Back
CSIS: US Vision for AI Safety
webCredibility Rating
4/5
High(4)High quality. Established institution or organization with editorial oversight and accountability.
Rating inherited from publication venue: CSIS
This CSIS interview provides insight into the US government's institutional approach to AI safety through AISI, relevant for understanding how federal policy is translating AI safety concerns into concrete regulatory and evaluation frameworks as of 2024.
Metadata
Importance: 62/100organizational reportanalysis
Summary
A CSIS interview with Elizabeth Kelly, Director of the US AI Safety Institute (USAISI), discussing the US government's strategic approach to AI safety, the role of AISI in evaluating frontier AI models, and international coordination on safety standards. Kelly outlines the institute's priorities including developing evaluation frameworks, engaging with AI developers, and building global partnerships to manage AI risks.
Key Points
- •The US AI Safety Institute serves as the primary federal body for developing technical standards and evaluation tools for frontier AI model safety.
- •AISI focuses on pre-deployment testing of advanced AI systems in collaboration with leading AI developers to identify potential risks before public release.
- •International coordination is a key priority, with AISI working with counterparts like the UK AI Safety Institute to harmonize safety evaluation approaches.
- •Kelly emphasizes the need to balance promoting AI innovation with managing safety risks, avoiding overly prescriptive regulation while maintaining meaningful oversight.
- •The institute is building capacity in red-teaming, evaluation methodologies, and technical standards to inform both domestic policy and global AI governance norms.
Cited by 2 pages
| Page | Type | Quality |
|---|---|---|
| US AI Safety Institute (now CAISI) | Organization | 91.0 |
| AI Safety Institutes (AISIs) | Policy | 69.0 |
Cached Content Preview
HTTP 200Fetched Apr 9, 202653 KB
The U.S. Vision for AI Safety: A Conversation with Elizabeth Kelly, Director of the U.S. AI Safety Institute
Skip to main content
The U.S. Vision for AI Safety: A Conversation with Elizabeth Kelly, Director of the U.S. AI Safety Institute
Transcript
— July 31, 2024
Available Downloads
Download the Transcript
341kb
Image
The U.S. Vision for AI Safety: A Conversation with Elizabeth Kelly, Director of the U.S. AI Safety Institute
July 31, 2024 • 10:00 – 11:30 am EDT
Hosted by Wadhwani AI Center
This transcript is from a CSIS event hosted on July 31, 2024. Watch the full video here.
Gregory C. Allen: Good morning. I’m Gregory Allen, the director of the Wadhwani Center for AI and Advanced Technologies here at the Center for Strategic and International Studies.
AI safety has been one of the most incredible topics – incredibly talked about topics in AI policy for several years now. And we have the perfect person to outline the U.S. vision for AI safety. And that is Elizabeth Kelly, the inaugural director of the U.S. AI Safety Institute, a new institution that was created just in the past few months and is going to be the foundation for setting a lot of the policy, guidance, and other documents around AI safety here in the United States. And also has an important role to play internationally.
Elizabeth Kelly, thank you so much for joining us today.
Elizabeth Kelly: Thank you for having me.
Mr. Allen: We’re going to get really into the weeds on what the AI Safety Institute is up to, but before we go there, I wanted to hear a little bit about you and how you came to occupy the role as the inaugural director of the AI Safety Institute.
Ms. Kelly: So, I am thrilled to lead the amazing interdisciplinary team that we have at the U.S. AI Safety Institute. I’ve spent most of my career in public service, including most recently at the White House National Economic Council, where I was the – one of the driving forces behind the AI executive order, and helped lead a lot of our AI policy engagement globally. Before that, I was in the private sector where I helped start and then scale a startup, from seed stage through acquisition.
I think that this role is really a combination of those two things. It’s how do you create a nimble startup that is able to tackle big things, and how do you operate inside a government with all of the complexities
... (truncated, 53 KB total)Resource ID:
887c0e40caa39cf4 | Stable ID: sid_a9xs1JJs6s