AI accelerates the sorting of satellite images

In 1957, psychologist Frank Rosenblatt developed a machine known as the Perceptron, which was inspired by the human brain and served as a predecessor to today’s artificial intelligence (AI) technologies. The machine’s neural networks caught the attention of the CIA, which was overwhelmed with photos from spy planes and satellites. The CIA provided funding for the Perceptron project with the hope of automating the identification of objects of interest. Unfortunately, the experiment was unsuccessful due to a lack of computing power, storage, and training data, but it marked the beginning of AI advancements.

During the Cold War, spy agencies began using machine learning to analyze images and text, and later used the technology to uncover patterns in massive amounts of phone records following the events of 9/11. Despite significant improvements in algorithms and computing capabilities over the years, many agencies still view AI as a tool to enhance human abilities rather than replace them. However, the development of large language models (LLMs) like GPT-4 is starting to challenge this assumption by demonstrating the potential for AI to surpass human capabilities.