When one hears the word “government”, thoughts of inefficiency are at the forefront of many American’s minds. The lines at the DMV are horrendously long. The TSA (the very organization that is supposed to prevent terrorist attacks) has somewhere between a 70-95% failure rate when it comes to detecting weapons.
Inefficiency and bureaucracy are ultimately at the root of many of America’s problems. The tremendous inefficiency of the immigration system, for example, has given rise to the current problem of having 11 million illegal immigrants present in the United States. An overloaded bureaucracy has led to large errors in governmental spending – for example, a Pentagon oversight resulted in a pair of washers being shipped for one million dollars. And the ultimate root of bureaucratic failure is man. Man makes errors, man works at a finite pace, and man cannot work alone.
Yet, many jobs in government can be boiled down to something that machine learning engineers around the world are tackling – classification. Neural networks have been trained to tell thousands of different image classes apart from each other. Apple’s Face ID technology is similar – after a few seconds of scanning, it learns an accurate 3D image of a face – making it a viable candidate for tasks like facial recognition at border control.
The solution, therefore, is relatively simple – automation. Whereas man can only work eight hours a day, an algorithm can work 24-7, with only a meager wage of a power supply. Whereas man is prone to errors, a robot can do a task for hours on end, unfailingly and with precision that no man can match. Robots can communicate with each other at the speed of light, eliminating communications delay, and robots can work infinitely faster when compared to an overworked desk worker. While automation may require a substantial initial investment, it will pay off infinitely in terms of huge cost savings.
Furthermore, the US government already has a major edge in resources when it comes to the application of Deep Learning. The US government undoubtedly has tons of high-quality data at its disposal, and access to powerful computers capable of training highly advanced models. The government is capable of hiring experts in the field with its vast financial resources.
We shouldn’t hastily try and automate every decision-making process in the government – even Machine Learning experts don’t understand how exactly neural networks work, and trusting these “black boxes” to make critical decisions would be foolish. But there are tasks that can be augmented with the use of automation – facial recognition, document validation, recognition of handwritten characters on forms, and many other tasks.
The American government must adopt these new technologies to increase efficiency and optimize cost. We must not buy into Hollywood fears of an “AI Apocolypse” or a robotic government enslaving humanity. Instead, we must realize what machine learning is, and apply it to optimize the various processes involved in running a country.