Microsoft: ‘Skeleton Key’ attack unlocks the worst of AI • The Register
With the exploitative creation of guardrails and how quickly users can work around them, it is time to curate data to eliminate the need for guardrails.
Related Posts
Supercomputing Around the World: A look at the Top500 as of 11/2020
With the release of the Nov 2020 Top500 list of supercomputers, it is time again to have a look at the distribution of supercomputers worldwide. Notable new entries…
Progress on AI racial issues
https://www.theregister.com/2022/05/01/ai_models_racist/
Four Steps to Ensure GenAI Safety and Ethics
Four Steps to Ensure GenAI Safety and Ethics (hpcwire.com) ContractPodAI has a good set of 4 basic principles to consider. While these do not address things like training…
AI ethics: a call to faculty
https://dl.acm.org/doi/10.1145/3478516. Integrating ethics into artificial intelligence education and development.