Idaho buckeye21h ago
David Chisnall (*Now with 50% more sarcasm!*)

I was just in a meeting where someone used a thing called Fathom to get an 'AI' summary of the meeting. Aside from some understandable typos arising from not understanding terms of art and replacing them with common English words, one of the key points that it concluded was that A was faster than B. It reached this conclusion because it missed one of the digits in the time for A. This completely inverted the key takeaway from one important section of the meeting.

Do not use plausible-nonsense generators for anything important.

Mar 30 at 11:45amWeb
Show threadVolker Stolz1d ago
@david_chisnall I guess it was easy to spot when everyone reading it was surprised that x86_64 came out ahead…
Show threadTurre22h ago

@david_chisnall I have many similar experiences from folks using Gemini meeting notes.

And as with everything with this LLM bullshit, the notes appear legit on a quick glance: blah blah agreed to blah blah and then blah. You only know it completely missed the most important thing in the meeting, or reversed it, if you actually were there.

Show threadTroy21h ago
@Turre @david_chisnall I’ve yet to see a Gemini summary that’s correct or useful
Show threadKiric20h ago
@david_chisnall I've experienced similar bullshittery from WebEx. I clearly and firmly stated that another person was the owner of a doc and not myself. That person agreed in the next utterance. The summary had it backwards. 
Show threadLori M Olson20h ago

@david_chisnall @briankrebs I’ve used it, and this is my experience. But I also copied that summary into meeting minutes and fixed the errors before sharing.

It’s about as good as letting an intern do the meeting minutes. Needs supervision. Wish it learned like interns though.

Show threadXCM16h ago

@david_chisnall
Another risk is that these LLMs can be easily influenced to give more weight to an opinion rather than others.

https://www.schneier.com/blog/archives/2025/11/ai-summarization-optimization.html

AI Summarization Optimization - Schneier on Security

These days, the most important meeting attendee isn’t a person: It’s the AI notetaker. This system assigns action items and determines the importance of what is said. If it becomes necessary to revisit the facts of the meeting, its summary is treated as impartial evidence. But clever meeting attendees can manipulate this system’s record by speaking more to what the underlying AI weights for summarization and importance than to their colleagues. As a result, you can expect some meeting attendees to use language more likely to be captured in summaries, timing their interventions strategically, repeating key points, and employing formulaic phrasing that AI models are more likely to pick up on. Welcome to the world of AI summarization optimization (AISO)...

Schneier on Security
Show threadHayley 12h ago
@david_chisnall friends and I found an AI slop performance evaluation which missed the m in ms, leading to a concerningly very large speedup
Show threadsamaaron5h ago

@david_chisnall Not sure this is a strong argument against plausible -nonsense generators.

Feels more like a shout-out for the importance of verifying important things.

A real human taking notes might also misplace a decimal point or get an arrow inverted on a flow diagram.

Show threadLouis Dureuil5h ago

@david_chisnall the worst with this is that people are subjecting you to their use of the tool.

I wish I could opt out from being summarily summarized

Show threadPhil4h ago
@david_chisnall Seems the A"I" summary thing the company I work at is not that different.
I regularly read the summary of a meeting I don't join (and don't have to). There is no regular word for a customer's name, so something else is used every time, including the name of a colleague.
It has also happened that people who attended were surprised at something that was supposedly said in that meeting (and they didn't get the summary), but now, it's in there and if somebody checks it, it's a fact.
Show threadPxl Phile3h ago

@david_chisnall the last time I accidentally used this resulted in:

"A said x
B said y
This resulted in a spontaneous discussion."

No shit, Sherlock.

What was the consensus? 🪿
WHAT WAS THE CONSENSUS OF THE DISCUSSION?! 🗡️🪿💨