Curated Digest: Don't write for LLMs, just record everything
Coverage of lessw-blog
A recent post on LessWrong challenges the growing narrative that individuals should write publicly to secure a digital legacy or train future LLMs, advocating instead for comprehensive private recording.
In a recent post, lessw-blog discusses the evolving relationship between our digital footprints and the capabilities of Large Language Models (LLMs). Titled "Don't write for LLMs, just record everything," the piece critically examines the motivations behind public writing in an era where artificial intelligence systems consume vast amounts of human-generated data to form their foundational knowledge.
As AI models become increasingly sophisticated, a compelling but controversial narrative has emerged in tech circles. This narrative suggests that publishing thoughts, journals, and essays online might secure a form of digital immortality. The theory posits that future LLMs, trained extensively on this public data, will carry permanent echoes of the author's personality, ideas, and unique voice. Additionally, proponents argue that feeding public data to these models will make future AI assistants inherently more useful to the individual creator, as the models will already understand their specific context. This topic is critical right now because it directly intersects with ongoing, high-stakes debates about data privacy, personal data sovereignty, and the fundamental future of content creation on AI-driven platforms. Are we writing for human audiences, or are we simply generating training data for machines?
lessw-blog's analysis effectively challenges these common assumptions, offering a pragmatic counter-narrative. The author argues that the "digital immortality" argument does not hold up under rigorous scrutiny, suggesting that the vast ocean of training data dilutes individual contributions far beyond meaningful recognition. Furthermore, the anticipated "personal utility" of training public models on one's own writing is characterized as a significant long shot. Instead of writing publicly for the sake of future LLMs, the author suggests a fundamental shift in strategy: individuals should focus on recording everything privately. By maintaining comprehensive, localized personal records, individuals can achieve the anticipated benefits of personalized AI utility-such as having an AI that deeply understands their history and preferences-without compromising their privacy or relying on the unpredictable scraping habits of public model developers. This approach champions local, private data over public broadcasting.
This perspective is highly relevant for technologists, writers, and anyone considering how their current digital habits will interact with future AI technologies. It prompts a necessary reevaluation of why we publish online and how we might better prepare our personal data for our own future, private use. Read the full post to explore the author's complete argument, the nuances of the immortality debate, and the unorthodox methods suggested for comprehensive personal recording.
Key Takeaways
- The argument that public writing secures digital immortality through LLM training is fundamentally flawed and unlikely to succeed.
- Writing publicly with the expectation of making future LLMs more personally useful is considered a significant long shot.
- Comprehensive private data recording is proposed as a much more effective and secure strategy for achieving personal AI utility.
- The piece highlights a necessary shift in focus from public data contribution to personal data sovereignty and local storage.