Yevgeny Zamyatin’s 1924 novel We depicted a future society ruled by absolute mathematical order, where citizens lived in glass buildings for constant surveillance. This vision of totalitarianism, achievable through technology, has echoes today.

While online tracking for advertising is known, Large Language Models (LLMs) present a new, more destructive threat. These tools grant corporations and governments unprecedented power to watch and manipulate individuals. LLMs are parasitic, feeding on every online interaction - chatbot prompts, social media posts, and more - to grow stronger and enhance their surveillance capabilities.

The ability of industries and governments to analyze vast amounts of data has surged with LLMs. They can collate information from numerous platforms to build detailed predictive models of individuals. Governments already use these tools for tracking, and corporate use behind closed doors remains a concern.

Furthermore, LLMs absorb user interactions, learning individual expression styles and questions to become more personalized, and thus more addictive. This detailed user profile allows companies to peer into habits, ideas, and interests, enabling not just advertising power but the potential to influence user behavior. The line between a "behavioral nudge" and manipulation is thin when LLMs can precisely tailor responses to appeal to users and shape their thinking.

The unease surrounding constant observation stems from the invasion of our private lives. LLMs, or rather their owners, can see more than mere eyes or windows, as users voluntarily input hidden thoughts, insecurities, and questions. While LLMs themselves are algorithms, the data they process can be accessed by those who hold the keys.

In this age, lives are increasingly surveillable and manipulable at scale. However, individuals retain agency. Opting out reduces the data available to LLMs and their owners. The adage "necessity is the mother of invention" is reversed; invention is now the mother of necessity. LLMs often fulfill needs created by their own existence, diminishing human capacity and fostering dependency. This dependency makes individuals more susceptible to manipulation through tailored, potentially biased, answers.

It is crucial to distinguish between appropriate technical uses of LLMs and those that replace substantive human capabilities like writing, thinking, analyzing, and decision-making. For tasks like sorting technical data or automating rote tasks, LLMs may be useful. However, when considering LLM use, one must ask if it diminishes capacity, reduces freedom, or disconnects from reality.

A general rule of thumb is to avoid using LLMs for tasks previously achievable with other tools. Writing, reading, analyzing, communicating, artistic production, and moral decision-making are human domains. Collaboration and conversation with others are far more enriching than LLM interaction. Engaging with mentors, reading books, or discussing ideas with people offers a richer, more authentic experience than relying on LLM-generated content.

While tech companies profit from user addiction and dependency, individuals still have the option to opt out. Reading books, thinking in silence, observing nature, talking with friends, and struggling with problems foster genuine growth and resilience. If an LLM genuinely assists in a worthwhile endeavor that couldn't be achieved otherwise, its use might be justified. Otherwise, engaging with the rich, real world remains the superior choice.