Back
Even Superhuman Go AIs Have Surprising Failure Modes
webCredibility Rating
4/5
High(4)High quality. Established institution or organization with editorial oversight and accountability.
Rating inherited from publication venue: FAR AI
Landmark FAR.AI study showing that even superhuman AI systems can have exploitable blind spots; widely cited as empirical evidence that capability does not imply robustness, relevant to AI safety evaluation and adversarial testing research.
Metadata
Importance: 75/100blog postprimary source
Summary
FAR.AI researchers demonstrate that superhuman Go AI systems (including KataGo) can be reliably defeated by an adversarial attack algorithm that exploits unexpected blind spots, despite these AIs vastly outperforming human players. The work reveals that high capability in a domain does not guarantee robustness, with implications for AI safety and evaluation methodology.
Key Points
- •Adversarial testing found simple, exploitable failure modes in superhuman Go AIs that humans can recognize but the AI cannot defend against.
- •The attack algorithm is cyclic and does not require human-level Go knowledge, suggesting capability gaps are not just about domain expertise.
- •Findings highlight that superhuman performance on standard benchmarks does not imply robustness against adversarial strategies.
- •Results have broad AI safety implications: deployed systems may have hidden vulnerabilities not revealed by normal evaluation.
- •Open-source code and project page released to enable further adversarial robustness research.
Cited by 1 page
| Page | Type | Quality |
|---|---|---|
| Deep Learning Revolution Era | Historical | 44.0 |
Cached Content Preview
HTTP 200Fetched Mar 20, 202631 KB
[We updated our website and would love your feedback!](https://www.far.ai/about/website-feedback)
[](https://www.far.ai/)
# Even Superhuman Go AIs Have Surprising Failure Modes
[Full PDF](https://openreview.net/forum?id=LX3VAhXNTw) [Project](https://goattack.far.ai/) [Source](https://github.com/AlignmentResearch/go_attack)
Citation
July 15, 2023
[Adam Gleave](https://www.far.ai/about/people/adam-gleave)
[Euan McLean](https://www.far.ai/about/people/euan-mclean)
[Kellin Pelrine](https://www.far.ai/about/people/kellin-pelrine)
[Tony Wang](https://www.far.ai/about/people/tony-wang)
[Tom Tseng](https://www.far.ai/about/people/tom-tseng)
Summary
**FOR IMMEDIATE RELEASE**
### **FAR.AI Launches Inaugural Technical Innovations for AI Policy Conference, Connecting Over 150 Experts to Shape AI Governance**
WASHINGTON, D.C. — June 4, 2025 — FAR.AI successfully launched the [inaugural Technical Innovations for AI Policy Conference](https://far.ai/events/event-list/technical-innovations-for-ai-policy-2025), creating a vital bridge between cutting-edge AI research and actionable policy solutions. The two-day gathering (May 31–June 1) convened more than 150 technical experts, researchers, and policymakers to address the most pressing challenges at the intersection of AI technology and governance.
Organized in collaboration with the [Foundation for American Innovation (FAI)](https://www.thefai.org/), the [Center for a New American Security (CNAS)](https://www.cnas.org/), and the [RAND Corporation](https://www.rand.org/), the conference tackled urgent challenges including semiconductor export controls, hardware-enabled governance mechanisms, AI safety evaluations, data center security, energy infrastructure, and national defense applications.
"I hope that today this divide can end, that we can bury the hatchet and forge a new alliance between innovation and American values, between acceleration and altruism that will shape not just our nation's fate but potentially the fate of humanity," said Mark Beall, President of the AI Policy Network, addressing the critical need for collaboration between Silicon Valley and Washington.
Keynote speakers included Congressman Bill Foster, Saif Khan ( [Institute for Progress](https://ifp.org/)), Helen Toner (CSET), Mark Beall ( [AI Policy Network](https://theaipn.org/)), Brad Carson ( [Americans for Responsible Innovation](https://ari.us/)), and Alex Bores ( [New York State Assembly](https://nyassembly.gov/)). The diverse program featured over 20 speakers from leading institutions across government, academia, and industry.
Key themes emerged around the urgency of action, with speakers highlighting a critical 1,000-day window to establish effective governance frameworks. Concrete proposals included Con
... (truncated, 31 KB total)Resource ID:
3cc31d303002fcbb | Stable ID: ODM2NzhiOT