#12: AI-Generated Zoom Summaries: Yay or Nay?
Welcome to another episode of The Junction. I'm excited.
Today we are talking about AI generated zoom
summaries, all the things you need to know.
We use zoom.
We've used zoom for a very long time.
We've zoomed for a long time.
We've been active zoomers,
zoom adopters, early adopters.
Not that we're not a virtual workspace by any
stretch, but we work with a ton of clients.
We're in consulting, and so our clients
and partners are all over the world.
But what do you think about all this stuff that's
coming out, about them updating their terms of service?
And they're now using and learning off customer data
that just happened within the last couple of months.
It's not shocking.
It's not shocking at all.
Everybody wants data.
If you've done any research on any
of these large language models, everything relies
on some kind of human created data.
I think we talked about this a little
bit in one of our last episodes.
If you don't have data to train the model on,
then you don't have anything to train with, right?
So then it's like, well, let's start generating.
How do we generate more data?
Well, what if I had a ton of recordings where people
were talking every day, all day long, and if I had
access to that, then I could train all sorts of models.
So this is a really interesting move by Zoom.
It doesn't shock me in the least bit.
And they've paired it with this idea of a summary.
Like, we're going to summarize your call.
And that's something that we want to do.
We've been doing we're doing it with our transcripts,
playing around with it, from recruiting to sales calls
to prospecting, I mean, you name it.
Gong has been doing a little
bit of this, like watching Keywords.
I want to sidebar about that.
How has Gong been doing it for so long?
If you do need such massive amounts of the they're
looking at the transcripts and then finding like, keywords.
I think you and I took one of their calls, right?
They're like, yeah, how many times
did Mel say marketing, right?
And they're like, 13.
Well, then they can link that word or the number of
times that you say it to each spot in that video.
What Zoom is doing now is they're taking
the whole transcript and they're giving everybody a
summary of what the call was about.
How do you think the summary is?
We've experimented with it since it was available.
It's not it needs to learn.
They need a lot of data.
Now, is it learning based off of all of this,
anonymized customer data, or is it learning on our account?
You have to enable zoom summaries, but
the summary capability is in mass, right?
They're not looking at so for Venn
technology, it's not training on Ven.
And here's Ven's summary training in mass, right?
I believe they're anonymizing the recording data and
they're training their model based off of that,
and everybody's going to get access to an
enhanced version of Zoom Summaries, right? Right.
Now, did you ever do Dragon easy Speak?
Dragon Speakeasy?
It was like 2004.
You speak into a mic and it basically records your
voice and then types it for you, like word processing.
Okay, if only the listeners could
see my face right now. They can, though.
This is also why we make we've got synergy, right? No.
Yeah.
I have not done those things.
I'm willing to bet that some of
our listeners out there are not familiar.
But what is the connection there?
Where are you going with that?
Well, the most challenging thing was taking
audio and converting it into text. Okay.
Right now, Zoom has gotten fairly good at that.
I mean, you still catch the one off words where it's
like, I didn't say the F word, but there it is.
Sometimes it misses it, and that's probably more so
some kind of internet connection, mic related thing.
There's a hiccup. Whatnot?
But now they need to take the transcript and they
need to dump it into their large language model.
And the only way they get access to that
is if at least what they say right, is
if you enable these Zoom call Summaries, you are
now automatically saying that it's okay for Zoom to
look at all your recordings, take the transcripts right.
And then put that into their large language model
to train it on all the things you're saying.
It's actually a really interesting concept because once they've
trained it on all this data, all these natural
conversations that people are having, it'll be really interesting
to see what comes of that.
Because the majority of the large language models are text
based conversations where I'm chatting with you or it's reading
reddit posts and we go back and forth and we're
arguing about, I don't know, chess, who knows?
Whatever people argue about.
Very contentious landscape.
Yes, you better watch out.
But Zoom recordings, a lot of them
are like business related business conversations.
It's usually b to b, I would say, for the most part.
Right.
And so they're going to have this kind of
like, general business person that's really the large language
model that is is going to be really good
at having a one on one conversation with somebody
the same way that we're talking right now.
Anyway.
But I imagine that when we see what comes of
this, people will start to question how'd you do this?
How much of our data did you take?
They're not necessarily very specific other than they
say that you're giving them access to it
so that they can enhance the product. Right.
Well, that doesn't mean that they won't also come
out with another product that they sell sure.
Based on all the data that you gave them.
What I think is interesting, we've talked about
deploying something like this internally from pulling the
transcripts in and based on the type of
call or the stage that we're in.
So if it's an early stage sales call, having the
recording or the summary that comes out of the zoom
transcript, and if we're using a tool like OpenAI to
do it, but basically saying for these types of calls,
we want you to not only summarize it, but then
we want you to answer these five questions what was
the thing we're selling?
What was the sentiment of the person on the phone?
Whatever typical kind of things we would want to see
recapped that the sales rep does not have time to
recap on and then being able to package that up
and drop it into Slack Channel, something like that.
I think that's something that's really interesting to
me, that we would be able to do,
like by stage or by call type.
Oh, for sure.
Well, there's two avenues here, right?
Like, if you give your recordings to Zoom, who
knows what they're going to do with it.
Oh, by the way, I want to spend half a
million dollars buying this software, so on and so forth.
Right?
And now I'm talking about intent. Data.
Well, Zoom Info, different company is very interested
in that right now, maybe they anonymize It,
but somebody is interested in this particular piece
of software to the tune of 500 grand.
Got to be really careful about maybe what
you say if you enable this, because you
may be unknowingly giving other companies interest in
what you were talking about on the phone.
But back to what you were thinking.
If you can have that full recording and you
can summarize all the things that you're saying, the
transcript or the summary ideas, it's interesting.
It's good for the people that didn't weren't
in on the call, or maybe you had
a long weekend and you can't remember either.
That's great.
What we found that might even be more powerful is
people tend to at least in the sales cycle, right?
Like, oh, yeah, they were totally in.
They gave me a verbal half a mil, we're ready to roll.
And if you go back and listen to the call, they're
like, yeah, we're talking to a couple more other folks, and
we're going to make a decision in the next six months.
And it's like, whoa, whoa, that was totally different.
Right?
So let's gauge the sentiment on the
phone call for this specific speaker.
When the decision maker talked, did they sound excited?
Did they sound ready to buy, or
did they mention anything about waiting?
Are they ready to buy? Right now?
You can ask questions of the transcript.
That isn't necessarily a summary, and
that's where I found it.
So you're interacting with it. Oh, yeah, okay.
So that's different.
That differentiates a product like Gong from what we're
talking about versus because it can do that today.
It can give you the sentiment.
And this was mentioned this many times, and it can
give you kind of that application of training hey, sales
manager, help train up this rep on these things.
Oh, yeah.
But if you have the ability to then
actually have that conversation like we do in
Chat GPT after the call, that's pretty powerful. Right?
Well, it's great for people that weren't on the
call that want to ask questions about the call,
but they can't get a hold of that person. Right.
Because maybe they're on PTO, but they also don't have if
it was like two hour long phone call, maybe they don't
have 2 hours to go back and find that one thing.
Right, right.
So now you can ask the large language model,
did anybody mention anything about a ballpark range?
And this is one that I'm actually very interested in, because if
we've already told them it's 20 to 50 grand and I'm about
to get on the phone and I'm going to tell them it's
50 to 100, they're going to be like, hold up.
Brent told me that it was 20 to 50, and
I'm like, oh, shoot, I should have watched the video
or read through the transcript or asked a question.
But now Brent's out of the Bahamas, and I'm
at my own left to my own devices.
So it just really speeds up that knowledge
transfer when the knowledge transfer doesn't happen.
Well, I guess we'll see how this continues to evolve.
I know, like you said, we're experimenting with some
of the stuff in house on our end.
And as with anything, we weigh kind of the
cost to go build and do ourselves versus some
of the capabilities that are out there.
So again, kind of like important
to just keep these conversations going. Right.
One of the advantages of having an AI committee, which I
think let's just do an episode on that one time.
Totally.
How to kind of put together your AI
committee, what they should do, what's the cadence.
But I think that we're just getting that off the ground.
But that's been helpful to kind
of keep these discussions alive.
And it's not just you or myself finding staying
up to date with some of these trends, but
we've got multiple people in different parts of the
organization that are keeping tabs on AI.
And if we need to pivot direction because you know
what might not be worth it to build that, let's
go and adopt this feature in The Tool. Absolutely.
You know what's crazy is I think meetings
maybe I'm maybe slightly biased, but I think
meetings go really well in person.
And I don't know if you saw this headline, but
Zoom just announced to people that are within like 50
miles of their offices that they must come in.
And it just made me chuckle a little bit because Zoom
is all about, well, you all can meet remotely, but if
you happen to work for us in this one little area,
you need to come into the office to maybe use Zoom. Wow.
No, I missed that.
Just slightly ironic.
No, it is ironic.
I wonder if they're sitting in the
conference rooms with each other using Zoom,
because that would make me chuckle.
I mean, there has to be some I don't know.
Again, you look at, like, the we know what the
value add is for in person in office collaboration.
Again, we use Zoom all the time to interact
with clients and partners, but there is something about
having those quote unquote water cooler talks.
Oh, yeah, we've got a dartboard in here.
That's kind of where some
of those conversations happen.
You can hear it on the other side of the studio.
Or Nintendo Switch.
Nintendo Switch.
We've got break room.
Yeah, I don't think anyone
touches the foodswell machine. We don't.
We almost switched it out for the shuffleboard.
The shuffle board would have been way cooler.
I love shuffleboard.
I almost told Brent to do it, but it's fine.
We'll get know when we're on the up and up.
Let's have a Zoom call about it. Nice.
All right, so let's move on.
I love the headline section, so thanks for
kind of teeing us up for that.
I did miss that headline.
This one is really interesting.
I have to ask before I read the headline
chase, did you ever play Dungeons and Dragons?
Now, if I tell the truth, are you
going to ring me out on social media?
Actually, I should have paraphrased it and
not locked you down to the past.
Do you presently play it's?
Like, maybe you still play?
I really liked the idea of Dungeons
and Dragons telling stories, building out characters.
I just didn't have friends that really bought into that
whole thing, and so I have never played an in
person tabletop version of it, but I have played lots
of video games, role playing, RPG games that are based
on the Dungeon and Dragons rule set.
Okay, so do you know where I'm going with this, then?
Have you seen anything about this?
I actually saw this headline today before
you threw it on, so okay.
Yeah, I mean, it was happening on news.
There was, like, six pages of Google. Yeah.
So Dungeons and Dragons tells Illustrators to stop
using AI to generate artwork for fantasy franchise.
What's your take?
This really doesn't shock me a whole lot because
I saw another headline maybe somewhere that some court
in the United States has ruled that you can't
copyright something that wasn't created by a human.
And I didn't really dig into the really, though, I didn't
make it down that far, at least in my head.
But I like the general idea.
We want to say that original works
that were generated by somebody's own hand
constitute something that can be copyrighted.
And so if I'm paying you to use your own hand
to create art for me, I'm not saying, hey, I want
you to go create art and use AI that now makes
it something that you didn't do with your own hand, because
that's not what I'm paying you to do.
So I can understand and enjoy
maybe the idea behind this. Right.
Like, well, we want to pay for original
works of art, not something that generates it
just because I typed in a sentence.
Yeah, but there's also to something like when you
think about art that's hanging on the wall, original
art sitting in museums right, from a long time
ago, was all done by hand.
And that's because it's got that human element.
It was handcrafted. Right.
Maybe that's the word I'm really looking for.
Things that I really enjoy
or really value are handcrafted.
Things that I throw away real fast or
things the beer you're drinking right now, like
the beer, I mean, the can. Right.
I think of the timu stuff that I buy for $2.
Well, it's worth $2 because it's not handcrafted. Sure.
And I feel like AI generated creative works
of art are kind of the same thing.
Like, well, I spent 2 seconds typing up
some words, and I utilize that to submit
my artwork for a Dungeons and Dragons card
set or book or whatever thing they're generating.
So I think I'm in favor of this.
What about you, though?
What's your take here?
Because I feel like you maybe have
a hotter take than I do.
No, I would say we're in line.
I think it's important to protect the artists
like you're talking about, especially really original artwork.
I was actually just out of town for the weekend,
visiting some friends and family, had a baby shower to
attend, and we went in this little local community.
There was a bunch of original artworks
out in San Juan Capistrano in California.
And, I mean, I'm just, like flipping through all
of this artwork, know, marveling at the price of
some and gawking at the price of others.
But there's also such a huge appreciation as I hold it
and I can see the brushstrokes and know someone somewhere in
this town or across the world the real deal yeah.
Was holding this.
They were looking at something as they were
crafting it, or they had an image in
their mind, and that's really powerful, and I
think we absolutely need to protect that.
Where I think it starts to get a little
slippery is we use tools like Illustrator and InDesign.
Canva we use a number of different tools here at Vend?
Do our own graphic design?
Granted, we're not doing the level of graphics and books
and things like Dungeons and Dragons, but you create one
thing one time, you can do a copy paste and
apply it elsewhere, and you don't necessarily I don't believe
that, to be like plagiarizing your own work.
So there is that level of being able to
put an original concept together and then maybe leveraging
some of these tools to get you that type
of style or artistic flair across multiple mediums.
So I don't want to say that you shouldn't ever do it.
There's very specific use cases for it.
And obviously this particular illustrator was doing
some experimentation and maybe thought this was
going to fly out on the radar
or wasn't thinking about the implications.
Yeah, well, part of some of the lawsuits that have
come out is that the lawsuit is against the model
effectively training off of the content that was handcrafted, that
was paid for by those individuals or organizations.
Right.
And this is maybe at least as far as the
lawsuit goes, this is where I'm kind of conflicted because
the large language model is simply like looking at what
you did and then making its own, altering it.
Well, that's what we I mean, we all do that, right?
So what's the difference of this machine
checking out all your work, right.
And then creating a version of that?
Now it's its own version.
And I think that's kind of where I struggle.
And I personally hope they lose those because
I really don't see a true difference.
If I did that and I happened to
do it very fast, just was that good? Right.
They're not going to be able to sue me
simply because I can paint faster than somebody else.
But I think it really does
go back to this handcrafted idea.
Was the work originally handcrafted?
If so, that's the copyright.
If it wasn't generated by hand,
it was generated by something else.
Well, that's not copyrightable.
You also have to consider, where was this individual?
So were they employed by a company that thereby
all of those rights are actually belonged to the
company I've had in previous roles, did actual created
illustrations that went on to become corporate training material
for years after I had left the company.
And I will occasionally get a text from someone
former colleague like, hey, does this look familiar?
They're still using it.
You're famous, right? Somewhere.
So I see that and in one regard, I'm
like, oh my gosh, they're still using it.
That's so cool.
And then there's that little joke
narrative, like, where's my residual?
Right?
Once upon a time, I signed a contract
that I was working for said company.
And anything that I do, same goes here for Venn.
Anything I do here, it's not mine, really. Right.
It's under the umbrella of Venn technology.
So there's also like that consideration. Right.
And I think that I can understand both sides.
That's where it's kind of tricky.
Like, I don't want to weigh in all in on one side.
I'll let the courts handle it.
I guess that's what they do.
That's what they're supposed to do. Right?
There's like lawyers and judges.
I sit at my keyboard all day. Yeah.
What do I do?
Okay, well, let's jump into the next headline then.
So this one's super interesting.
Speaking of travel and going on trips,
a new frontier for travel scammers.
AI generated guidebooks.
There's this whole article by the New
York Times, like debunking this Amazon Travel
Book scam essentially, or AI generated content.
These travel scammers are leveraging a number
of tools to put together these guidebooks
that have been ranking in the top.
Like, you type in Paris 2023 Guidebook, right?
And these AI generated content pieces are showing up.
They're buying them and then they're getting
them and saying that this is trash.
But it's very misleading because they're showing as
being almost like ranked or promoted on Amazon.
There's a lot of five star listings which
the New York Times kind of dug into.
So yeah, I'll stop there.
What's your take?
Okay, so this isn't the ultimate direction that I'm
going to go, but I love the I don't
know, just the creative ingenuity that you have.
So if you read the full article, right,
they had to come up with content.
And I'm not saying that you should
go and do this yourself because ultimately
they're going to get found out, right?
And they're going to get kicked off and
then their account's going to be banned. Who knows?
But I have to appreciate the ingenuity here.
They created a bunch of content using, right?
They created a bunch of images using AI, okay?
People have been doing that now past six
months or however long if you're an OG.
Then they created some kind of
image for the front page.
They've got some images in the book. I imagine.
They've got this fake author.
They've got this fake bio.
But you know what really kind of just
makes me smile is that they were able
to put all of these things together, right?
The content, the images, the author, the picture.
Hey, let me list it on Amazon.
Oh, by the way, Amazon has like a paperback generator deal,
so I don't really have to print my own books.
Or they can get it for free on Kindle Unlimited.
Then you slap in some fake reviews
and you put all those together, right?
And you net, I don't know.
There's no way this person made
more than a couple of $100.
But that's just for one book.
What if you did that times 100 books, right?
Well, if you made a couple hundred bucks
times 100, now we're talking about $100,000.
Like, well, man, that's actually
a decent chunk of change.
I think these folks, whoever they were, just
ultimately put all this together as a quick
way to grab some quick money.
And Amazon just hasn't found a way to quickly
find these out, to quickly get rid of them.
Because ultimately it does look like a bunch of spam.
They didn't have a Microsoft red
AI team or whatever, right?
They have a red team to go in and be like,
how do I scam people out of a bunch of money
when people wants to go to just I honestly the irony
of it and for anyone listening, you should go look this
up and check out the New York Times article because also
just I thought where you were going to go with it
was just the way that the New York Times article looks.
They confront you with all these images of these fake
travel guides, and as you scroll through, it's very interactive
and asks questions and kind of tease things up.
But the irony of all of it is this person that
essentially was Scammed, she was trying to spend, like, $10 less
on a guide, and she ended up going back to the
OG, Rick Steves, who's apparently the master original content creator for
travel guides and his take on all of this.
You got to be boots on the ground
to know exactly what you're doing out here.
I'm not worried.
And this just continues to go back to what
you've said from the very beginning about original content
creators being the ultimate form of authenticity.
That's how you cut through the
noise of AI generated content. Oh, absolutely.
Absolutely.
And I don't know if you actually went, I'm
on Amazon right now, and I typed in France
travel Guide, and there's a ton like, holy smokes.
Well, stay away, Rick Steves.
That's your guy I was looking for.
The one by Mike Steves is
the guy in this article, right?
And I bet it's already been pulled down because
it's New York Times, but I didn't know travel
books were such, like, a huge kind of they
do touch on that mean, I I use Instagram.
Like, you can go to Pinterest and you've
got kind of these travel bloggers and stuff.
So I think people have moved away from some
of the more handheld paperback versions of things.
But as I say that I consume paperback
books just maybe not in travel form.
Well, that's kind of the thing that this blog
pulls out or not this blog, this article, right?
If you look, if you scroll down,
maybe you'll pull up this article.
If you scroll down, there's an image of an
image where this lady had written a review.
And if you zoom in enough, you can see it.
It's an image of the Notre Dame Cathedral.
And then right below that, there's, like what is that
one, two we'll call it four sentences on one page.
Like the font is ginormous, right?
This is like a little kids book, and
you spent $17, right, and then looked at
I mean, if I could read this, it's
probably like, just basic information about Notre Dame.
It's like I could pull this up
on Wikipedia or with a Google search.
And that's what the reviews said.
The One Star reviews were like, I might as
well have just done a Google search copy pasted
into Word Doc and printed it out.
Well, if there's one thing about this handcrafted original content
thing that, you know, going off on, it's that if
you want to be unique in the space that you're
talking about or that you're in, you're now going to
have to be even more unique.
Because the moment that your content is ingested
by the AI let's circle all the way
back around to the beginning of the show.
I've now come up with some really unique content,
and now all of these large language models have
digested it, and then they start spewing that stuff.
Well, now my content is maybe original, but it's no
longer unique because everybody else is saying it too.
So Rick Steves needs to be worried.
He needs to be worried if
he stops generating unique content.
And, I mean, how many people are really out
there traveling the world, coming up with I don't
know, but it sounded like a pretty great lifestyle.
He actually has a whole team of fact
checkers that their job is to travel.
We should apply using his guide, and they go to
these places, and they fact check that they're still there.
They eat at restaurants.
Let's do that.
And then we can do, like, a co sponsor thing
where he's on our show and we're on his show.
If he has a show. Okay.
Or he should start one, we can help him.
I'll add that to my list.
Okay.
Well, this has been fun.
I really enjoyed this one.
I know we kind of bounced around with this Dungeon
and dragons and travel themes, started out with Zoom recordings,
but there's just so much interesting stuff out there in
the news right now that we can't not cover some
of these kind of quirkier things that are a little
bit more outside of the lines.
And it was not really b to b oriented
headlines today, but totally quick sidebar, I wasn't going
to say it, but I have to.
I did kind of get into Dungeon
Dragons for, like, a hot second.
I bought Gloomhaven, which is based off of Dungeons and
Dragons rule set, but it is very much like Tabletop.
You got some characters, and it's this Gin box.
If you've ever played and you want to play, hit
me up, because when I tried to play, the only
person that would play with me and she didn't really
even want to play was my wife.
And once we got it all out, she
was like, yeah, I can't do this.
So if you're out there and you want to
play glue maven one day, just let me know.
But yeah, maybe one day I'll be
really cool and get to play.
That's one of the most authentic things I've
ever heard you say one day that I'm
not cool yet or that I'm already above
you're cutting through that AI noise big time. Oh, man.
All right. Well, thanks, Chase.
This has been really we. Until next time.
You know the drill.
Send us an email junction@ventechnology.com,
give us your take.
Laugh with us.
Send us other articles, or cry
with us, or cry that never.
Happens off air.
We would love to know what your questions are.
What other quirky articles you have or
new and emerging AI and automation trends.
We are here for all of it.
Until next time, keep it automated.
