AI軍事應用領域辨識與資源配置問題之探討：軍事智能化觀點

近年來人工智慧與軍事的整合已加速國際間的軍備競賽。據此，本研究聚焦於臺灣國防產業導入人工智慧決策模式之建構，主要探討AI應用於軍事項目的辨識開發，同時也嘗試解決資源配置的問題，減少資源虛耗，在現今共機不斷擾台的威嚇下深具探討意義。本研究為建構台灣軍事AI投資應用領域之評選決策模式，區分兩階段進行，第一階段為應用領域辨識與評選，先辨識出適合台灣之AI軍事應用領域作為備選方案，再透過管制影響評估觀點及軍事智能化相關研究文獻，建構軍事投資評選指標，計四大構面17項指標，最後應用DANP法計算指標相對權重，並透過ELECTRE II、Fuzzy VIKOR、SAW等多準則決策方法進行方案排序。第二階段為資源配置問題之模擬，由於決策者對評選方案狀態資訊無法確認，故以不確定型決策準則（樂觀準則、賀威茲準則及拉普拉斯準則）進行最終方案之篩選，再以0-1目標規劃解決資源配置的組合問題。結果顯示，「主觀效能」為產官學專家認定影響方案優序程度最高之構面，而「技術獲取不確定性」、「防衛戰力整合性」及「科技人才成本」為最關鍵前三名之指標；此外，投資項目中以「網路安全攻防」為國防最應優先投資發展之AI軍事應用領域。

關鍵字

人工智慧；多準則決策；不確定型決策；目標規劃

並列摘要

In recent years, the integration of artificial intelligence and military has accelerated the international arms race. Accordingly, this research focuses on the construction of the artificial intelligence decision-making model introduced in Taiwan's defense industry. It mainly discusses the identification and development of AI applications in military projects. At the same time, it also tries to solve the problem of resource allocation and reduce resource waste. The intimidation of the author is of great significance for discussion. This research is to construct a decision-making model for the selection and selection of Taiwan's military AI investment application fields. It is carried out in two stages. The first stage is application field identification and selection. First, the AI military application fields suitable for Taiwan are identified as alternatives, and then the regulatory impact assessment is conducted. Views and military intelligence related research literature, construct military investment selection indicators, count 17 indicators in four major dimensions, and finally calculate the relative weight of the indicators using the DANP method, and use ELECTRE II, Fuzzy VIKOR, SAW and other multi-criteria decision-making methods to sort the plans. The second stage is the simulation of the resource allocation problem. Since the decision maker cannot confirm the status information of the selection plan, the final plan is screened by the uncertain decision criteria (optimistic criterion, Hewitz criterion and Laplace criterion), and then Solve the combination problem of resource allocation with 0-1 goal planning. The results show that "subjective effectiveness" is the dimension that industry, government, and academic experts consider to affect the program's highest degree, while "technological acquisition uncertainty", "defense force integration" and "technical talent costs" are the most critical top three. In addition, among the investment projects, "Internet security attack and defense" is the AI military application field that should be the top priority for national defense investment.