What Does Trust in AI Look Like?
Andy Jordan is President of Roffensian Consulting S.A., a Roatan, Honduras-based management consulting firm with a comprehensive project management practice. Andy always appreciates feedback and discussion on the issues raised in his articles and can be reached at [email protected]. Andy's new book Risk Management for Project Driven Organizations is now available.
Last month I wrote an article about Anthropic’s recent announcement that they were going to introduce a watermark to help identify content generated by their Claude AI engine. There were some great comments raising some interesting issues, so I thought it was worth highlighting some of those issues here. Hopefully this piece will also generate further comments and debate!
The details matter
A couple of the comments on the original article raised the point that not all AI content is created equal. There is a big difference between a copy-and-paste of an entire response to a query, and the refinement and editing of a largely complete human generated piece of content. In the first scenario, an AI watermark is clearly appropriate; in the second, if the content is flagged as AI, that could be misleading—potentially with significant negative consequences.
An example from the comments is if a human user asks AI to assist in refining a résumé. That could result in a rejection by a potential employer’s software because the incorrect assumption is that if there is an AI watermark, then that means the résumé is 100% AI generated.
Anthropic has published some more details since I wrote that original piece. But in truth, it’s still not completely clear how that situation will be addressed. The way that the watermark is applied …
Please log in or sign up below to read the rest of the article.
|
"You're talking to someone who really understands rock music." - Tipper Gore |
