Real word AI challenges cannot always be predicted in pre-deployment tests!
- 1 day ago
- 1 min read
National Institute of Standards and Technology (NIST) just realeased a report reiterating what I have been saying for a long time. Testing AI systems in alpha and beta modes is not equivalent to real-world deployment challenges. You must monitor and measure for governance post-deployment which is challenging given the scale and complexity of adoption! We often see mission drift beacuse of values and commercial interests conflicting, or contract subclauses (especially military once) creating tensions.
Infographic via @Google NotebookLM but suggest read the report!
Also adding the link of the report: https://lnkd.in/dD5D9YYG
Comments