As I discussed in my previous article on Accelerate Screening, AI is an increasingly essential tool for helping global governments improve operations at their ports of entry. By augmenting border patrol and law enforcement professionals with artificial intelligence (AI), global governments can improve traffic flow through their ports of entry without sacrificing security.
In fact, when used to deliver services and capabilities that are impossible for humans, AI can even help global governments get closer to their goal of inspecting 100 percent of the cargo that arrives at ports of entry.
…global governments need to communicate to their customs employees that AI is not going to replace inspection officers. However, officers who use AI will replace officers who do not use AI.
However, despite AI’s benefits to the cargo inspection process, some global governments may find it difficult to implement the technology at their ports. Based on discussions with our government customers, there appear to be two large challenges governments will need to overcome to make AI a part of their cargo inspection processes: culture and data.
AI as a Tool and not a Replacement
Unfortunately, there is a misconception that introducing AI into customs inspections means eliminating customs officers’ jobs. Considering the hysteria from academia, social media, and the mainstream media about how AI will replace us, this is somewhat predictable, albeit very wrong.
It is important for customs and law enforcement professionals at ports of entry to realize that AI is not replacing them in their jobs. Rather, AI will be used to accomplish certain tasks. Each job consists of many tasks, AI will take over some of the more redundant ones, freeing customs professionals to focus on the more mission-critical ones. In other instances, AI will perform functions that are downright impossible for a human to perform.
Digital economists have predicted that almost every job will have some tasks that algorithms can perform better. Therefore, customs officers will do more tasks that humans do better, such as personal interviews, putting inspection results into context, and making final decisions based on ambiguous inspection data. Algorithms will then assist with more mundane tasks, such as detecting a pattern of illicit cargo in X-ray imagery or correlating current inspection data with historical examples.
By acquiring screening data from industry partners, global governments can ensure that the quality is suitable for algorithm training and that the data is properly staged for fast access and algorithm training.
Ultimately, global governments need to communicate to their customs employees that AI is not going to replace inspection officers. However, officers who use AI will replace officers who do not use AI.
This is the tougher AI adoption challenge for governments to overcome because it requires changing individual perceptions and opinions, and overcoming inherent concerns about change and obsolescence. But it’s not the only challenge.
Garbage in, Garbage out
There’s a saying in the scientific and technology community that goes, “Garbage in, garbage out.” This saying has nothing to do with actual refuse and everything to do with the quality of the data being analyzed and the findings identified. If poor data is used, then all findings and correlations that are identified will also – invariably – be of poor quality or no consequence.
Unfortunately, the biggest obstacle to developing AI algorithms for customs inspection is the lack of access to quality, curated scan data. There are mountains of data that are available to global governments, but not all of it is high quality. Much of it is coming from a single source. And very little of it has been cleaned and curated in such a way that will make it usable and effective for training AI algorithms and models.
This creates a “garbage in, garbage out” problem for government data scientists. The available data either isn’t properly prepared and vetted for use or – in far worse circumstances – inaccurate or altered.
While this is the biggest obstacle for global governments, it’s not the most difficult to overcome. In fact, all they have to do is work with their existing industry partners. The companies that many global governments are already turning to for turn-key inspection solutions will often make data available to their government customers for use in the training of AI algorithms.
Based on discussions with our government customers, there appear to be two large challenges governments will need to overcome to make AI a part of their cargo inspection processes: culture and data.
Often, this data has already been curated, meaning the partner has gone through the process of data aggregation, cleaning, augmentation, and conversion to non-proprietary formats. In addition, the partners have often removed sensitive information and applied digital signatures to prevent data alteration. They’ve also taken the time to perform all necessary labeling and annotation.
By acquiring screening data from industry partners, global governments can ensure that the quality is suitable for algorithm training and that the data is properly staged for fast access and algorithm training.
While the use of AI to augment human personnel at ports of entry and along international borders is incredibly beneficial for global governments, these two challenges could keep some from embracing this revolutionary technology. While acquiring the data needed to train AI algorithms is the largest challenge, it’s also a relatively easy one to solve. The culture issue will invariably take more time.
To learn more about data services that can help accelerate the AI programs at your ports of entry, click HERE.
The author, Denis Dujmic, PhD, is a Data Scientist and the VP of Algorithm Researchat S2 Global.