A Roadmap for Robust End-to-End Alignment

2018/09/04 by Lê Nguyên Hoang, Hoang, Lê Nguyên
Computer Science · #Artificial Intelligence (cs.AI) #Complexity and Algorithms in Graphs #FOS: Computer and information sciences #Machine Learning and Algorithms #Optimization and Search Problems

paper · pdf · doi:10.48550/arxiv.1809.01036

openalex publication_date 2018/09/04 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28

Abstract

This paper discussed the \it robust alignment problem, that is, the problem of aligning the goals of algorithms with human preferences. It presented a general roadmap to tackle this issue. Interestingly, this roadmap identifies 5 critical steps, as well as many relevant aspects of these 5 steps. In other words, we have presented a large number of hopefully more tractable subproblems that readers are highly encouraged to tackle. Hopefully, this combination allows to better highlight the most pressing problems, how every expertise can be best used to, and how combining the solutions to subproblems might add up to solve robust alignment.

Citations

• Efficient Estimation of Word Representations in Vector Space
• Experimental evidence of massive-scale emotional contagion through social networks
• Neural Architecture Search with Reinforcement Learning
• Rank Analysis of Incomplete Block Designs: I. The Method of Paired Comparisons
• AI Safety Gridworlds
• Exposure to opposing views on social media can increase political polarization
• Recurrent World Models Facilitate Policy Evolution
• Supervising strong learners by amplifying weak experts

Related

• On the Essence and Prospect: An Investigation of Alignment Approaches for Big Models
• Superalignment with Dynamic Human Values
• AI Alignment: A Comprehensive Survey
• A Practical Analysis of Human Alignment with *PO
• Insights into Alignment: Evaluating DPO and its Variants Across Multiple Tasks