Zhiyuan Institute of Strategy and Defense Mu Jian/Compiled
From: the website of the Center for American Strategic and International Studies
[Zhiyuan Guide] The editor of this article is selected from an article written by the Center for Strategic and International Studies (CSIS) Technology and Intelligence Special Research Group using emerging technologies to improve the strategic intelligence analysis capabilities of intelligence analysts. The original title is The Analytic Edge: Leveraging Emerging Technologies to Transform Intelligence Analysis. This article first studies the new methods of integrating technology into the intelligence analysis process, then discusses the key obstacles and limitations of integrating artificial intelligence and other technologies into strategic analysis, and finally discusses intelligence analysis in the era of AI and big data. How should personnel improve their own strategic value, and provide decision makers with the maximum value of intelligence, as well as the significance of building a future intelligence system.
If the United States invests in technological transformation now, intelligence analysts in 2030 will look back and forth at their counterparts in 2020 with incredible, even sympathetic and compassionate eyes. With the world's leading artificial intelligence, cutting-edge data analysis, and unlimited cloud computing capabilities, future analysts will be able to almost continuously master their target combat environment. In all fields of information from open source intelligence to highly confidential intelligence, they will be able to quickly mine, integrate, visualize and use high-quality data within all the information. They will be able to quickly provide high-level, data-rich advice to policymakers. Regarding the expedient analysis techniques adopted by the predecessors and the outdated analysis process of "reading, writing, thinking", future analysts may shrug their shoulders, thinking that these are the legacy of the past era, and the speed and scale of the big data era. incompatible.
However, intelligence analysts in 2020 have neither time nor interest to consider this fantasy future scenario. As the amount of data increases exponentially, their ability to process data will increase slightly. Their monitors are full of multiple intelligence arrangements, isolated shared drives, manually managed spreadsheets and databases, and error-prone kmz files, and no interface to synthesize data. They were overwhelmed by the many "new tools" and "artificial intelligence solutions" provided to them, and were disappointed in their practicality and usability for strategic analysis. In the case of uninterrupted customer requirements and schedules, analysts will default to using the small set of trusted isolated information sources and time-tested intelligence technology they are used to to collect information, and provide a "good enough "Information products that are more or less punctual."
Although the current scenario is not as bleak as the above-mentioned, and the future may not be as bright as the above-mentioned, the analysts of the intelligence department are indeed completely behind the technology level curve in 2020. The explosive growth of data and disruptive technologies, the rapid evolution and emergence of new global threats, and the acceleration of the decision-making cycle of policy makers are likely to disrupt the intelligence analysis process. How the intelligence department can quickly integrate advanced technology into full-source intelligence analysis will be very important for its competitiveness in the future intelligence environment and for providing timely, accurate and relevant analysis products. While envisioning and shaping future intelligence analysts, intelligence agencies can and must use emerging technologies to enhance the capabilities of analysts.
Limitations of New Technology in Intelligence Analysis Work
For analysts, although the advantages of artificial intelligence and related technologies may be huge, intelligence agencies face some key obstacles and limitations in applying these tools to full-source analysis. The broad challenges of technology acquisition, digital infrastructure, and data architecture identified by the research team in the first phase hinder intelligence collection and processing tasks, which will also affect intelligence analysis tasks. But structural obstacles are neither the only obstacle nor the main obstacle. The main reasons that hinder the adoption of emerging technologies for analysis are the technology’s own limitations in meeting analysts’ intelligence and interpretability standards, as well as the cultural and institutional preferences of analysts and institutions on their traditional intelligence and analysis methods.
Data acquisition and fusion
Accurate and insightful artificial intelligence applications need to capture, organize, and manage certain data. The sheer amount of potentially relevant intelligence and data used for analysis may exceed the processing, filtering, and absorption capabilities of a highly skilled analyst. The challenges faced by the standardization and fusion of data sets will further complicate the use of data. These data sets are both obtained through confidential methods and collected from public sources.
Keep up with the pace of data development
The speed and scale of the intelligence department's fusion of intelligence and data processing tools must keep up with the ever-increasing number and diversification of intelligence and data. Even if artificial intelligence is optimized and streamlined, the proliferation of sensors, intelligence streams, and open source intelligence data (5G networks and IoT devices accelerate this process) will still overwhelm the processing power of analysts. The inability to capture and analyze real-time data will put intelligence analysts behind the standard curve in providing situational awareness to decision makers.
Data Fusion
For analysts, the best artificial intelligence applications will be able to use both secret and open source intelligence data to train algorithms and gain insights, but incompatible architecture and security barriers may hinder "low-end" and "high-end" data Fusion. As with data, machine learning algorithms and open source models may face similar obstacles when ported to confidential systems and integrated into analytical workflows.
Data label
The best artificial intelligence applications also require a lot of data, and this data requires a lot of labels and tags-this is a lengthy, time-consuming, and mostly manual task1. Unlike the private sector, which can crowdsource and hire “gig economy” labelers, the confidential data sets of the intelligence service need to be labeled internally, and most of the work is done manually by analysts. Although it is not impossible to accomplish these tasks in a short period of time, as data continues to grow exponentially, manual labeling and marking will be overwhelmed2.
Algorithm limitations
Analysis relies on rigorous intelligence techniques and clear explanation/reasoning of the logic, evidence, assumptions and inferences used to draw conclusions. The complexity of strategic analysis, standards and requirements for transparency and intelligence assurance, as well as the inherent challenges of modeling and analysis processes and functions, will impose theoretical and practical limitations on the current use of artificial intelligence analysis workflows.
Strategic analysis modeling
Complex intelligence technology and strategic analysis cognitive skills are inherently difficult to define, standardize, and replicate, and therefore difficult to model, which imposes practical limitations on artificial intelligence applications. Put the meaning of new intelligence in the environment and identify it, weigh and connect the data to form an intelligence graph, organize the intelligence logically and convincingly demonstrate and evaluate-this analysis process combines art and science, and standardization The intelligence and savvy of intelligence technology applications and intelligence researchers, as well as judgments drawn from hard data, deductive reasoning, cultural expertise and long-term honed analytical intuition. If the analysis task cannot be digitized, the ability to apply artificial intelligence will be limited3.
deviation
Insights generated in artificial intelligence require analysts to help shape, improve and guide algorithms and models, but analysts will have deviations in the way of conceptualizing intelligence issues, designing models, and selecting input data, resulting in biased and possibly inaccurate the result of. The transparency of inherent biases in the data, the way the models are used, and their impact on conclusions and confidence levels will be critical, but it may not be easy for customers to understand4.
Interpretability
In order to apply the results of artificial intelligence, analysts need to understand the logic, deviations, assumptions, and inferences of the algorithms and models that produce these results-these may or may not be known. Many of the most complex artificial intelligence applications and machine insights are derived from "black box" algorithms in which machine logic and processes are difficult to define (if not impossible). The lack of transparency in the evidence chain, where and how to use artificial intelligence, and validity conditions, etc. mean that the conclusions drawn by machines may be untrustworthy and unusable5.
Authenticity
Analysts must constantly assess the quality, accuracy, and relevance of intelligence, while learning how to measure a new factor that was once taken for granted: authenticity. Techniques to misclassify data through deception algorithms, and techniques that use generative hostile networks 6 to deeply falsify confidential and open source data may confuse analysts, leading to analysis errors and policy decision errors7. As adversaries become more adept at tampering with data, and launch targeted deceptions of false information at a faster speed and scale, ensuring the authenticity of data and intelligence will only become more difficult.
safety
Analysts will also face aggressive and targeted artificial intelligence countermeasures from the intelligence agencies of hostile countries. These intelligence agencies are designed to infiltrate and disrupt artificial intelligence systems and thereby affect analysts’ confidence in artificial intelligence tools and results. The rush to adopt artificial intelligence may bring vulnerabilities to a series of "anti-artificial intelligence" threats at the cost of strict artificial intelligence security standards, protocols, and testing requirements. These threats include "toxic" data injected into artificial intelligence models, and System completely hacked and manipulated 8. Even if the adversary cannot obtain access to this level and convince analysts that their artificial intelligence system is damaged and unusable, the same effect can be achieved9.
Analytically hate reform
Although the technical obstacles are huge and real, the biggest obstacle to the application of artificial intelligence may be the analyst himself. In the field of intelligence analysis, deep-rooted are institutional, bureaucratic, and cultural preferences, as well as the tried-and-tested intelligence techniques and skills they consider to be the global gold standard. Underinvestment in digital intelligence, uncertainty in the value of artificial intelligence and open source intelligence tasks, and cultural aversion to risk and change may prevent the most innovative analysts and departments from incorporating emerging technologies into their tasks.
Digital literacy
Analysts need to have basic digital skills in order to effectively use artificial intelligence and analytical tools in analysis, and to explain the insights derived from artificial intelligence to policy customers who are even less proficient in digital technology. To develop these skills, analysts need not only professional training, but also leadership and management support. However, institutional leaders need to strike a balance between investment in digital skills and traditional intelligence technology, language and other regional training, which will also be critical to the analytical capabilities of intelligence agencies.
Bureaucratic obstacles
AI investment requires years of commitment to implementation and integration, social capital management expenditures to obtain institutional support, and leadership's acceptance of risks and occasional failures. However, middle and high-level managers tend to only stay in their positions for 2-3 years. They may not be willing to spend their already tight time and resources on new technologies with uncertain task returns and risk of failure, especially if Their intelligence leaders and supervisory agencies discourage this kind of adventurous words.
Task value
If analysts and managers do not see clear and substantial "mission benefits" from technology, training, motivation, and leadership support may still not be enough to stimulate the enthusiasm for technology adoption. The marginal benefits of insights and productivity may not justify the time, expense, and opportunity costs required to acquire artificial intelligence and analytical capabilities. Analysts may also be provided with too many technical tools to see the value of any tools, especially those tools that are not specifically designed and customized to meet their unique analysis needs. Analysts who have confidence and trust in traditional intelligence technologies are more likely to give up and not adopt new technologies that they cannot adapt to.
Trust in non-traditional methods
The use of artificial intelligence capabilities requires open source intelligence as an important source of analysis and learning how to gain trust in machine-derived results. Obstacles to the adoption of new technologies are the intelligence agencies’ preference for secret reports that constitute judgments, suspicion of open source intelligence as diagnostic data (this suspicion will only increase with deep forgery and falsehood), concerns about artificial intelligence security, and Trust in time-tested intelligence technology rather than black box procedures. The preference for secret reports may be understandable, because the interception of a signal intelligence or a source of human intelligence may be the only way to distinguish plans and intentions. However, while waiting for the secret information to be collected and processed, ignoring timely and practical open source intelligence insights will cause analysts to lag behind the information needs and decision-making cycles of decision makers.
The Strategic Value of Professional Intelligence Analysts in the New Technology Era
Obviously, when weighing its advantages and limitations, emerging technologies such as artificial intelligence, cloud computing, and advanced analytics can automate key analysis tasks and create more strategic bandwidth for analysts. But what value does technology create for the analysis itself? High-level analysis must answer complex questions for decision makers (for example, what is the prospect of a conflict between an ally and a hostile force? Will a large-scale protest in country X evolve into civil war)? To answer these questions, you need to answer a series of interrelated sub-questions. These sub-questions must be linked into a coherent analytical story, namely: what happened, why, its impact, its prospects, and what benefits to the United States. influences. So, in what ways can new technologies help solve these problems?
The direct value of technology lies in the ability to solve the problems of how to capture, organize, correlate, and understand massive amounts of intelligence and data streams. These data are related to the analyst's country, problem, or goal of concern. Technology can also help analysts assess impact—detecting and measuring the impact of problems or actors on the combat environment. The lag of artificial intelligence and related technologies answers the reason. Understanding the driving factors, intentions, and motivations of foreign actors, as well as the history, background and personality that shape their behavior, are mainly the domain of anthropological experts. As artificial intelligence technology advances, it may become more and more capable of identifying these driving factors, thereby helping to predict the prospects of the problem. But for now, explaining the impact of intelligence on American policymakers will still be the unique advantage of human analysts.
Although emerging technologies will provide immense value to the analysis of intelligence agencies, another question will arise in the next few years: how valuable is the analysis of intelligence agencies to US policy? Although intelligence agencies will still enjoy many advantages, such as Know about intelligence collection, but the combination of high-quality open source intelligence, geospatial intelligence and signal intelligence that can be obtained from commercial channels, and data analysis will make the intelligence analysis competition more intense. Any well-trained and well-equipped organization will be able to conduct a full-source analysis of current events at a faster speed and lower cost, with a quality comparable to that of analysts in the intelligence department. In the future increasingly common sensor and continuous sensing information environment, the business sector's faster technology adoption rate and superior facilities of open source intelligence can give it an advantage over the intelligence sector in assessing rapidly changing global events.
The competitive advantage of intelligence analysts may be diminished in providing intelligence on threats and events in the coming years to US policymakers. However, since the intelligence agency's goal is to distinguish itself from open source intelligence, its value to US policy will not come from a slightly better current affairs analysis of the CNN. Although the intelligence department can and must provide timely analysis to maintain relevance to policy makers, the strength of the intelligence department should still be the expertise of its experienced analysts and the analysis they can provide to policy makers alone. , That is, they have unique and unparalleled insights into the causes and prospects of global events, their impact on American interests and emerging threats.
Of course, new technologies will still be very important. An intelligence department analyst who has mastered artificial intelligence and open source intelligence skills can quickly understand what is happening, understand secret intelligence and historical background, understand the reasons, and provide information on global threats, future scenarios and their impact on US policy Insights. The combination of emerging technologies, anthropological expertise and intelligence technology will place analysts in the intelligence department in a unique position to answer various tricky, usually technology-oriented questions that decision makers may raise in the next few years, such as :
Which ones are new?
As U.S. competitors increasingly use unconventional, indirect, and secret methods to gain strategic advantage, analysts must be able to discover new and increasing in the political, paramilitary, information, and economic fields. Gray Area" Activity 10. Analysts with expertise in artificial intelligence signal detection, pattern discovery and visualization tools, as well as enemy strategies, campaigns, and operational principles will be in the best position to discover new actions and identify the increasing but meaningful in the combat environment Changes provide early warning for US policy makers and reduce the risk of strategic surprises.
Which ones are true?
As foreign disinformation and influence activities accelerate (which are faster, larger, more complex, and seem more real), American policymakers will need intelligence agencies to help distinguish "true" And fiction". Analysts will need to have artificial intelligence capabilities, such as detecting synthetic and unreal deep fraud in a generative confrontation network, and sentiment analysis to measure the factors that influence actions. Analysts with basic technical skills and national expertise will be very suitable for evaluating opponents’ information warfare strategies and potential future actions.
What to do in the future?
Anticipated strategic intelligence is not to predict specific threats, but to envision and correctly assess the possibility of potential events and hostilities. Artificial intelligence-based modeling, war deduction, and scenario analysis can help analysts identify and discover potential course of action, predict enemy decision points, and discover all kinds of situations before low-probability but great impact on US interests. sign.
Empower current and future intelligence analysts
The intelligence agency’s ability to integrate and utilize innovative technologies in strategic analysis is essential for generating and maintaining decision-makers’ decision-making advantages against increasingly complex enemies and competitors. In order to maintain its analytical advantage, the intelligence department must simultaneously begin to envision, plan, and invest in future analysis tasks, and quickly adopt and absorb emerging technologies in today's intelligence technology.
It will be the intelligence analysts themselves who will connect current and future analysis needs. Due to the limited recruitment quota, long training time, low personnel mobility, and high retention rate, it is not easy for the staff of the intelligence department to be replaced by talents with new technologies11. Indeed, as Joseph Gartin, the former chief learning officer of the Central Intelligence Agency, asserted, “the labor force of the future is already here.” Analysts in 2020 may become future leaders and managers. For many For people, they will still be analysts by 2030. Intelligence leadership and important stakeholders—policy makers, Congress, technology and research departments—must provide these analysts with technology and training so that they can thrive today and lay the numbers for future success. Foundation, institutional advantages and cultural norms. So, what should we do now?
Use open source intelligence
In terms of providing information and promoting analysis and judgment, as well as in terms of strategic necessity in a big data era, intelligence agencies must redefine open source intelligence as a kind of basic intelligence along with traditional secret intelligence. In addition, open source intelligence can not only be used as an important reference information for secret intelligence evaluation, but it can also be used as an analysis task. The combination of high-quality open source intelligence, commercial geospatial intelligence, and signal intelligence means that intelligence analysis from all sources can now be performed at a non-confidential level. Rather than treating it as competition, the intelligence agency might as well consider "finished open source intelligence" as an opportunity to expand the coverage of intelligence and have an impact on new customers and stakeholders who may value intelligence Department’s intelligence technology and insights obtained through non-confidential means, including domestic law enforcement, foreign governments, technology and industry, and the general American public.
Pay attention to science and technology information
Information about foreign artificial intelligence systems and scientific and technological capabilities, plans, and intentions must also be considered as a basic intelligence task, collected and analyzed, which is vital to the planning and resources of future intelligence agency tasks. Intelligence agencies must be able to understand and predict emerging technologies—especially artificial intelligence, biotechnology, and quantum computing—and their applications in foreign policy, economic competitiveness, military, and intelligence operations. To do this, it is necessary to secretly collect the technological capabilities and applications of opponents, as well as formal open source intelligence on the sources of foreign technological innovations (including patents, partnerships, acquisitions and expansions). Analysts need more technical and tactical knowledge to understand foreign artificial intelligence systems, as well as their own artificial intelligence capabilities and limitations in collecting, locating, and acquiring data.
Integrated technology
Intelligence analysts need to develop a certain degree of digital intelligence in data science and artificial intelligence. However, working with real technical experts, such as data scientists, machine learning engineers, and product designers, can unleash the true nature of artificial intelligence analysis. potential. Incorporating data scientists into the analysis department will help data scientists understand and analyze problem sets, help analysts understand related artificial intelligence, jointly build and customize models, apply the correct tools to the correct data set, and generate Give meaningful results 12. Machine learning engineers and product designers need to contact the end users of analysts to understand how to design, build, and adjust software, tools, and interfaces to suit the unique requirements of analysts.
Spread the sporadic achievements
While creating digital infrastructure and institutionalized incentives for system-wide technology adoption, intelligence department leaders should also authorize individual supervisors and task centers to obtain, test, and adopt analytical tools that suit their unique mission needs. Certain analysis tasks, especially tasks that focus more on practical intelligence like counter-terrorism, will be more suitable for using artificial intelligence/machine learning. But the leaders of the intelligence department should determine the attributes, norms, and best practices of the technology transformation unit, and seek to disseminate lessons learned to inspire creative methods throughout the department.
Education policy maker
The value of intelligence analysis ultimately stems from the product's impact on policy customers, as well as customers' trust in the quality, clarity, and transparency of interpretation of their judgments. As intelligence agencies integrate artificial intelligence and data analysis into their products, analysts must be able to clearly and convincingly explain to decision makers how these technologies are applied, their relative weight in forming evaluations, and their impact The impact of the confidence level of key judgments. Analysts must become educators of artificial intelligence and analytical applications, and learn to build trust with strategic leaders who make key policy and action decisions based on artificial intelligence analysis.
In terms of enriching and promoting analysis and judgment, and in terms of strategic necessity in the era of big data, intelligence agencies must redefine open source intelligence as basic intelligence, just like traditional secret intelligence.
【1】U.S.technology and cloud-provider firm research interview.
【2】U.S.technology and analytics firm research interview.
【3】Paul R. Dougherty and H. James Wilson, Human + Machine: Reimagining Work in the Ageof AI (Cambridge, MA: Harvard Business Review Press, 2018),
【4】Joseph Gartin, “Thinking About theIC’s Talent Management Issues in an AI/ML Environment,” Elevated Debate, July8, 2020, https://elevateddebate.com/thinking-about-the-ics-talent-management-issuesin-an-ai-ml-environment/.
【5】Congressional Research Service,Artificial Intelligence and National Security (Washington, D.C.: CRS); PatrickTucker, “What the CIA’s Tech Director Wants from AI,” Defense One, September 6,2017, https://www.defenseone.com/technology/2017/09/cia-technology-director-artificial-intelligence/140801/.
【6】“Two neural networks are trained in tandem: one is designed to be a generative network (the forger) and the other a discriminative network (the forgerydetector). The objective is for each to train and better itself off the other,reducing the need for big labeled training data.” See National Security Commission on Artificial Intelligence, Interim Report.
【7】ODNI,The AIM Initiative; CRS, Artificial Intelligence and National Security; ErikLin-Greenberg, “Allies and Artificial Intelligence,” Texas National Security Review 3, no. 2 (Spring 2020): 56-76.
【8】U.S.AI security software and analytics firm research interview, interview by CSIS Intelligence and Technology Task Force, July 2020.
[9] Same as above.
【10】For more analysis on competition in the “grayzone,” see the CSIS Gray Zone Project at https://www.csis.org/grayzone.
【11】Gartin,“Thinking About the IC’s Talent Management Issues.”
【12】U.S.technology and analytics firm research interview.
1. Analyze the impact of "interpretability" on artificial intelligence from an embedded perspective!
2. [MCU] Register, standard library, HAL library, LL library, so many libraries! How do you ask me to choose?
3. How many steps are there to develop embedded projects with Linux?
5. Domestic integrated development environment helps domestic RISC-V break the monopoly of foreign giants in chip technology
6. When doing embedded development, how do you realize LCD display?
Disclaimer: This article is reproduced online, and the copyright belongs to the original author. If you are involved in the copyright issues of the work, please contact us, we will confirm the copyright based on the copyright certification materials you provide and pay the author's remuneration or delete the content.