Tracking (Expert/Influential) Predictions about AI
The author proposes creating a website to track the prediction track records of AI experts, including those who don't use existing forecasting platforms, to improve accountability and cut through noise. The goal is to provide a single place to see anyone's full prediction history, focusing on influential voices.
Introduction Tracking the track records of AI experts is important for understanding their past predictions and their effects. | 1:14Explained | |
Incentive Problem The current incentive structure encourages vague predictions, making it difficult to discern accurate forecasts from noise. | 1:21Explained | |
Proposed Solution Imprecise predictions should be tracked by either clarifying them or evaluating them based on perceived accuracy and uncertainty. | 1:22Explained | |
Platform Concept A website will aggregate predictions from existing platforms and the web, allowing users to view an expert's complete track record. | 1:34Explained | |
Distinction from Existing Platforms This platform tracks individuals who do not use forecasting platforms and offers an easy-to-use interface for viewing prediction histories. | 1:18Explained | |
Usage Note The site should prioritize inside views to avoid deference cascades and encourage independent assessment of predictions. | 1:26Explained | |
Call for Collaboration The author is seeking collaborators to build this tracking platform over a weekend. | 1:05Explained | |
Importance for Epistemics This tracking system is crucial for AI-related epistemic practices, particularly in anticipation of transformative AI. | 1:05Explained | |
Potential Risk While a risk exists of individuals misusing their track record, this can be managed by avoiding over-reliance on a few highly-rated forecasters. | 1:17Explained |