AI risk mitigation or designing "responsible AI"
goes a step beyond the latest and most incredible algorithms. These AI-induced risks
can include legal risk, breaking the law, organizational risk, losing
employees, and the chance of suffering a PR disaster. How do we really know if
the models are doing what they were supposed to? Let's not forget that some
models are incredibly complex, approaching trillions of parameters. Even
decades ago, when expert systems were the leading technology, we ran into
similar problems. At some level of complexity, it isn't easy to audit the AI
sufficiently. Achieving #explainability or #bias mitigation might be more
challenging to implement than the AI algorithms.
#aiexplainability #biasmitigation #aiethics #ai
#airesponsibility
https://www.technologyreview.com/2021/01/15/1016183/ai-ethics-startups/
No comments:
Post a Comment
Note: Only a member of this blog may post a comment.