AI's integration into scientific workflows has accelerated dramatically.
As some want to conduct scientific research almost entirely autonomously, how do we validate the accuracy and reliability of inferences made by or with AI?
Security Challenges:
-open source (often unverified) easy to acquire data sets are prevalent.
-open source software is the norm, with sprawling 3rd party dependencies.
- AI attacks are hard to detect.
-AI security issues are serious but less visibly dramatic.
The MATRIX Lab is focusing on helping scientists and infrastructure providers solve the challenges of AI Security for Science.
As some want to conduct scientific research almost entirely autonomously, how do we validate the accuracy and reliability of inferences made by or with AI?
Security Challenges:
-open source (often unverified) easy to acquire data sets are prevalent.
-open source software is the norm, with sprawling 3rd party dependencies.
- AI attacks are hard to detect.
-AI security issues are serious but less visibly dramatic.
The MATRIX Lab is focusing on helping scientists and infrastructure providers solve the challenges of AI Security for Science.