AI Capabilities

ai-capabilities (E5)

← Back to pagePath: /ai-transition-model/factors-ai-capabilities-overview/

Page Metadata

No page data found for "ai-capabilities"

Entity Data

{
  "id": "ai-capabilities",
  "type": "ai-transition-model-factor",
  "title": "AI Capabilities",
  "description": "The aggregate advancement of AI system capabilities—including reasoning, autonomy, generality, and domain expertise. Higher capabilities amplify both benefits and risks.",
  "tags": [
    "ai-transition-model",
    "factor",
    "capabilities",
    "scaling"
  ],
  "relatedEntries": [
    {
      "id": "misalignment-potential",
      "type": "ai-transition-model-factor",
      "relationship": "amplifies"
    },
    {
      "id": "misuse-potential",
      "type": "ai-transition-model-factor",
      "relationship": "amplifies"
    },
    {
      "id": "safety-capability-gap",
      "type": "ai-transition-model-parameter",
      "relationship": "affects"
    }
  ],
  "sources": [],
  "lastUpdated": "2026-01",
  "customFields": [
    {
      "label": "Model Role",
      "value": "Root Factor (AI System)"
    },
    {
      "label": "Character",
      "value": "Amplifier (neither inherently good nor bad)"
    },
    {
      "label": "Trajectory",
      "value": "Rapidly increasing"
    }
  ]
}

Canonical Facts (0)

No facts for this entity

External Links

No external links

Backlinks (3)

id	title	type	relationship
ai-uses	AI Uses	ai-transition-model-factor	shaped-by
tmc-compute	Compute	ai-transition-model-subitem	child-of
tmc-algorithms	Algorithms	ai-transition-model-subitem	child-of

Frontmatter

{
  "numericId": "E660",
  "title": "AI Capabilities",
  "description": "Root factor measuring AI system power across speed, generality, and autonomy dimensions.",
  "sidebar": {
    "label": "Overview",
    "order": 0
  },
  "lastEdited": "2026-01-03",
  "subcategory": "factors-ai-capabilities"
}

Raw MDX Source

---
numericId: E660
title: AI Capabilities
description: Root factor measuring AI system power across speed, generality, and autonomy dimensions.
sidebar:
  label: Overview
  order: 0
lastEdited: "2026-01-03"
subcategory: factors-ai-capabilities
---
import {DataInfoBox, FactorSubItemsList, PageCauseEffectGraph, EntityLink} from '@components/wiki';

<DataInfoBox entityId="E5" />

AI Capabilities refers to how powerful AI systems become across multiple dimensions. This is a key root factor in the AI Transition Model because capability levels directly influence the probability and severity of various scenarios.

**For detailed tracking of current AI capabilities, see the Capabilities section.**

## Key Dimensions

<FactorSubItemsList factorId="ai-capabilities" />

## Capability Categories

The Knowledge Base tracks capabilities across several domains:

| Capability | Status | Risk Relevance |
|------------|--------|----------------|
| <EntityLink id="E186">Language Models</EntityLink> | Rapidly advancing | Foundation for all other capabilities |
| <EntityLink id="E246">Reasoning</EntityLink> | Emerging | Key for general intelligence |
| <EntityLink id="E61">Coding</EntityLink> | Human-competitive | Enables self-improvement |
| <EntityLink id="E2">Agentic AI</EntityLink> | Early stage | Enables autonomous action |
| <EntityLink id="E356">Tool Use</EntityLink> | Growing | Expands action space |
| <EntityLink id="E277">Scientific Research</EntityLink> | Emerging | Could accelerate capability growth |
| <EntityLink id="E282">Situational Awareness</EntityLink> | Emerging | Key prerequisite for <EntityLink id="E274">scheming</EntityLink> |
| <EntityLink id="E278">Self-improvement</EntityLink> | Theoretical | Could lead to recursive improvement |
| <EntityLink id="E224">Persuasion</EntityLink> | Concerning | Enables manipulation at scale |
| <EntityLink id="E192">Long-horizon Tasks</EntityLink> | Early stage | Enables complex autonomous projects |

## Relationship to Scenarios

Higher AI capabilities primarily increase the probability and severity of **AI Takeover** scenarios:

- **<EntityLink id="E623">Rapid Takeover</EntityLink>**: Requires sufficient capability for decisive action
- **<EntityLink id="E619">Gradual Takeover</EntityLink>**: Enabled by increasing autonomy and generality over time

Capabilities also affect <EntityLink id="E671">Human-Caused Catastrophe</EntityLink> scenarios by enabling more powerful <EntityLink id="E42">Bioweapons</EntityLink>, <EntityLink id="E86">Cyberweapons</EntityLink>, and <EntityLink id="E35">Autonomous Weapons</EntityLink>.

## Current Trajectory

AI capabilities are advancing rapidly across all dimensions, driven by:
- Scaling laws (more compute, data, parameters)
- Algorithmic improvements (transformers, <EntityLink id="E259">RLHF</EntityLink>, reasoning chains)
- Hardware advances (specialized AI chips, larger clusters)
- Increased investment (≈\$100B+ annually in US alone)

**Key metrics** are tracked at [Epoch AI](https://epochai.org/) and [Stanford HAI AI Index](https://aiindex.stanford.edu/).