The Artificial Intelligence (AI) revolution is here and is rapidly becoming a central element of the Great Power Competition. Machine Learning (ML) and Natural Language Processing (NLP) capabilities of AI have swiftly taken the pole position in this race. China has declared its intent to be a ‘first mover’ in AI with the goal of leading the world in AI by 2030. China has systemic advantages that will help them achieve this goal unless the Congress, the Director of National Intelligence (DNI), and the Intelligence Community (IC) take aggressive action to accelerate the IC’s acquisition and integration of AI.
Simply put, NLP uses machines to enable humans to fully exploit all available information to make better-informed decisions faster. These machines distill massive volumes of narrative text, structure the content, identify entities, and produce summaries to surface new insights. NLP is not a new concept; it has been around since the 1950s. But only recently have the three essential components of NLP: computing power, advanced algorithms, and availability of massive volumes of data, matured. This perfect storm of technology has transformational potential.
The IC is exploring the application of ML/NLP to core mission execution activities. Many ML/NLP investments have focused on core mission execution functions such as strategic analysis. One IC Agency has made a substantial investment in a platform that will structure the vast volume of textual reporting inundating analysts every day and create a living knowledge base that summarizes issues of national intelligence interest and surfaces new insights and connections, and there is significant untapped potential for NLP at the enterprise level for mission-enabling activities.
Mission enablement, and all the support functions critical to mission execution, are ripe for ML/NLP. Potential applications include records management. logistics, and site security. From name tracing to research, wherever an intelligence officer is working with textual material, NLP can add value. Making this investment in mission enablement will free resources for mission execution.
Across the IC, there is agreement that adoption of AI is critical to our national security mission. The technology is mature and mission applications are plentiful. Yet, we struggle to identify, acquire, and implement AI capabilities at the speed necessary to retain our intelligence advantage. The challenges are numerous, but there are two critical pacing items.
- Acquisition. In this case acquisition has two components; selection of ML/NLP products and procurement. We lack a full picture of the ML/NLP industry, which results in lengthy market analysis or selecting from what is in sight and potentially missing a better solution. The IC needs current knowledge of the entire ML/NLP ecosystem, from the earliest start-up to mature industry leaders, to understand the full range of ML/NLP offerings and rapidly select the best product for the mission. Procurement hurdles present challenges that both slow acquisition and risk eliminating the best ML/NLP solutions. Small companies often have neither the experience to successfully navigate the bureaucratic processes nor the financial resources to sustain them through the first paid invoice. With Congressional and DNI support, the IC can create alternate procurement processes specifically designed to accelerate the acquisition of best of breed ML/NLP products and services.
- Authority to Operate (ATO). After a procurement contract is complete, The Agency begins a series of security reviews to determine whether the newly procured software should be deployed on a classified IT system. The myriad of ATO systems across the IC are locally optimized in a stove pipe structure that impede the dispersion of proven ML/NLP capabilities across the Community. Each IC organization maintains its own standards and ATO process that satisfy their perceived unique security requirements and are designed to reduce risk to the lowest possible level. Further, none of them are reciprocal. This means that an ML/NLP capability that has completed the acquisition and ATO process at DIA and is delivering unique mission value will have to endure a second, independent, ATO approval process at CIA, potentially taking 10-12 months before delivering similar value there. The IC must agree on a common set of security standards that thoughtfully absorb risk. Standards should weigh the risks of approving a capability against the risks and opportunity cost of a denying approval. This set of common standards enable reciprocity and will accelerate broader adoption of proven capabilities across the community.
Accelerating acquisition and ATO requires an ML/NLP-literate IC workforce. ML/NLP is not well understood and is often viewed with wariness and suspicion. To some it is a black box technology, others see it as a threat to their jobs, and to many, it is a faddish term that does not have clear connectivity to their missions. However, our adversaries are crystal clear on how they intend to use ML/NLP to create a strategic advantage over the US. An IC ML/NLP-literate workforce will understand the true potential of AI and recognize where it can improve the speed, accuracy, and completeness of their missions and use it to stay ahead of our pacing threats.
The Center for Strategic and International Studies recently released a paper Move Over JARVIS, Meet OSCAR that describes the challenges for integrating ML/NLP and presents actionable recommendations. The report addresses ML/NLP in the Open Source environment, but the obstacles and many recommendations also apply to the classified environment. It is well worth a read.
Effectively implementing AI tools such as ML/NLP is a team sport. China is marching steadily toward the goal of leading the world in AI by 2030, if not sooner, leveraging every licit and illicit tool to get there. We need the active collaboration of Congress, the DNI, Agency leaders, and industry to drive policy, prioritization, and authority changes so we can outpace China. We can do this as a partnership leveraging every licit tool at our disposal without compromising our values. Every day we risk falling a little further behind our adversaries, it is time to take action.