Hacker News Reader

Comment by adamddev1 | Hacker News Reader

These AI Note-Takers can also mangle the summaries. A few weeks ago I read anecdotes here about a doctor getting completely wrong information about a patient, and a manager getting upset because he was depending on a summary of something a client never said or agreed to, but the AI summary said he did. These things are downright dangerous.

Now come the replies saying, "as if human note-takers never made mistakes!"

View on HN

Actually I think one of the main values of note taking isn't that you get notes, it is that engaging your brain to summarize points makes you understand and remember the points better, even if you throw your notes away right after.

It is like writing: One of the main purposes of writing (for me and many others) is that it forces you to structure your thoughts. You could give that task to AI, get the resulting text, but your loss is the lack of clarity you could have gained from doing it yourself.

This is definitely a thing that's been studied in schools:

https://news.ycombinator.com/item?id=20268974 (2019, 238 comments)

However, note taking in a business sense is usually different than note taking in an academic sense. Business meetings are more often used to make decisions, unlike academics where lectures are intended to disseminate information.

I get the difference between business and academic note taking, but I still think it is benefitial to do it yourself in a business meeting, because as you write down the notes you may realize a point hasn't been made as clearly as you wish it was made, which allows you to intervene.

I think that's a good case for shared notes, that all the participants can see and agree on. However the AI notes I've seen are compiled long after the meeting is done.

Moreover, they capture all the smalltalk perfectly, but tend to plausibly mishear the parts that matter most - terminology, abbreviations, names of people. There's no mystery why this happens, of course, but the only transcript I trust is a transcript I personally checked by listening to the audio and fixing mistakes (doable at 1.75x speed, so not that bad). I catch outrageous mistakes sometimes, mistakes that completely change the meaning of what's being said (up to capturing the opposite of what's being said). So, all in all, even though modern speech2text models are really impressive, I am not sure the utility of _completely_ automated transcribers outweighs their dangers today.

It would be nice if they would at least say __unintelligable__ when they couldn't get things, but alas. I also notice it with YouTube auto generated subtitles and translations. They misrepresent SO much.

Reply on news.ycombinator.com