Executive Summary
- Long policy documents are unreliable for governing AI agents
- Experts cite issues with context models, quantization, and sampler quality
- Local inference may be the solution to common LLM defects
The Buzz Score
The Internet’s Verdict: 70% Hyped, 30% Skeptical
Understanding the Issue
Experts agree that long policy documents are not effective in governing AI agents. One expert notes,
This is a problem with long context models. To put it as simple and as bluntly as possible: just because they claim you can use 1M tokens in your context doesn’t mean its true and you should do that.
Real-World Experience
Another expert shares their anecdotal experience,
Yeah checks out with my anecdotal experience with Claude. It is pretty great at following instructions – for about 10 minutes, after which it seems to ignore things I told it before.
Solutions and Limitations
Local inference is proposed as a solution to the common defects in LLMs. However, the limitations of human working memory and the complexity of real-world policies also play a role in the failures of AI governance.
Focus Keyword: AI Governance
