Summary of Can Go Ais Be Adversarially Robust?, by Tom Tseng et al.
Can Go AIs be adversarially robust?
by Tom Tseng, Euan McLean, Kellin Pelrine, Tony T. Wang, Adam Gleave
First submitted to arxiv on: 18 Jun 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Artificial Intelligence (cs.AI); Machine Learning (stat.ML)
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 The proposed research aims to investigate the robustness of Go AIs against adversarial attacks by introducing natural countermeasures. Prior studies have shown that superhuman Go AIs can be defeated by simple strategies, particularly “cyclic” attacks. This paper explores three defenses: adversarial training on hand-constructed positions, iterated adversarial training, and modifying the network architecture. The results demonstrate that while some of these defenses protect against previously discovered attacks, none withstand newly trained adversaries. Furthermore, most reliably effective attacks found by these adversaries are variations of the same overall class of cyclic attacks. This study highlights two key gaps: efficient generalization of defenses and diversity in training. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Artificial intelligence systems can be vulnerable to clever tricks called “adversarial attacks”. A popular game like Go is often used as a test bed for AI research, but surprisingly, even superhuman Go AIs can be beaten by simple tactics. In this study, researchers tried to strengthen these AIs against such attacks by using different techniques. Unfortunately, they found that none of the methods worked well enough to protect against newly developed tricks. This shows that building robust AI systems is still a challenging task, especially in areas like Go where the AI has an edge. |
Keywords
» Artificial intelligence » Generalization
