Dear Editor,

Thanks to Christopher Kelley for his recent, excellent article on the possibility of using AI to serve as a future secretarial tool, e.g., taking minutes for official public meetings.

Christopher is right to be a little dubious! How can we trust a machine that makes mistakes, as AI technology certainly does, to represent what actually transpires in a meeting?

I offer two potential answers to Christopher. First, the AI technology we have right now is not ready for full-trustworthiness. Most AI we hear about today is generative - producing new language and analysis from prior databases. It must be trained in order to learn proper context and to scrutinize outlier databases carefully. Until this training “learning curve” is more developed, we are all in early days for AI technology and must be careful.

Another type of AI is transcriptive. This tool does not create new language or analysis, it merely captures the spoken language it hears and, possibly, edits it for clarity. It does this by using an editing scan prescribed by the human who is responsible for the accuracy of the presented final language in a document.

Such an editing scan might be one that searches only for “Action” items undertaken by those participating in a meeting. Other “context” scans might be performed as needed.

Crucially, the raw captured database itself would never be tampered with or deleted. So in case of an accuracy inquiry into a contested AI transcription, the original content is available.

However, the transcriptive AI tool must also be trained, often in the exact type of secretarial work that it is being asked to record. There are simply too many “irregulars” in languages (especially English!) for a non-contextual recorder to always understand correctly. Training will solve this problem.

In any event, one thing about public Board Secretaries of all types will never change: the actual, human Board Secretary will always have to formally certify that the presented minutes of any public meeting are accurate, even when it’s machine data.

Finally, news reporting by a live person, such as Christopher, in my opinion is also not replaceable by AI, because of the overwhelming contextual nature of journalism itself. This kind of writing will be eternally safe from machines, thank goodness.

Larry Tasaday

Eldora