Another Day, Another AI Fail
It's easy to point a finger, and criticise anything or anyone. I try to avoid doing this as it's not big, and it's not clever. However, I don't apply that standard to AI. I am not a fan, generally.
Having said that, I recently had an AI success when I tested it with a real-world challenge. But let's not dwell on that one.
I sit on a executive board for a large charitable trust. As one of the non-exec directors I help oversee the high-level running of the charity, and offer advice and direction as needed. Part of that role includes receiving many reports from working groups and other committees.
Yesterday's Board meeting included a set of minutes from one such sub-group. It had been automatically generated by AI from the transcript of a Teams call. At first glace it looked impressive.
However, when we came to discussions around the group's activity, an amendment was proposed to the draft minutes around accuracy. The minutes stated that Anne (name changed) was one of the attendees at the meeting. She had also raised two points, and had proposed an action item. However, Anne, who was in this Board Meeting, wanted to point out that she did not attend that meeting as she was busy, and had therefore given no input to any proceedings!
Quite where the AI tool came up with the context is a mystery.
We reminded everyone they are free to use AI tools, but a human must always review the AI outputs for accuracy.
People are too quick to assume the AI sausage machine churns out accurate information.
Leave a Comment; Reply via the Fediverse; or send a message if you have replied with your own blog post and I will mention it here.