社会ガイド
軍事および防衛における AI
Defense organizations study AI for sensing, information analysis, logistics, and decision support.
このガイドを聴くAIによるナレーション · ~2 min
概要
These uses differ from autonomous weapon functions, which raise questions about human judgment, testing, and control.
ディープダイブ
Project Maven was announced as a DoD effort to use computer vision and machine learning to help process full-motion video. That announcement does not establish that every present-day system works the same way or is used for targeting. DoD Directive 3000.09 covers autonomous and semi-autonomous weapon systems and emphasizes appropriate human judgment over use of force, plus rigorous verification, validation, and testing. NSPM-11, issued in 2026, directs a review and update of the directive. Autonomy labels alone do not identify who chooses a target or authorizes engagement. The ICRC-hosted policy analysis describes meaningful human control as a debated concept and notes there is no agreement on a single definition. It is a broader international policy discussion, not a claim that DoD Directive 3000.09 requires a human to approve every engagement. The 2023 directive defines a semi-autonomous weapon system as intended to engage targets selected by an operator, and an autonomous weapon system as able, after activation, to select and engage targets without further human intervention. NSPM-11 directed an update to Directive 3000.09 within 90 days and annual review thereafter; the memo’s timetable is not proof the update was completed.
現実世界の実装
Project Maven was announced as an effort to apply computer vision and machine learning to full-motion video analysis.
Predictive-maintenance models can support equipment decisions; results depend on system and mission.
DoD Directive 3000.09 sets policy for covered autonomous and semi-autonomous weapon systems.
A review team can test perception models against expected operating conditions and record limitations.
軍事および防衛における AI の未来
Policy and technical work will continue to address testing, human judgment, and accountability. Since NSPM-11 directs an update to Directive 3000.09, check whether a revised directive has been issued before describing current policy. Do not present research or planned programs as deployed capability.
実践する
なぜそれが重要なのか
- リスクと安全性. AI による壊滅的な被害も日常的な被害も、誰がリスクを理解し、誰が行動できるかにかかっています。
- より明確な判決. 国民と専門家のリテラシーは、強力な安全政策が政治的に可能かどうかを左右します。
- 誇大広告を打ち破る. 明確な説明は、誇大広告、研究室の PR、曖昧な倫理劇場に囚われることを減らします。
気をつけてください
- 能力が複雑になる一方で、実存的なリスクを SF として扱います。
- 高度な自律性の下での調整による表面製品の安全性を混乱させる。
- 英語以外や専門家ではない聴衆には、低品質の情報源しか提供されません。
次のステップ
- 製品の危害、誤使用、制御不能/調整不良のリスクを分離します。
- どのような証拠がタイムラインと重大度についてのあなたの見方を変えるかを尋ねてください。
- マーケティング上の主張よりも、一次情報源と具体的な評価を優先します。
- 意識だけでなく、キャリア、政策、資金、スキルなど、行動経路を 1 つ特定します。
探検を続けましょう
よくある質問
軍事および防衛における AI とは何ですか?
Defense organizations study AI for sensing, information analysis, logistics, and decision support. These uses differ from autonomous weapon functions, which raise questions about human judgment, testing, and control.
In international policy discussions, what is the goal of “meaningful human control” over autonomous weapons?
The term is used in ongoing international debate about human involvement in autonomous weapon use. Its meaning is contested and should not be treated as a statement that DoD Directive 3000.09 mandates approval for every engagement.
Project Maven は主に何をするために設計されましたか?
Project Maven は、コンピューター ビジョン AI を応用して、ドローンが収集した膨大な量の監視ビデオ内のオブジェクトを自動的に識別し、フラグを立てました。
What does DoD Directive 3000.09 emphasize for covered autonomous weapon systems?
The directive sets policy for human judgment and rigorous system verification, validation, and testing.
Why does DoD policy call for rigorous verification and validation?
Directive 3000.09 calls for rigorous verification, validation, and testing to assure intended performance and control.
Under DoD Directive 3000.09, what characterizes a semi-autonomous weapon system?
The 2023 directive defines a semi-autonomous system as one that, once activated, is intended to engage only individual targets or target groups selected by an operator. This differs from the directive’s autonomous-system definition.