The essence of music lies not in reading it, but in hearing it. While algebra is inaudible, its power can be felt through the machines it drives.
Systems Sing. It’s not about any fixed code or straightforward process; it’s about creating a dynamic entity that can follow our instructions. This entity isn’t real—it’s a metaphor.
But thinking this way reveals patterns everywhere. Consider the process of sending a traditional letter—something rare nowadays. It involves multiple steps to move the letter from sender to recipient. This is similar to sending an email or transferring data over the internet. Seeing the world this way might seem peculiar, but as Friedrich Nietzsche supposedly said, “Those who were seen dancing were thought to be insane by those who could not hear the music.”
Randomness
The inherent sense of a machine in action can provide insight into almost any phenomenon, including randomness. Even a coin flip, which seems random, follows a complex process influenced by variables like force, angle, height, coin size, mass, and air resistance and so on. Generating “random” numbers on a computer involves running a complicated function. None of these are truly random processes. In April 2024, Avi Wigderson was awarded the Turing Award for linking randomness to hard-to-compute mathematical functions. He developed a process that produces “pseudorandom” bits from a complex function, indistinguishable from truly random bits. Thus, randomness is just computation beyond our prediction.
Managing Complexity
Recent advancements in artificial intelligence, particularly through machine learning, show us how to handle such complexity. Take translation, for instance. A bilingual person can easily understand text in one language and translate it into another. Their brain, acting as the machine, follows a process to convert one language into another. They might not understand the entire process, but it happens, nonetheless. For a computer to translate text, a simple word-for-word translation won’t work due to different language structures and contextual meanings. Linguistic tools help, but they only go so far. A person understands languages from growing up bilingual, exposed to their complexities. Similarly, machine learning trains language models on vast data. These models use a complex neural network with artificial neurons and weighted connections. Properly trained, the neural net predicts the next word in a sequence, translating from one language to another.
Avoiding Unintended Consequences
We might not fully understand the workings of a trained neural net, just as our bilingual person doesn’t grasp their complete translation process. However, we can simulate the process to predict the next word’s probability. If trained perfectly, the neural net’s predictions would be indistinguishable from a human’s. Like Wigderson’s work on complexity functions and pseudorandomness, predicting word probabilities captures the complex calculations involved.
Conclusion
Machine learning advancements enable us to manage complex processes. However, these models are still susceptible to errors, misinformation, and bias since they fail to grasp the practical contextual realities of experience. Nonetheless, they help us manage the randomness inherent in complex systems.
Sphere Economy
Sphere Economy, founded on decades of practical contextual experience, transforms complex random behaviours within and between operational functions into a more organized three-dimensional structure. This structure enhances relational functionality, improves performance, eliminates waste, reduces risk and cost, and uncovers missed opportunities.
Christopher Gleadle