Machine intelligence, powered by large language models (LLMs) such as ChatGPT and Microsoft Copilot, has become a ubiquitous and influential force in our society since 2022. Many people, including myself, use these LLMs for various purposes, such as data analysis, image creation, trivia, and writing. These machines are going to have an impact on creative industries as seen in various on going lawsuits but the impact, for now, may not be as dramatic as many have decried.
Personally, I never use AI to generate entire articles from scratch. Instead, I write a rough draft in my own words and then use the LLMs to polish and proofread it. This is what I consider to be a responsible and ethical use of machine intelligence for writing. It requires acknowledging and recognizing the limitations and biases of these LLMs, which are essentially sophisticated statistical imitations of their source data. These LLMs are not trained on some comprehensive and universal collection of knowledge, like the library of Trantor from Isaac Asimov’s Foundation. Rather, they are trained on data sets that are curated from a specific zeitgeist, such as popular sources like the New York Times, Getty, and George RR Martin. These data sets enable the LLMs to produce grammatically correct and concise sentences, or to summarize and analyze content. But they do not enable them to create truly original or creative content. The output of these LLMs is always a reflection of their source data, and therefore subject to question and critique. They are not designed to manifest creativity, but to mimic it within the guardrails of the prompt.
Creativity, in my view, is still the domain of Homo Sapiens. We are the ones who can generate novel and meaningful ideas, expressions, and solutions. Machine intelligence can save us time and provide us with a powerful analytical tool, but it cannot replace our skill, talent, and imagination. Over the next decade, AI will transform many industries and jobs, requiring us to adapt and learn new skills and duties. The creative industries will be especially affected by the quality and design of the data sets that feed the LLMs. But I believe that human creativity will always have an edge over machine intelligence, as long as we do not overestimate the latter or underestimate the former.
Machine intelligence is a remarkable and useful technology, but it is also a dumb and fallible one. It can do amazing things, but it can also make stupid mistakes. It can help us, but it can also harm us. We should use it as a tool, not as a replacement. We should be aware of the ethical and legal implications of using it. And we should always remember that it is, after all, just a machine. A machine that we can control, but also a machine that we should respect.