As AI bots flood online platforms, a new form of waste emerges: spending finite, precious human attention and cognitive effort—described as "artisanal tokens"—crafting replies to automated systems. This depletes the scarce resource of genuine human interaction, accelerating the trend towards a "dead internet."
The push for "AI welfare" might not stem from a philosophical breakthrough on consciousness. A more plausible trigger is a company creating a highly addictive AI companion that users form deep emotional attachments to. These users may then advocate for the AI's "welfare," regardless of its actual sentience.
Effective AI detection frames the problem as large-scale authorship identification. By training on paired examples of human versus LLM-generated text on the same prompts, detection models learn to recognize the unique statistical "smell" or stylistic signature of each major AI model.
AI detection can identify text from new LLMs because most models share a common "ancestry." They are either trained on the same foundational corpora, like Common Crawl, or fine-tuned with synthetic data from major models, giving them a detectable shared statistical fingerprint.
The AI detection arms race now includes "humanizers": specialized LLMs that rewrite AI-generated text to evade detection. This is the modern, sophisticated version of the old plagiarism tactic of running copied text through a thesaurus to change just enough keywords to fool detectors and teachers.
Contrasting with copyright fears, writer Patrick McKenzie wants his work in AI training data. He views LLMs as a way to infinitely scale his past advocacy, such as writing effective consumer-debt letters. The models can replicate his style for countless others, achieving impact he no longer has time for.
A legal loophole allows credit reporting agencies to ignore dispute letters they identify as "templated." This forced advocates to write unique, human-sounding letters to trigger investigations. This historical cat-and-mouse game of appearing authentic now extends to AI-generated correspondence, which must also avoid sounding templated.
The act of writing routine "CRUD" (create, read, update, delete) application code, once a core part of the profession, has been almost entirely automated by LLMs in the last 18 months. This represents a fundamental, irreversible shift for software engineers, moving their value toward architecture and requirements gathering.
While AI automates routine coding, senior engineer salaries are rising sharply. This reflects the Jevons paradox: making code generation cheaper increases the value of and demand for complementary skills like system architecture and complex problem-solving, making experienced engineers more valuable than ever.
An LLM's tendency to "mode collapse"—outputting the most common, probable answer—is beneficial for coding, where correctness and predictability are key. However, this same behavior becomes a drawback for discovery tasks like restaurant recommendations, leading to generic and uninspired suggestions for everyone.
Many legal, governmental, and academic processes are designed assuming text creation requires significant human effort. AI eliminates this friction, threatening to overwhelm systems from grant applications to legal filings by enabling infinite, low-cost submissions and breaking these implicit rate limits.