Guardian Angels: LLM Personalization for Productivity and Security · Gwern.net
I propose an approach for highly personalized LLMs, for near-future productivity gains and personal info/cybersecurity against increasingly powerful LLMs: they should, in the spirit of uploading, try to emulate the user’s values and preferences in order to amplify the principal—not replace them. I discuss a package of techniques and proposals to accomplish such ‘guardian angels’; dynamic evaluation of LLMs combined with active learning and elicitation and heavy inner-monologue search/data-augmentation.
--- title: "Guardian Angels: LLM Personalization for Productivity and Security" description: "I propose an approach for highly personalized LLMs, for near-future productivity gains and personal info/cybersecurity against increasingly powerful LLMs: they should, in the spirit of uploading, try to emulate the user’s values and preferences in order to amplify the principal—not replace them. I discuss a package of techniques and proposals to accomplish such ‘guardian angels’; dynamic evaluation of LLMs combined with active learning and elicitation and heavy inner-monologue search/data-augmentation
saved by
- Justin Wang
- Ruozhen Gong
- Claire Wang
- Freeman Jiang
- Rishi Kothari
- Yixiong Hao
- Uzay Girit
- Lydia Nottingham
- Vincent Cheng
- Timothy Kostolansky
- Fang Ting
- Abhay Sheshadri
related reading
- LLM Daydreaming · Gwern.netgwern.net
- Quantifying Truesight With SAEs · Gwern.netgwern.net
- LLM Powered Autonomous Agents | Lil'Loglilianweng.github.io
- Human-like Neural Nets by Catapulting · Gwern.netgwern.net
- Patterns for Building LLM-based Systems & Productseugeneyan.com
- Writing for LLMs So They Listen · Gwern.netgwern.net
- Lean Software Scaling Laws · Gwern.netgwern.net
- llm assistant personas seem increasingly incoherent (some subjective observations) — LessWronglesswrong.com
- Simulated Users & Sad LLMs1a3orn.com
- Taking LLMs Seriously (As Language Models) — LessWronglesswrong.com
- Modifying LLM Beliefs with Synthetic Document Finetuningalignment.anthropic.com
- What We’ve Learned From A Year of Building with LLMs – Applied LLMsapplied-llms.org