Yeah, how can it know what part of the meeting I want to remember? Plus sometimes what I write is thoughts / action items that occur to me because of the meeting, but were not mentioned during the meeting.
I would like a tool like this that made a summary/notes and included the full transcript, so they could be searched. This would be good for people who attended and those who didn't, or for seeing if something was discussed a while ago.
Ideally the same system would work on zoom calls, in person meetings, and in meetings with both (some in person, some on zoom). For me the hybrid workplace is one that has the advantages of both, not one that is either/or.
AI will never replace jobs that require real human care or intent. We don't want an AI taking care of us when we're sick, though they'd probably be very good at it. Business notes may be a nice case, but notes taken by someone who cares about the project is going to be better than the AI's, even if the AI does a better job, because the care o the person taking the notes is evident.
> We don't want an AI taking care of us when we're sick
Debatable, what if...
- there are not enough doctors to see patients
- it meant I could be seen immediately vs having to wait 1+ hours
- my doctor had instant access to all medical knowledge and the wisdom of 1M+ hours of practice
- there is a pandemic where a robot is much easier to sterilize between patient visits
There are people already willing on a personal / social level to see an artificial doctor, doctors are already incorporating AI into their practice to help reduce errors, it's hard to believe it will never happen. The context of the visit or stay will play into acceptance
If the AI objectively does a better job at summarizing the meeting, then maybe someone who cares about the project should use the AI instead of taking notes by hand.
Eah statement in this boils down to a variation over an appeal to nature [1], in that each argument here is basically that an AI can't do a good job because it's not human / natural. While an AI might well not do a good job now, you've provided no meaningful argument for why it can't match - or surpass - people.
An actual argument might be that it can't fully replace my own notes because my own notes can, and sometimes do, include notes about my thoughts and opinions that the AI can't possibly know without being told. But that does not negate the possible value in terms of freeing me from taking notes about what was actually said during the meeting.
With respect to being taken care of: I want to be taken care of in whichever way produces the best outcome for me. If that means care handled by an AI, then I want the AI.
I'd gladly take an AI that can monitor my vitals and physical activity and automatically recommends a treatment option, especially if it's substantially cheaper than a person.
> After the meeting, jamie generates the summary for you in seconds
This is cool, but sometimes you want only a summary, sometimes you want a clear list of decisions and actions (with owners), and sometimes you literally want a word-for-word transcription of who said what.
I've seen wildly varying company cultures regarding their approach to note taking. I worked for a company once where the SVP and above folks were treated as kind of god-kings, where every word they spoke had to be written down and agonized over. Felt like those guys following Kim Jong-un around with little notebooks. Even something casually mentioned would become a real project with dozens of staff, just because it was uttered by a SVP during a meeting once. Other places I've worked barely took any notes, and your team's priorities and expected work was passed around by word of mouth like a terrible game of telephone. Everywhere is different!
Yeah, the act of physically writing (as in not typing) words on paper helps me to remember things better and distill the important parts in my head. Just reading someone else's notes is not as useful for me.
That's interesting - thanks for sharing! From what we've heard many people actually lose attention, if they have to take notes while focussing on the meeting
The act of writing notes with pencil and paper is a ritual that loads information into my brain. But I’m in my 30’s, and while I grew up around and constantly using computers, they were not so ubiquitous as to replace pencil and paper note taking ritual. I wonder what the next generation’s experience has been, is typing their ritual?
I’m in my 40s, but I often transcribed school notes from written to text, and the act of typing them really cemented it in my memory. I even recall flicking my fingers during recall, air typing words. In college I typed everything having a laptop, and I think it had similar value.
I guess the language models have an advantage here because they don't need to understand the TLAs and "internal lingo" of McCompany X, they can just fake it good enough (like everyone else does in orgs).
But in highly technically important meetings, y'all might want to proofread in case the wrong TLA gets used.
I mean, to be clear, these will be strictly less valuable than human-generated notes.
But on the other hand, I think they'll be strictly more valuable than no notes.
And as someone who has a terrible time trying to both note-take and participate in a meeting at the same time, this seems like it could be super helpful.
Huh. So I've never taken meeting notes. I guess you write down some things people said. Do you then refer back to that later, or is it a memory aid thing?
I don't feel like I'm missing anything, but equally the op has built a product to automate this, so it's got to be important in some contexts. Would like to know what I'm missing
I've always took notes because I cannot remember anything in detail after a weekend.
Taking notes doesn't have to be painful, for me it's just one `diary` command away and it will open/create the appropriate note file in my favorite editor:
Those should be sent out to all attendees of the meeting (or written in a place that's accessible to everyone). Beyond that, each person needs to take notes for themselves for other things they want to remember.
Most managers do all their work through meeting summaries/notes. It is usually a symptom to a larger problem given that if people just showed up to meetings that mattered to them, they would be driven to do the things needed to get done from said meeting.
> I dont know how it would be possible to build anything reasonably complex with groups of people without taking meeting notes.
I don't imagine it is. I've never seen anything reasonably complex successfully built by way of meeting, notes or not. The actual work always happens when ideas are formally written down and iterated upon asynchronously.
> Every single meeting would be everyone hallucinating what was said before in previous meetings, and lots of lying.
Yup. Even with notetaking, that's all you can get out of meeting. You simply don't have immediate access to all the information you need (critical in complexity) in the heat of the moment to present useful thoughts, so anything that does come up is bound to be invalid anyway. Noting the invalid expression made during the meeting doesn't serve any purpose. As such, your notes are only as good as the trash that it is. It's a nice break from work, but that's about it.
I don't think every meeting needs notes, but there are three levels that I've found useful (as a dev). I don't take all of these at every meeting, but I make at least one of them at most.
- An email to the group summarizing my action items, which I also add to my own to-do list. This way, if someone thinks I'm supposed to be doing something, they'll usually reply to the email to let me know. This also doesn't need to be recorded during the meeting; it can happen shortly thereafter (but it's often more efficient to write it during).
- A list of useful bits of information that came up. "X is a thing; I should read more about it", "This person is working on Y; that's who to go to with questions", that sort of thing.
- An outline of substantial discussions and key decisions.
I don't need to write down every word of a twenty minute discussion of what color to make the button, but it's handy to have a reminder that "There was an extensive discussion of the color of the button. The agreement was that it should be red." If I'm asked to take notes for a group, this is what I generally produce; I've never had any complaints.
FWIW, it's been shown that just the act of taking notes enhances the memory of the person doing it, whether they refer back to the notes at a later time or not.
As to the value of note taking in general, for devs, I equate it to telling someone to "put that in the ticket, don't verbally instruct what needs to be done". Documenting decisions made, who will do what, etc. can all be valuable and prevent needless discussions later about who's memory is correct about X (i.e. how something is supposed to function). Some people have great memories, but others do not, and even if you think yours is not in need of this kind of assistance, having something to reference back to can be a big help, especially when you can say something like "hmm, that's interesting, in my notes, I wrote down we decided to go with Y, but you are saying Z, when did that change?"
I should note that taking notes is a skill like any other, done poorly it can probably have the opposite effect of causing distraction and missing things that are being discussed.
My luddite? solution: make note taking as painless as there is, make it one word command away in the terminal, and that command organize the notes/diary in dates:
It's a pip package and by installing you can invoke the `diary` command to directly edit the day's note or diary (as long as the python package bin are in your shell PATH). You can also configure location where they are stored or the shell editor you want to use.
There are dozens of these tools, probably 5+ that are popular enough to be regarded as "stable" (give or take acquisitions). Products in saturated areas should either show price prominently to save time on comparisons, or explain well why they're unique (not fluff "we cool kidz", "our product has _good_ quality recognition", etc.).
These types of features will make their way into every meeting product and be frictionless to the end-user where all they do is show up. Those in the middle will be squeezed out quickly.
While an interesting concept, an AI has no point of view and can only summarize with the context given. It won't be able to assume the right person for the action item if not said, won't know who has what expertise if those aren't there, and is limited to words spoken. I'm sure in the future big tech companies will run their own AI models (if not already) to supplement this knowing. Definitely not worth what the offering is today at $37/mo for a few bullet points that takes seconds to write.
The main value of any note taking, summarization, and creative act in general is the process. Not the output.
In addition to this, I found with Otter and Gong that it doesn’t understand internal jargon very well. When you say “the fizzbuzz microservice is blocked by Istio” the note-taking is appalling - yet those are key to understanding the point being made.
because of this issue, when you upload a video to a service I run, it populates and biases the speech-to-text (ASR... using whisper 'prompt' parameter) towards terms in the title and descriptions of the video (from youtube) - I have found this works incredibly well, and am pretty shocked the default youtube transcriptions don't seem to do the same.
One big cohesive company AI could extract topics from documents and appropriately cue the other AIs in the system with keywords to ensure that internal jargon is understood.
No problem. As an interface designer I can assure you that nobody wants to have to track content horizontally and vertically while trying to parse it. Unless there is something you're trying to convey with a carousel that couldn't be conveyed with a list, don't use one. Simplicity is a virtue in content layout.
It definitely won’t make telling jokes as risky as the jester taunting the king
Imagine, at the rate of wokeness, any joke or non factual piece of information you present is now searchable via text or even related terms.
Now imagine your boss or HR person doesn’t really like you and needs an excuse to get you in trouble. They’re few queries away from finding something that from some angle can make you look like a monster.
We briefly utilized a product similar to this, however the issues you mentioned remained and it was awkward to explain that meeting notes were going to some rando saas. The situation became even more embarrassing when it started joining arbitrary private meetings that were scheduled. We immediately got rid of it and won't again use another product like this.
If you believe that hand-written notes improve your memory and comprehension, as I do, then the real value of this kind of tool is simply to make conversations searchable, and easily shareable.
Semi-ADHD person here. Personally I feel like I lose engagement unless I'm actively taking notes on a meeting.
Definitely doesn't have to be pen-and-paper but organized notes help me focus and keeps continuity of context, whereas my natural tendency is not to record things into my brain very well.
A few years ago I designed a product like this, but more focused on sales calls. We called it DiscoveryCall.ai and and my friend built a proto-type. It was cool, but the sales cycle was something I didn't want to tackle.
A few ideas & learnings if they can help.
1. If you focus on a specific domain of conversation, you can pull conversation data and "summary" into a more valuable data structure. For sales it might be identifying the Decision Maker, budget or launch date. Project meetings will have budgets, timelines, approvers etc as well.
2. My thesis was that getting data into a structured format was more interesting than summary. Sales people need to enter their data into Salesforce. Sales people hate doing this and do it poorly. It's valuable if an application can process the recorded conversation into a completed SFDC record/object.
3. It might be possible to rely on the user to prompt the application correctly in the conversation. In a project management scenario, the user might want to collect launch_date as a field. The user would ask, "When is the launch date?"
The response might be, "I don't know, we're thinking about March or maybe April or even May."
The user of the application can then repeat. "OK, let's pencil in the launch_date for April 15th."
During onboarding or setup for this type of conversation and data extraction, the user would set these keywords like "Pencil In" to then save the date mentioned in the conversation.
Anyways, that's what I remember. If any of that is interested or you want to see any old notes and wireframes, hit my email.
59 comments
[ 3.4 ms ] story [ 122 ms ] threadI would like a tool like this that made a summary/notes and included the full transcript, so they could be searched. This would be good for people who attended and those who didn't, or for seeing if something was discussed a while ago.
Ideally the same system would work on zoom calls, in person meetings, and in meetings with both (some in person, some on zoom). For me the hybrid workplace is one that has the advantages of both, not one that is either/or.
Debatable, what if...
- there are not enough doctors to see patients
- it meant I could be seen immediately vs having to wait 1+ hours
- my doctor had instant access to all medical knowledge and the wisdom of 1M+ hours of practice
- there is a pandemic where a robot is much easier to sterilize between patient visits
There are people already willing on a personal / social level to see an artificial doctor, doctors are already incorporating AI into their practice to help reduce errors, it's hard to believe it will never happen. The context of the visit or stay will play into acceptance
An actual argument might be that it can't fully replace my own notes because my own notes can, and sometimes do, include notes about my thoughts and opinions that the AI can't possibly know without being told. But that does not negate the possible value in terms of freeing me from taking notes about what was actually said during the meeting.
With respect to being taken care of: I want to be taken care of in whichever way produces the best outcome for me. If that means care handled by an AI, then I want the AI.
[1] https://en.wikipedia.org/wiki/Appeal_to_nature
This is cool, but sometimes you want only a summary, sometimes you want a clear list of decisions and actions (with owners), and sometimes you literally want a word-for-word transcription of who said what.
I've seen wildly varying company cultures regarding their approach to note taking. I worked for a company once where the SVP and above folks were treated as kind of god-kings, where every word they spoke had to be written down and agonized over. Felt like those guys following Kim Jong-un around with little notebooks. Even something casually mentioned would become a real project with dozens of staff, just because it was uttered by a SVP during a meeting once. Other places I've worked barely took any notes, and your team's priorities and expected work was passed around by word of mouth like a terrible game of telephone. Everywhere is different!
But sometimes most of the value is in the act of creating it.
I just type into org-mode. The notes are rarely useful later, but I keep them in a folder that I can grep through.
But in highly technically important meetings, y'all might want to proofread in case the wrong TLA gets used.
I mean, to be clear, these will be strictly less valuable than human-generated notes.
But on the other hand, I think they'll be strictly more valuable than no notes.
And as someone who has a terrible time trying to both note-take and participate in a meeting at the same time, this seems like it could be super helpful.
I don't feel like I'm missing anything, but equally the op has built a product to automate this, so it's got to be important in some contexts. Would like to know what I'm missing
Taking notes doesn't have to be painful, for me it's just one `diary` command away and it will open/create the appropriate note file in my favorite editor:
I've also packaged it as a pip package for distribution: https://github.com/Aperocky/diarycli
Most tools are great at the minutes, but cannot have the context to do 'notes'
1. Decisions and supporting details
2. Action items with dates and assignees
Those should be sent out to all attendees of the meeting (or written in a place that's accessible to everyone). Beyond that, each person needs to take notes for themselves for other things they want to remember.
Every single meeting would be everyone hallucinating what was said before in previous meetings, and lots of lying.
I don't imagine it is. I've never seen anything reasonably complex successfully built by way of meeting, notes or not. The actual work always happens when ideas are formally written down and iterated upon asynchronously.
> Every single meeting would be everyone hallucinating what was said before in previous meetings, and lots of lying.
Yup. Even with notetaking, that's all you can get out of meeting. You simply don't have immediate access to all the information you need (critical in complexity) in the heat of the moment to present useful thoughts, so anything that does come up is bound to be invalid anyway. Noting the invalid expression made during the meeting doesn't serve any purpose. As such, your notes are only as good as the trash that it is. It's a nice break from work, but that's about it.
Most formal meetings start with approval of the minutes of the previous meeting.
- An email to the group summarizing my action items, which I also add to my own to-do list. This way, if someone thinks I'm supposed to be doing something, they'll usually reply to the email to let me know. This also doesn't need to be recorded during the meeting; it can happen shortly thereafter (but it's often more efficient to write it during).
- A list of useful bits of information that came up. "X is a thing; I should read more about it", "This person is working on Y; that's who to go to with questions", that sort of thing.
- An outline of substantial discussions and key decisions. I don't need to write down every word of a twenty minute discussion of what color to make the button, but it's handy to have a reminder that "There was an extensive discussion of the color of the button. The agreement was that it should be red." If I'm asked to take notes for a group, this is what I generally produce; I've never had any complaints.
As to the value of note taking in general, for devs, I equate it to telling someone to "put that in the ticket, don't verbally instruct what needs to be done". Documenting decisions made, who will do what, etc. can all be valuable and prevent needless discussions later about who's memory is correct about X (i.e. how something is supposed to function). Some people have great memories, but others do not, and even if you think yours is not in need of this kind of assistance, having something to reference back to can be a big help, especially when you can say something like "hmm, that's interesting, in my notes, I wrote down we decided to go with Y, but you are saying Z, when did that change?"
I should note that taking notes is a skill like any other, done poorly it can probably have the opposite effect of causing distraction and missing things that are being discussed.
https://github.com/Aperocky/diarycli
It's a pip package and by installing you can invoke the `diary` command to directly edit the day's note or diary (as long as the python package bin are in your shell PATH). You can also configure location where they are stored or the shell editor you want to use.
Digital, if you want to go a bit modern; with speech-to-text postprocessing if you want to go modern.
As far as selecting salient points from the transcript goes, the ancient tech of "Copy" and "Paste" usually does the job just fine.
While an interesting concept, an AI has no point of view and can only summarize with the context given. It won't be able to assume the right person for the action item if not said, won't know who has what expertise if those aren't there, and is limited to words spoken. I'm sure in the future big tech companies will run their own AI models (if not already) to supplement this knowing. Definitely not worth what the offering is today at $37/mo for a few bullet points that takes seconds to write.
The main value of any note taking, summarization, and creative act in general is the process. Not the output.
One big cohesive company AI could extract topics from documents and appropriately cue the other AIs in the system with keywords to ensure that internal jargon is understood.
Imagine, at the rate of wokeness, any joke or non factual piece of information you present is now searchable via text or even related terms.
Now imagine your boss or HR person doesn’t really like you and needs an excuse to get you in trouble. They’re few queries away from finding something that from some angle can make you look like a monster.
Better just to say as little as possible.
This has been the norm for the past decade.
Please stop joining important things to giant mystery machines, it's extremely annoying to watch.
https://journals.sagepub.com/doi/abs/10.1177/095679761452458...
Definitely doesn't have to be pen-and-paper but organized notes help me focus and keeps continuity of context, whereas my natural tendency is not to record things into my brain very well.
A few ideas & learnings if they can help.
1. If you focus on a specific domain of conversation, you can pull conversation data and "summary" into a more valuable data structure. For sales it might be identifying the Decision Maker, budget or launch date. Project meetings will have budgets, timelines, approvers etc as well.
2. My thesis was that getting data into a structured format was more interesting than summary. Sales people need to enter their data into Salesforce. Sales people hate doing this and do it poorly. It's valuable if an application can process the recorded conversation into a completed SFDC record/object.
3. It might be possible to rely on the user to prompt the application correctly in the conversation. In a project management scenario, the user might want to collect launch_date as a field. The user would ask, "When is the launch date?"
The response might be, "I don't know, we're thinking about March or maybe April or even May."
The user of the application can then repeat. "OK, let's pencil in the launch_date for April 15th."
During onboarding or setup for this type of conversation and data extraction, the user would set these keywords like "Pencil In" to then save the date mentioned in the conversation.
Anyways, that's what I remember. If any of that is interested or you want to see any old notes and wireframes, hit my email.