美国最新《人工智能机器学习战略计划》-24页_3mb

S&T Artificial Intelligence & Machine Learning Strategic Plan Summary

Core Content

The U.S. Department of Homeland Security (DHS) Science and Technology Directorate (S&T) has developed a Strategic Plan for Artificial Intelligence (AI) and Machine Learning (ML) to guide the integration of these technologies into DHS missions. The plan outlines three strategic goals aimed at driving innovation, facilitating adoption, and building a skilled workforce, all while ensuring ethical, legal, and privacy-compliant use of AI/ML.

Main Objectives

The Strategic Plan is informed by national guidance and the DHS Artificial Intelligence Strategy, and is aligned with the broader mission of S&T to safeguard the nation through science, technology, and innovation. It emphasizes the responsible and effective use of AI/ML to enhance homeland security capabilities, mitigate risks, and improve public trust.

Key Information

I. Executive Summary

• S&T aims to support DHS mission needs through AI/ML research, development, test, and evaluation.
• The Strategic Plan outlines three main goals to guide the implementation of AI/ML across DHS and the Homeland Security Enterprise (HSE).
• A subsequent Implementation Plan will detail how the Strategic Plan will be operationalized.

II. Purpose

• The Strategic Plan defines S&T's vision, mission, goals, and objectives for AI/ML.
• It identifies focus areas for AI/ML that S&T will address to support its role as a research and development arm and science and technology advisor to DHS and the HSE.

III. Introduction

A. AI/ML in DHS Mission Context

• DHS missions include securing borders, managing cyber and physical risks, preventing crime, and responding to disasters.
• AI/ML can improve efficiency and effectiveness in these areas, such as processing sensor data, scanning cyber activity, and modeling disaster impacts.
• However, AI/ML also introduces new risks, including privacy violations, bias, and adversarial attacks.

B. AI/ML in S&T Mission Context

• S&T is tasked with advancing science and technology to support DHS missions.
• The AI/ML Strategic Plan aligns with the 2021 S&T Strategic Plan and the DHS AI Strategy.
• S&T will research and advise on AI/ML to ensure ethical and effective use across the Department and the HSE.

C. S&T AI/ML Vision

• S&T aims to be a trusted advisor for AI/ML, providing expert technical guidance for research, acquisition, and implementation.
• It anticipates the impact of AI/ML on DHS and the HSE, including adversarial use by threats.
• S&T will foster assessment and acquisition of AI/ML capabilities by DHS Components.

D. Definitions of AI/ML

• Artificial Intelligence (AI): Automated, machine-based technologies that can make predictions, recommendations, or decisions based on human-defined objectives.
• Machine Learning (ML): A subset of AI that learns from training data rather than explicit programming.
• ML's effectiveness is supported by large datasets, computational power, and technical advances.

E. Strategy Development Process

• The S&T AI/ML Strategy Working Group, led by the Technology Centers Division, convened in Summer 2020 to identify research and development priorities.
• The goals and objectives were informed by workshops with DHS stakeholders in October and November 2020.
• The plan aligns with national and international AI principles, including those from the OECD and G20.

Goals and Objectives

Goal 1: Drive Next-Generation AI/ML Technologies for Cross-Cutting Homeland Security Capabilities

• Objective 1A: Advance Trustworthy AI

  • Focus areas: Explainable AI, Privacy Protections, Bias Detection, Public Trust, and Countering Adversarial Uses.
  • Importance: Ensures AI/ML systems are reliable, compliant, and trusted by both DHS and the public.

• Objective 1B: Advance Human-Machine Teaming

  • Focus areas: Optimizing Human-in-the-Loop Architecture and Enabling Collaboration Across Heterogeneous Systems.
  • Importance: Enhances human-AI collaboration and leverages diverse data sources.

• Objective 1C: Leverage AI/ML for Secure Cyberinfrastructure

  • Focus areas: Model Lifetime Management, Threat Detection and Response, and Secure Shared Computations.
  • Importance: Ensures cybersecurity is strengthened through AI/ML capabilities.

Goal 2: Facilitate Use of Proven AI/ML Capabilities in Homeland Security Missions

• Objective 2A: Identify, Evaluate, and Transition Existing AI/ML Capabilities

  • Focus areas: Developing or adopting proven capabilities, matching them to mission needs, and conducting pilot studies.
  • Importance: Enables rapid adoption of AI/ML solutions that are technically mature and mission-relevant.

• Objective 2B: Enable AI/ML throughout DHS Components and the HSE

  • Focus areas: Guiding on accessible tools, advising on technical architecture, participating in governance, and managing automated data governance.
  • Importance: Supports organizational learning and ensures ethical and effective implementation.

Goal 3: Build an Interdisciplinary AI/ML-Trained Workforce

• S&T will recruit and train experts to enhance AI/ML competence within the workforce.
• It will also provide guidance to the broader DHS and HSE communities on training opportunities.
• Emphasis is placed on interdisciplinary collaboration across technical, legal, policy, and social science domains.

Conclusion

• The S&T AI/ML Strategic Plan outlines a comprehensive approach to integrating AI/ML into DHS missions.
• It emphasizes ethical, legal, and privacy considerations alongside technological innovation.
• The plan supports the DHS AI Strategy and is informed by national and international AI principles.
• A subsequent Implementation Plan will detail the operationalization of the Strategic Plan.

Appendices

• A. Acronyms: Definitions of key terms used in the document.
• B. References: Cited national and international AI strategies and reports.
• C. Endnotes: Additional context and citations for the strategic goals and objectives.