Artificial intelligence (AI)

Artificial intelligence (AI)

Artificial intelligence (AI) remains one of the fields of technology that continues to witness rapid growth around the world. As a result, it continues to capture the attention of various stakeholders such as policymakers, commercial investors, and defence intellectuals, among many others. That is, artificial intelligence is a vital global issue that focuses on addressing security challenges in future. Many industries continue to implement AI technologies aimed to transform and improve predictions and decision-making. Such techniques include machine learning (ML), information, data, and sensor fusion.

Nevertheless, the deployment of AI presents problems such as operations and monitoring, as well as verification and validation. Addressing these problems will help solve practitioners’ issues and concerns such as trust, effectiveness, and robustness of the AI methods. Blasch’s article focuses on the deployment of AI systems in national security by exploring how the application of standards would address practitioners’ trust-related issues.

Article Summary

The article explores the application of best practices in the deployment of AI systems aimed to boost widespread acceptance of such technologies. The report has several sections, such as a review of the motivation for the development of the AI process, standards, verification and validation, operations and monitoring, strategic recommendations, as well as a conclusion about AI product label. In particular, the authors focus on the analytic standards as outlined in the Intelligence Community Directive 203. They use the same rigorous criteria to compare between machine and human outputs.

The article provides applicable standards of the development and deployment of AI system at each stage of the process. Nevertheless, the authors concentrate on the implementation of AI in national security and argue that advances in such technologies would help build trust between practitioners. They say that advances in AI have resulted in widespread multimedia applications such as natural language processing (NLP) and computer vision (CV) for industrial, commercial, and intelligence, among other different domains. However, Blasch et al. ‘s article reveal that the opaque nature of AI systems in national security reduces the ability of human beings to understand how the system generates results. The authors note that people often rely on ‘black boxes’ to predict outcomes, which in turn inform their decisions. Based on this fact, they warn users to refrain from such practice due to its potential dangers.

The article explains that AI must ensure transparency, trust, and explainability across all stages of deployment and testing. In doing so, the approach will help decision-makers within the Intelligence Community (IC) to improve national security. Through computer vision, NLP, and data transmission, policymakers can utilize various applications of AI to solve massive, novel, and multimodal tasks. They recognize the design, evaluation, test, and operational relevance as a core component of product development that an AI system incorporates to generate rough ideas. Besides, they encourage the technical community to adhere to guidelines, metrics, mandates, policies, and best practices to ensure successful deployment of advances in technologies.

On motivation, the authors note that sophisticated AI applications are becoming more prevalent in the national security environment. With the help of NLP, an AI system can analyze and generate a natural language needed to control the flow of exponentially increasing unstructured information. Analysts must use IARPA’s better program to regulate the extraction of semantic data aimed to detect fake news. For instance, the authors argue that the use of automated systems help identify fake-news because the machine uses the learning models to differentiate between factual and false stories.

ICD 203 allows analysts to use coherent reasoning, relevant information, and consistent syntax and language aimed to improve understanding of the interpretation of results. Evidential reason and information fusion reduce the likelihood of uncertainty of outcomes. As a result, adoption of interpretable models that encourage visualizations would help generate simple results with logical justification for human consumption. Best practices encourage accurate assessments of spatial, modal, and temporal analysis. In this way, the development of an analytics product will solve the interaction issues between the human and machine. ICD 203 provides maintenance details of the model to enable users to understand the accuracy, fairness, and timeliness, among other performance metrics.

The verification and validation methods are essential in determining the relevance and usefulness of natural language processing and artificial intelligence techniques. In particular, verification ensures that a system produces the product or service that complies with imposed regulations and conditions. On the other hand, the audit focuses on virtual testing to ensure that the product or service meets the technical readiness level in line with stakeholder requirements.

After the validation and verification, a product should pass through operations and monitoring stages for data cleaning and labelling. At this stage, the designers should focus on security, privacy, and potential risk of adversarial attacks. For this reason, it is essential to maintain operational robustness to ensure that products reflect the desired needs of sectors such as intelligence, healthcare, and infrastructure. Hence, there is a need to adopt effective training methods that would enable workers to understand and efficiently operate tools in the workplace. The article proposes a multimodal analysis of AI from the development to deployment stages. The model should combine dashboards, data statements, and information screens to leverage the interaction between human and machines. The development process should be simple, transparent, trustworthy, and understandable.

Analysis

The authors use a straightforward approach to explain the role of AI in fighting terrorism and propaganda that might breach national security. Using America as an example, they have demonstrated the need for detective systems to scan texts that advocate for terrorism and any form of propaganda. The opaqueness of the AI system makes it a viable security tool for national security even though it reduces the ability of a human to understand how the machine generates such results. The authors propose an NLP system to engender trust and confidence at the same time reducing reliance on black boxes to predict security outcomes. A standardized system will protect privacy, civil liberties, and American values. In this regard, intelligence analysts should ensure integrity, accuracy, relevance, and objectivity to build this trust against politicization and bias in line with AI standards.

The authors recognize the need for standardization to offset the effects of many developments in artificial intelligence. The article focuses on transparency and trust associated with AI methods. The unification will help develop the reasoning strategy based on the hierarchy of needs. Blasch et al. encourage strict adherence to a variety of guidelines, mandates, and policies across all stages of development of an AI system aimed to identify the needs of the process. The authors seek clarification on how to quantify uncertainty based on probabilistic approaches. They have a justified concern intended to reduce errors while matching the real-time security needs.

The article proposes many standards such as computer vision and cybersecurity to guide intelligence community. However, the authors believe that a stronger intelligence community will require delineation. Providing standards ensures conformity in the industry, as researchers and policymakers follow specific guidance for international systems. The development of a standardized framework ensures transparency of the AI systems Standards for Intelligence Community. The article pinpoints some of the best practices for artificial intelligence in national security. In particular, the ICD203 ensures analytical transparency while observing the hierarchy of needs. It allows analysts to remain consistent throughout their decision-making activities.

The article relies on theoretical and empirical studies to justify the need for validation and verification processes. Such models ensure that the AI system detects the bias and to develop a mechanism for interpretability. The authors also rely on trade studies to justify simulation and modelling rather than virtual scenarios in making predictions. In other words, Blasch et al. explore several data approaches to find efficient solutions in line with the real-time needs of stakeholders. The use of empirical and theoretical studies also helps determine measures of performance and effectiveness, fundamental determinants of product usefulness. These studies enable authors to describe the efficacy of a product through a food label. In this regard, the article demonstrates that the percentages and F scores are essential statistical measures of confidence, uncertainty, and completeness.

Blasch et al. provide a straightforward discussion of the gaps and opportunities associated with the deployment of AI technologies. The vivid analysis prepares analysts to understand the security threats in the complex landscape beyond the national borders. The article offers a critical view of artificial intelligence from a global perspective. The report assesses features of artificial intelligence technologies, thus providing areas that need transformative reforms. As such, the authors provide a framework that focuses on the robustness and resiliency of the AI systems to boost national security.

The report is a helpful frame for both decision-makers and policymakers to examine systems-level opportunities and threats associated with AI. However, the authors did not broaden the definition of security to capture new dangers and risks such as economic and environmental security. Instead, they precisely focused on cybersecurity as a pressing threat to national security. The report utilizes real-world examples to emphasize the security implications of AI, such as promoting terrorism activities and propaganda. In this regard, the report highlights advances in AI as a critical policy consideration for relevant stakeholders.