AI Use Cases: State Department
The US State Department is using artificial intelligence and deep learning/neural networks in identifying manipulated photos and video, arms control and new media, and identifying classified content.
We started the search by reviewing the US State Department’s Agency and Bureau pages. From this search, we were unable to locate essential information on the use of AI or deep learning in the state department. However, a more in-depth search led us to a report stating that the previous presidential administration’s state department articles and press releases can be found in the state department archives. We looked into the archives and found one article that stated the uses of artificial intelligence. However, even if we had the name of the original article, it was impossible to retrieve the information from the archive. We decided to use a cache tool to retrieve the article from the archives, and use it on this report.
We then looked into various news reports, case studies, and press releases from reliable publications such as CNN, the Washington Post, Politio, and other scholarly institutions such as Stimson. We also looked into IT publications such as Security Affairs, WIRED, and ITProportal. From this search, we were able to find articles that talked about cyber attacks in the state department, but none of the articles talked about how it was stopped. We also found articles that talked about the use of satellites by the state department. However, no information stated as to how they were used. The useful articles we found was a study commissioned by the state department bureau, and another study conducted by the state department, all of which are included in this report.
AI Use Cases: State Department
Identifying Manipulated Photos and Video
“Fake News” is becoming a new way for terrorists to spread terror in the United States. Various terrorist organizations are known to create this “Fake News” with the use of doctored photos and videos, and disinforming the public by sending it to them in private chat, in where it cannot be monitored in real time. The state department found that AI and deep learning technologies are being used to create these “Fake News,” for media release, and as time passes by, they become increasingly difficult to detect even with the use of modern tools. The state department is worried that the “Fake News,” might disinform the public because they all look too authentic.
MADCOMs or machine-driven communication tools are chatbots that study several online conversations and develop speech patterns to replicate human conversations. However, in its current stage, its conversation is still distinguishable that it is not human, but if MADCOM continues to learn, it is predicted that it will learn to chat like a human being. Matthew Chessen, the senior policy advisor of the US State Department, believes that MADCOMs can determine the tide of battle in this “Media War.” Additionally, he also believes that AI and deep learning technologies can evaluate a society’s culture by programming its inhabitants to what to do and say. With this idea in mind, the state department is planning to partner with companies to use MADCOM with a larger data set, so that it can better its ability to detect misinformation and “Fake News” on the internet with the use of deep learning algorithms.
Arms Control and New Media
The US State Department is looking into the use of AI as a way to control social media and use it to aid in arms control. In a report by the Bureau of Arms Control, Verification and Compliance (ACV) stated that new media functions could help assist in arms control. This includes content creation, social environments to share information, problem-solving by bringing people together to find solutions, and data mining. Moreover, the report talks about developing existing machine algorithms for the use of satellite imagery to detect locations of large weapons, such as surface-to-air missiles.
New Media is predicted to go beyond its former purpose of monitoring and reporting. The functions of new media will include analysis, confidence building, and public diplomacy efforts. Additionally, using new media in arms control is undoubtedly feasible, because the underlying technologies for it to work already exists in other fields such as business and research. All that is left is to cultivate these capabilities as a form of arms control.
Identify Classified Content
One of the state department’s declassified study found certain human errors when it comes to labeling classified information. In one instance, a machine learning program found that a piece of classified information was labeled unclassified. This led to the idea of using AI as a way to prevent these existing human error in natural security. Furthermore, AI can be used to monitor cables and emails, to see if some of this information is deemed classified to the public.