Back
ML Safety — Center for AI Safety Research Hub
webmlsafety.org·mlsafety.org/
The official hub of the Center for AI Safety's ML Safety initiative; useful as an entry point for researchers new to the field or seeking structured resources, courses, and community connections.
Metadata
Importance: 62/100homepage
Summary
MLSafety.org is the homepage for the ML Safety research community, a project of the Center for AI Safety (CAIS), organizing resources, education, courses, and competitions focused on reducing risks from AI systems. It frames ML safety across four pillars: Robustness, Monitoring, Alignment, and Systemic Safety. The site serves as a hub for researchers and non-technical audiences seeking to engage with AI safety work.
Key Points
- •Defines four core ML safety research areas: Robustness, Monitoring, Alignment, and Systemic Safety with concrete subtopics.
- •Hosts the ML Safety Course, newsletter, seminar series, and SafeBench competition for benchmark development.
- •Project of the Center for AI Safety (CAIS), connecting researchers via Slack, Twitter, and events like NeurIPS socials.
- •Covers technical topics including adversarial robustness, interpretability, value learning, power aversion, and cooperative AI.
- •Provides funding opportunities, reading resources, and community infrastructure for the AI safety research ecosystem.
Cited by 1 page
| Page | Type | Quality |
|---|---|---|
| Multipolar Trap Dynamics Model | Analysis | 61.0 |
1 FactBase fact citing this source
Cached Content Preview
HTTP 200Fetched Mar 20, 20264 KB
[SafeBench](https://www.mlsafety.org/safebench)
# ML Safety
The ML research community focused on
reducing risks from AI systems.
## What is ML Safety?
ML systems are rapidly increasing in size, are acquiring new capabilities, and are increasingly deployed in high-stakes settings. As with other powerful technologies, the safety of ML systems should be a leading research priority. This involves ensuring systems can withstand hazards ( **Robustness**), identifying hazards ( **Monitoring**), reducing inherent ML system hazards ( **Alignment**), and reducing systemic hazards ( **Systemic Safety**). Example problems and subtopics in these categories are listed below:
### Robustness
Adversarial Robustness, Long-Tail Robustness
### Monitoring
Anomaly Detection, Interpretable Uncertainty, Transparency, Trojans, Detecting Emergent Behavior
### Alignment
Honesty, Power Aversion, Value Learning, Machine Ethics
### Systemic Safety
ML for Improved Epistemics, ML for Improved Cyberdefense, Cooperative AI
[Learn more](https://www.mlsafety.org/resources#readings)
## ML Safety Projects
We organize AI/ML safety resources and education for researchers and non-technical audiences.
[\\
\\
Seminar Series (Coming Soon)\\
\\
](https://mlsafety.webflow.com/events#virtual-events) [\\
\\
The Newsletter](https://newsletter.mlsafety.org/) [\\
\\
NeurIPS 2023 Social](https://www.mlsafety.org/events/neurips/2023) [\\
\\
Competitions and Prizes\\
\\
](https://safe.ai/competitions) [\\
\\
ML Safety Course](https://course.mlsafety.org/)
## Get Connected
Stay in the loop and exchange thoughts and news related to ML safety. Join our [slack](https://join.slack.com/t/ml-safety-workspace/shared_invite/zt-1k73lo3ap-mGEykgb8crvuTY_10lBclQ) or follow one of the accounts below.
[Follow](https://twitter.com/ml_safety)
**ML Safety @ml\_safety**
General Announcements
[Follow](https://twitter.com/topo
... (truncated, 4 KB total)Resource ID:
48fda4293ccad420 | Stable ID: ODg3MGQyNG