Proving safety for narrow AI outputs - Evan Miyazono

Proving safety for narrow AI outputs
By ​Evan Miyazono, Founder of Atlas Computing
Guaranteed Safe AI Seminars, July 2024
​User demand for new AI capabilities is growing even as risks from foreseeable AI capabilities become clearer. To navigate this, we are identifying domains where AI can provide better capabilities with quantitative guarantees against objective safety criteria. Evan presents progress at Atlas Computing, mapping a path to the use of language models to generate software, formal specifications, and proofs that the software satisfies the specification
Slides: docs.google.co...
---
About the series:
​​​​The monthly seminar series on Guaranteed Safe AI brings together researchers to advance the field of building AI with high-assurance quantitative safety guarantees.
www.horizoneve...