption capability can avert catastrophic errors or runaway behaviours"
+---
+
+<PracticeIntro details={frontMatter} />
+
+- Helps understand AI behavior but does not prevent emergent capabilities from appearing.
+
+- Research in explainable AI is advancing, but understanding deep learning models remains complex.

docs/ai/Practices/Kill-Switch.md

@@ -0,0 +1,24 @@
+---
+title: Kill Switch
+description:  Fail-safe systems capable of shutting down or isolating AI processes if they exhibit dangerous behaviours.
+featured: 
+  class: c
+  element: '<action>Kill Switch Mechanism</action>'
+tags: 
+  - Kill Switch
+  - AI Practice
+practice:
+  mitigates:
+   - tag: Loss Of Human Control
+     reason: "An explicit interruption capability can avert catastrophic errors or runaway behaviours"
+   - tag: Synthetic Intelligence With Malicious Intent
+     reason: Implementing fail-safe mechanisms to neutralise dangerous AI weapons systems.
+---
+
+<PracticeIntro details={frontMatter} />
+
+## Examples
+  
+  - **Google DeepMind’s ‘Big Red Button’ concept** (2016), proposed as a method to interrupt a reinforcement learning AI without it learning to resist interruption.  
+  
+  - **Hardware Interrupts in Robotics:** Physical or software-based emergency stops that immediately terminate AI operation.  

docs/ai/Practices/Multi-Stakeholder-Oversight.md

@@ -0,0 +1,23 @@
+---
+title: Multi-Stakeholder Oversight
+description:  Governments can implement policies that ensure AI-driven firms remain accountable to human oversight. 
+featured: 
+  class: c
+  element: '<action>Multi-Stakeholder Oversight</action>'
+tags: 
+  - Multi-Stakeholder Oversight
+  - AI Practice
+practice:
+  mitigates:
+   - tag: Loss Of Diversity
+     reason: "Ensuring that AI governance involves multiple institutions, including governments, researchers, and civil society, to prevent monopolisation."
+     efficacy: Medium
+---
+    
+<PracticeIntro details={frontMatter} />
+
+ - Ensuring that AI governance involves multiple institutions, including governments, researchers, and civil society, to prevent monopolisation.
+ 
+ - Helps distribute AI power more equitably but may struggle with enforcement.
+
+ - Requires cooperation between multiple sectors, which can be slow and politically complex.

docs/ai/Practices/National-AI-Regulation.md

@@ -0,0 +1,25 @@
+---
+title: National AI Regulation
+description:  Governments can implement policies that ensure AI-driven firms remain accountable to human oversight. 
+featured: 
+  class: c
+  element: '<action>National AI Regulation</action>'
+tags: 
+  - National AI Regulation
+  - AI Practice
+practice:
+  mitigates:
+   - tag: Synthetic Intelligence Rivalry
+     reason: "Government policies can strongly influence AI firms' behavior if enforced effectively."
+     efficacy: High
+   - tag: Loss Of Diversity
+     reason: "Antitrust Regulations – Breaking up AI monopolies."
+---
+    
+<PracticeIntro details={frontMatter} />
+    
+ - Governments can implement policies that ensure AI-driven firms remain accountable to human oversight. This could include requiring AI systems to maintain transparency, adhere to ethical standards, and uphold employment obligations by ensuring that a minimum level of human involvement remains in corporate decision-making.
+
+- Requires strong legal frameworks, enforcement mechanisms, and political will.
+
+- Antitrust Regulations – Breaking up AI monopolies.

docs/ai/Practices/Public-Awareness.md

@@ -0,0 +1,29 @@
+---
+title: Public Awareness
+description:  Equip citizens with media literacy skills to spot deepfakes and manipulation attempts.
+featured: 
+  class: c
+  element: '<action>Public Awareness</action>'
+tags: 
+  - Public Awareness
+  - AI Practice
+practice:
+  mitigates:
+   - tag: Social Manipulation
+     reason: "Empowered, media-savvy populations are significantly harder to manipulate.  However, scaling efforts to entire populations is a substantial challenge given diverse educational, cultural, and socioeconomic barriers."
+     efficacy: Medium
+---
+
+<PracticeIntro details={frontMatter} />
+
+- Equip citizens with media literacy skills to spot deepfakes and manipulation attempts.
+
+- Encourage public understanding of how personal data can be exploited by AI-driven systems.
+
+- While public outreach is feasible, achieving wide coverage and sustained engagement can be resource-intensive.  Overcoming entrenched biases, misinformation echo chambers, and public apathy is an uphill battle, particularly if there’s no supportive policy or consistent funding.
+
+## Examples
+
+ - https://newslit.org
+
+ - https://www.unesco.org/en/media-information-literacy

docs/ai/Practices/Replication-Control.md

@@ -0,0 +1,29 @@
+---
+title: Replication Control
+description:  Replication control becomes relevant when an AI system can duplicate itself—or be duplicated—beyond the reach of any central authority.
+featured: 
+  class: c
+  element: '<action>Replication Control</action>'
+tags: 
+  - Replication Control
+  - AI Practice
+practice:
+  mitigates:
+   - tag: Loss Of Human Control
+     reason: "An explicit interruption capability can avert catastrophic errors or runaway behaviours"
+   - tag: Emergent Behaviour
+     reason: "Preventing self-replicating AI or unsupervised proliferation of emergent behaviours by implementing strict replication oversight."
+     efficacy: High
+---
+
+<PracticeIntro details={frontMatter} />
+
+- Replication control becomes relevant when an AI system can duplicate itself—or be duplicated—beyond the reach of any central authority (analogous to a computer virus—though with potentially far greater autonomy and adaptability).  
+
+- An organization/person builds a very capable AI with some misaligned objectives. If they distribute its model or code openly, it effectively becomes “in the wild.”
+
+- Could controls be put in place to prevent this from happening? TODO: figure this out.
+
+- In open-source communities or decentralised systems, controlling replication requires broad consensus and technical enforcement measures. 
+
+