Summary of Towards a Law Of Iterated Expectations For Heuristic Estimators, by Paul Christiano et al.
Towards a Law of Iterated Expectations for Heuristic Estimators
by Paul Christiano, Jacob Hilton, Andrea Lincoln, Eric Neyman, Mark Xu
First submitted to arxiv on: 2 Oct 2024
Categories
- Main: Artificial Intelligence (cs.AI)
- Secondary: Machine Learning (cs.LG)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary This paper by Christiano et al. (2022) introduces a novel concept called heuristic estimators, which are hypothetical algorithms that estimate mathematical expressions from given arguments. A heuristic estimator takes in a mathematical expression and a formal argument, then outputs an estimated value of the expression. The authors propose two informal principles for these estimators: they shouldn’t be able to predict their own errors, and they should satisfy iterated estimation and error orthogonality properties. These principles are formalized as mathematically rigorous properties that ideal heuristic estimators should possess. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper is about a new way to estimate mathematical expressions using computer algorithms. The idea is called “heuristic estimator” and it takes in a mathematical equation and some extra information, then tries to guess the answer. The authors think that these estimators shouldn’t be able to predict how well they’re doing, which seems fair. They also propose two rules for good heuristic estimators: they should be able to estimate their own estimates correctly, and the errors in their estimates should be unrelated to the actual values. |
