Artificial Meetings: Possible merits of a terrible idea

Amidst all the angst about AI safety, LLM hallucinations, and bizarre chatbot behavior, it is easy to lose sight of some basic scenarios that seem pretty compelling knowledge worker scenarios:

  1. Improve Search results by improving understanding of web pages and summarizing search results
  2. Summarize large documents and e-mails
  3. Prioritize and sort your Inbox
  4. Convert bullet lists (presentations) into text documents
  5. Summarize text documents into presentations

A more advanced scenario would be if the AI monitored your work, your calendar, the mails and documents you produced and consumed, your check-ins, and then wrote at least a draft of your weekly status report (and even if you hate sending status reports, see my post about Sending Status Report to Yourself).

All of these offer an opportunity to save time and remove “busy work.” Of course, I’m not so naïve to think that there isn’t a real risk of the opposite happening: when mail and documents are easy to produce, it is likely your inbox could be flooded with new things to read.

That aside, the thing that really took up a lot of my time at work was meetings. So, what could AI do for meetings?

The obvious thing is to have AI take notes and write summaries. This could be especially valuable if one has to miss the meeting.

But let’s take that further. What I really want is not to attend so many meetings.

How about if you could train a bot (by reading all your documents and perhaps chatting with you) to represent your positions in the meetings? Then you could do something else and send the bot. Once everybody does that, no more meetings!

The nerd comedy “Real Genius” has a scene like this: https://youtu.be/wB1X4o-MV6o

Okay, obviously a terrible idea, especially once the AIs start making decisions out of the meetings, leading directly to Charles Stross’s Vile Offspring. And just because it’s a terrible idea, I guarantee there is at least one startup working on it right now.

But the more I thought about it, the more I liked it for a simple reason.

Maybe we should all keep this criteria in mind when we decide whether we should call a meeting or attend one: is it valuable enough that a bot couldn’t do the meeting?