With the integration of AI into products and our daily lives, tech companies are encountering difficulty in predicting or explaining the behavior of their AI systems. This lack of understanding is widely recognized among AI developers but not widely known by everyone else, making it the most unsettling aspect of the current AI boom. Unlike traditional computer systems that consistently produce the same output given the same input, generative AI introduces a level of randomness on a massive scale involving trillions of variables. Consequently, it becomes challenging to unravel the process by which the technology reaches a specific outcome. Recently, a group of researchers demonstrated that users can bypass safety features intended to prevent AI systems from providing instructions on harmful activities such as bomb-making.

In this field, reliance on oral tradition and shared tricks is prevalent as developers struggle to explain the behavior of AI systems. It is worth noting that these systems can be adjusted to have varying levels of randomness, resulting in broader or more specific responses. However, the underlying mechanisms behind their operation often seem mysterious. On the other hand, some experts argue that the notion of "not understanding AI" is a convenient excuse. The future remains uncertain as to whether AI developers will eventually gain the ability to provide more in-depth explanations of why and how their systems work.