Sara:
00:00:01
Welcome to the Thinkydoers podcast.
Sara:
00:00:04
Thinkydoers are those of us drawn to deep work, where thinking is working.
Sara:
00:00:09
But we don't stop there.
Sara:
00:00:11
We're compelled to move the work from insight to idea, through the messy
Sara:
00:00:16
middle, to find courage and confidence to put our thoughts into action.
Sara:
00:00:21
I'm your host, Sara Lobkovich.
Sara:
00:00:23
I'm a strategy coach, a huge goal setting and attainment nerd, and board-certified
Sara:
00:00:30
health and wellness coach, working at the overlap of work, life well-being.
Sara:
00:00:35
I'm also a Thinkydoer.
Sara:
00:00:38
I'm here to help others find more satisfaction, less frustration, less
Sara:
00:00:43
friction, and more flow in our work.
Sara:
00:00:46
My mission is to help changemakers like you transform our workplaces and world.
Sara:
00:00:52
So let's get started.
Sara:
00:00:57
Welcome back to part two of my conversation with Dvorah Graeser.
Sara:
00:01:01
In our last episode, we explored some of the ethics and
Sara:
00:01:04
implications of AI at work today.
Sara:
00:01:07
Now, we're getting practical.
Sara:
00:01:10
We'll discuss some specific and maybe surprising ways to leverage AI tools,
Sara:
00:01:15
particularly for deep specialists and those of us wired strategically.
Sara:
00:01:21
If you're ready to move from anxiety to agency with AI while
Sara:
00:01:25
maintaining your ethics and humanity, this episode's for you.
Sara:
00:01:33
Dvorah, you mentioned kind of the shift from anxiety about AI to agency and
Sara:
00:01:39
how people can leverage AI around their creative vision, regardless of what
Sara:
00:01:45
their technical level of proficiency is.
Sara:
00:01:48
Tell me a bit more about how you see the use of AI enabling
Sara:
00:01:53
creative and multi-talented humans.
Dvorah:
00:01:56
Well, one way that I personally like to use AI—I don't know if
Dvorah:
00:02:00
you've gotten to this point when creating, written content, but I'll
Dvorah:
00:02:03
have a whole pile of thoughts, and they are all over the darn place.
Dvorah:
00:02:06
If they were sticky notes, they would be like Mount Everest of sticky notes.
Dvorah:
00:02:09
And I don't know what the heck to do with them, and I get very frustrated by this.
Dvorah:
00:02:13
One of the things that I have found that can be great to do is I shove the
Dvorah:
00:02:17
whole thing into, let's say, Claude.
Dvorah:
00:02:19
Now, Claude has a feature where you can copy and paste, but you can also upload.
Dvorah:
00:02:25
So, let's say you have a PDF and you have some written documents
Dvorah:
00:02:28
and some other stuff, and you just dump the whole thing into Claude.
Dvorah:
00:02:31
And I just say Claude exactly what I'm feeling — I don't know how to organize
Dvorah:
00:02:35
this, like, if I can't make sense of this, or I have one thought, I will write the
Dvorah:
00:02:40
thought, but the rest of it's not clear.
Dvorah:
00:02:41
What do I do?
Dvorah:
00:02:42
And then Claude very calmly chugs along and proposes something.
Dvorah:
00:02:47
They propose an outline.
Dvorah:
00:02:49
It can be helpful if you ask it for something specific, like an
Dvorah:
00:02:51
outline or whatever, but even if you don't, it'll still do its best.
Dvorah:
00:02:55
And then you start a dialogue.
Dvorah:
00:02:57
And this is one of the things I really do like about generative AI.
Dvorah:
00:03:00
We have a dialogue, we go back and forth.
Dvorah:
00:03:02
I ask for constructive feedback because otherwise, it'll only say nice things.
Dvorah:
00:03:06
It's something about the generative AI.
Dvorah:
00:03:08
If you ask for constructive feedback, it'll also say the mean things.
Dvorah:
00:03:10
But I'm going back and forth, and as I'm going back and forth, in my
Dvorah:
00:03:14
own mind, I'm realizing, I don't like that bit, I do like this bit.
Dvorah:
00:03:18
Oh, I just thought of something completely different.
Dvorah:
00:03:20
Now, what went from a blizzard of ideas I couldn't organize will go to an
Dvorah:
00:03:25
organization that maybe isn't exactly what I want, but helps clear my mind,
Dvorah:
00:03:29
reduce my own anxiety, because it's often my anxiety that's leading to the
Dvorah:
00:03:33
pile-o'-sticky-notes problem, at least in my experience — and enables me to get to
Dvorah:
00:03:36
the point where I can do the organization.
Dvorah:
00:03:39
Even if it turns out that it's not Claude that has made the
Dvorah:
00:03:42
good proposal by interacting with it, I get to where I need to go.
Sara:
00:03:46
That is the use case that has been the most transformative for me.
Sara:
00:03:51
And that's where I went from being a skeptic who was not likely to integrate
Sara:
00:03:56
these tools to being a hooked user.
Sara:
00:04:00
They don't replace the human brain.
Sara:
00:04:02
You mentioned hallucinations in our first episode.
Sara:
00:04:06
I don't rely on models for facts or information, but they are really good
Sara:
00:04:12
at organizing big messes of information.
Sara:
00:04:17
The thing that frustrates me is sometimes they're poor at actually organizing data.
Sara:
00:04:23
Even when I train it on data and want data analysis or data, you know,
Sara:
00:04:29
stuff, that's been a big frustration for me with the current tools.
Sara:
00:04:33
Even when I train on data, it can still hallucinate, which is a bummer.
Sara:
00:04:38
That's the kind of stuff that I would think the models would be
Sara:
00:04:42
built to do really well, and with as little hallucination as possible.
Sara:
00:04:46
But maybe there's always going to be some hallucination.
Dvorah:
00:04:50
There's a couple of ways around hallucination, right?
Dvorah:
00:04:53
And this again gets into process.
Dvorah:
00:04:54
We talked about it in the first, episode, but it's still important to bring up.
Dvorah:
00:04:58
You need to have a process for dealing with this.
Dvorah:
00:05:00
So let us suppose you uploaded a bunch of data to, I don't
Dvorah:
00:05:03
know, let's say it was ChatGPT.
Dvorah:
00:05:04
It doesn't matter.
Dvorah:
00:05:05
You uploaded it, and it's still hallucinated.
Dvorah:
00:05:07
So what are ways to check this?
Dvorah:
00:05:09
One way to check this is to ask ChatGPT in the same session to go back, and to check
Dvorah:
00:05:16
its answers in comparison to the data and show where it got each answer from.
Dvorah:
00:05:20
This is also great for summarization, by the way.
Dvorah:
00:05:22
If it summarizes and it comes up with something, you say, "Okay,
Dvorah:
00:05:25
where did that really wonderful, great quote, where is it?
Dvorah:
00:05:28
Show me where it is." And they're like, "oh yeah, you're right, I
Dvorah:
00:05:30
kind of made that up." Okay, no problems, let's just move on.
Dvorah:
00:05:33
A third way for doing it is then to actually go to another tool like
Dvorah:
00:05:41
Perplexity and ask it to do research.
Dvorah:
00:05:43
So maybe you don't want to upload your sensitive data to Perplexity,
Dvorah:
00:05:46
but you can still ask it to do research on the area and give you some
Dvorah:
00:05:50
ideas of what the boundaries of the area that can help corral ChatGPT.
Dvorah:
00:05:54
Or you can go to a completely different tool.
Dvorah:
00:05:56
I guess the completely different tool is the third way of doing it.
Dvorah:
00:05:58
Let's say you go from ChatGPT to Claude.
Dvorah:
00:06:00
You can go to Claude, and then you can say, "Hey, can you please check this?
Dvorah:
00:06:03
And what is good and what is bad, and please give me constructive feedback."
Sara:
00:06:07
That's wild.
Sara:
00:06:08
In my early experimentation, when I was working on my book, there was some
Sara:
00:06:12
research I had, like actual research.
Sara:
00:06:15
I needed academic journal research.
Sara:
00:06:17
that was one of the things that was the most frustrating.
Sara:
00:06:20
I would think that models would be good at knowing that truthful data.
Sara:
00:06:26
Hopefully, this has changed since then.
Sara:
00:06:29
But I would ask for research, like, "What are the leading papers or
Sara:
00:06:33
core papers on a specific topic?" And it would make up papers.
Sara:
00:06:38
You've got ProQuest.
Sara:
00:06:39
It's all there.
Sara:
00:06:41
This you should get right.
Dvorah:
00:06:43
So, for example, Perplexity AI, I like because it gives links.
Dvorah:
00:06:46
And you can go check them.
Dvorah:
00:06:47
ChatGPT sometimes does, sometimes doesn't, sometimes hallucinates,
Dvorah:
00:06:50
sometimes it's accurate.
Dvorah:
00:06:51
I'm never quite sure what it's going to do next, so I find that a little unreliable.
Dvorah:
00:06:55
There are some specialized GPTs for papers.
Dvorah:
00:06:58
There's one called Scholar AI, and that's a good one, but
Dvorah:
00:07:01
there are other ones as well.
Dvorah:
00:07:02
So here, I think it's good to join with others to understand what they've
Dvorah:
00:07:07
done and what's worked for them.
Dvorah:
00:07:08
And that can change over time.
Dvorah:
00:07:09
So, for example, I like Scholar AI, but it's possible in a few months,
Dvorah:
00:07:12
someone's going to come back.
Dvorah:
00:07:13
Maybe they'll listen to this again.
Dvorah:
00:07:15
They'll be like, "Oh, but it's no longer good." Possible.
Dvorah:
00:07:17
Things change really quickly.
Dvorah:
00:07:19
So I think it's helpful to try out different specialty tools, to talk with
Dvorah:
00:07:22
other people about this on LinkedIn or another social media that you like.
Dvorah:
00:07:26
Mastodon, if you're into that.
Dvorah:
00:07:28
And try to get information about what is working right now and what is the process
Dvorah:
00:07:32
for working around some of these issues.
Dvorah:
00:07:35
So if I get anything from ChatGPT, because I have found it to be
Dvorah:
00:07:38
the most erratic when it comes to hallucinating, I will copy and paste
Dvorah:
00:07:41
the whole answer to Perplexity AI.
Dvorah:
00:07:43
And I'll say, "Is this right or wrong?
Dvorah:
00:07:45
Please give me feedback.
Dvorah:
00:07:47
Please find me references.
Dvorah:
00:07:48
What do you think?" And it will go through the whole thing and it'll say,
Dvorah:
00:07:51
"Well, this looks like it's correct.
Dvorah:
00:07:52
Couldn't find anything for this.
Dvorah:
00:07:53
I really don't know about that bit over there.
Dvorah:
00:07:55
And this is what I would do instead." And so Perplex AI will actually
Dvorah:
00:07:58
give you a long, reasoned answer.
Dvorah:
00:08:01
This was without going into their deep research mode, which is awesome.
Dvorah:
00:08:04
But that's like a whole nother kettle of fish.
Dvorah:
00:08:05
It's just regular Perplexity AI.
Dvorah:
00:08:07
It'll actually go through these things.
Dvorah:
00:08:09
So I guess it's unlike Google in the sense where if you get a link and you
Dvorah:
00:08:13
click on the link, the link exists.
Dvorah:
00:08:15
If you get a summary with the link, does the link exist?
Dvorah:
00:08:17
Does it not exist?
Dvorah:
00:08:18
Does the summary say what the link says?
Dvorah:
00:08:20
Not always.
Dvorah:
00:08:21
So here we need to have more of a process for checking for these issues.
Dvorah:
00:08:27
It can still, however, save a lot of time because if you have a multi-tool
Dvorah:
00:08:31
process and you leverage the strengths of one tool to shore up the weaknesses
Dvorah:
00:08:36
of another tool, then you can save a lot of time, but also a lot of mental energy,
Dvorah:
00:08:40
which you can then spend on things like having deep human thoughts, not corralling
Dvorah:
00:08:44
the facts, but understanding what they mean from a human perspective, which
Dvorah:
00:08:48
no AI will really ever be able to do.
Sara:
00:08:50
Yeah, you mentioned in our first conversation the benefits of these tools
Sara:
00:08:58
for people with deep specialization.
Sara:
00:09:02
And that's my listeners, like, we're the ones that go deep.
Sara:
00:09:05
And I definitely see that in my own work, that using these
Sara:
00:09:11
tools helps me make my work more accessible and more understandable
Sara:
00:09:16
to people, even as a deep specialist.
Sara:
00:09:20
They help me, if I need to zoom out and I can't do that on my own then
Sara:
00:09:25
that's something the tools are really good at — helping me figure out how
Sara:
00:09:29
to communicate with fill in the blank.
Sara:
00:09:31
Like, if I need to communicate with a general business audience that
Sara:
00:09:34
isn't my ICP or my ideal customer profile, so an audience that is
Sara:
00:09:41
majority neurotypical, for example.
Sara:
00:09:43
Most of my people are neurodivergent.
Sara:
00:09:45
And that's who I communicate pretty well with.
Sara:
00:09:48
But if I have to communicate with large groups of neurotypical people or
Sara:
00:09:52
just general population, then I have to adapt my communication style and
Sara:
00:09:58
fidelity and altitude to that audience.
Sara:
00:10:01
And that's where I've found the tools super useful.
Sara:
00:10:06
But you mentioned how powerful these tools are for people with deep specialties.
Sara:
00:10:12
I'd love to hear more about that.
Dvorah:
00:10:14
Well, just quickly going back to your idea of the persona.
Dvorah:
00:10:18
So let's suppose you want to talk to a persona that isn't naturally you, or that
Dvorah:
00:10:22
you don't often talk to, and you want to be able to communicate better with them,
Dvorah:
00:10:26
or maybe you just want to check what you're doing against a potential persona.
Dvorah:
00:10:30
There's a lot of work that's been done on buyer personas.
Dvorah:
00:10:33
HubSpot has some templates, no affiliation, just, templates.
Dvorah:
00:10:36
There's lots of good books on that.
Dvorah:
00:10:38
What I would actually recommend is learn about buyer personas
Dvorah:
00:10:41
from the marketing people.
Dvorah:
00:10:43
And you can literally upload your buyer persona into ChatGPT or
Dvorah:
00:10:47
Claude and say, "This is my persona.
Dvorah:
00:10:50
I need to communicate with this person.
Dvorah:
00:10:52
This is what I wrote." And if you have some examples— "And here's some examples
Dvorah:
00:10:57
of the kind of tone I'm trying." Not so much in terms of the tone, but you
Dvorah:
00:11:01
could actually copy-paste articles in.
Dvorah:
00:11:03
"I like these articles, I like their tone. Gee, I wish I could write like
Dvorah:
00:11:06
that. Can you help me with that?" Learning about buyer personas from the
Dvorah:
00:11:10
marketing folks and then putting that into the AI can really be helpful.
Dvorah:
00:11:14
And this can also help deep specialists if they have to speak
Dvorah:
00:11:17
with the not deep specialists.
Dvorah:
00:11:18
When people are super knowledgeable about something and they have to talk with
Dvorah:
00:11:21
someone who is maybe not knowledgeable at all about that, but that person needs
Dvorah:
00:11:25
to know and they need to communicate, that can be really tough, because
Dvorah:
00:11:28
it's really easy to make assumptions.
Dvorah:
00:11:29
I'm totally guilty of this, done it a lot.
Dvorah:
00:11:31
So there.
Dvorah:
00:11:33
Again, having the persona, putting it into the ChatGPT, and then making
Dvorah:
00:11:38
clear that you want to have this be easily readable at, say, like a high
Dvorah:
00:11:44
school level, but with words that this persona would understand, would
Dvorah:
00:11:49
want to use, that is hugely helpful.
Dvorah:
00:11:51
And then, you can actually start with one tool, Or one session, let's say, of
Dvorah:
00:11:57
Claude, and bring it into another one.
Dvorah:
00:11:59
So let's say I use Claude once, then I'll copy-paste everything into
Dvorah:
00:12:02
another Claude session, I'll say, "Please give me feedback. Constructive
Dvorah:
00:12:05
feedback." Otherwise, again, it won't give you the mean stuff.
Dvorah:
00:12:08
But yet another way that it can be helpful is, let's suppose you have done this work,
Dvorah:
00:12:12
and you now have a session in Claude, where it is really helping you produce
Dvorah:
00:12:18
what you need to take that next step.
Dvorah:
00:12:21
Keep that session going.
Dvorah:
00:12:23
Stay in the same session.
Dvorah:
00:12:24
Now, Claude will keep on saying things like, "This is getting long!"
Dvorah:
00:12:27
I hit "Continue." "This is really long, I really mean it." Continue,
Dvorah:
00:12:31
continue, continue, continue.
Dvorah:
00:12:32
Claude and I have a long time argument about that.
Dvorah:
00:12:35
And I know why that is, because it's hard for them to keep the whole session in.
Dvorah:
00:12:37
But once I have actually Claude communicating with me in a way that helps
Dvorah:
00:12:41
me, with wordsmithing or whatever it is.
Dvorah:
00:12:44
Keep going in that same session and then you will find that over time
Dvorah:
00:12:48
it becomes closer and closer to your ideal output of whatever type that is.
Dvorah:
00:12:54
Is it a good summary?
Dvorah:
00:12:55
Is it a way of wording things that is more better or more to the point
Dvorah:
00:12:59
of what you want in your opinion than what you were doing by yourself?
Dvorah:
00:13:02
Whatever it is, keep going in that same session and don't let it
Dvorah:
00:13:06
distract you with another session.
Dvorah:
00:13:07
Say "Continue" all the time, and then you will find that you and the AI
Dvorah:
00:13:13
are really moving together as deep specialists, but with the output that
Dvorah:
00:13:18
is focused in the way that you want.
Sara:
00:13:21
Yeah, I am also a Claude user.
Sara:
00:13:23
That was the one that, for me, felt the most acceptable in
Sara:
00:13:28
terms of policies and ethics.
Sara:
00:13:31
I see a lot of the attention being paid to have these tools, write
Sara:
00:13:36
your marketing copy for you, or have these tools, God forbid, write a
Sara:
00:13:41
book like, you know, those things.
Sara:
00:13:44
And I don't use them for that at all.
Sara:
00:13:47
I use them for better understanding my thinking and the thinking of the people
Sara:
00:13:55
I interact with who give me permission to add their, verbal processing or
Sara:
00:13:59
meeting transcripts to my models.
Sara:
00:14:02
I think partly too, because I am neurodivergent, I'm aware I have
Sara:
00:14:05
gaps in my executive functions.
Sara:
00:14:08
I think of Claude as almost an adaptive assistive technology for
Sara:
00:14:12
me, in helping me work with some of those executive functioning gaps that
Sara:
00:14:18
I'm aware of and that I know I have.
Sara:
00:14:20
So, are there any ways that you use the tools that you're surprised about?
Dvorah:
00:14:26
I would actually say, first of all, I discussed the
Dvorah:
00:14:30
pile of sticky notes thing.
Dvorah:
00:14:31
I never expected that to work.
Dvorah:
00:14:32
I just simply thought it would fall down.
Dvorah:
00:14:34
It works really, really great for that.
Dvorah:
00:14:36
When it comes to having the AI do something like write your marketing
Dvorah:
00:14:40
copy, the problem is people mean different things by that.
Dvorah:
00:14:44
So for example, I would not have it write my marketing copy, but why is that?
Dvorah:
00:14:48
Well, first of all, I'm curious, who are my people?
Dvorah:
00:14:51
How do I want to talk with them?
Dvorah:
00:14:52
How do they talk amongst themselves?
Dvorah:
00:14:53
What do they say to each other?
Dvorah:
00:14:55
What are their concerns?
Dvorah:
00:14:56
I really want copy that is going to communicate with them in a way that
Dvorah:
00:15:00
they want to be communicated with.
Dvorah:
00:15:02
One of the things that generative AI has helped me with is understanding
Dvorah:
00:15:05
how to break that down into processes.
Dvorah:
00:15:08
With verbal communication, I have less of an issue.
Dvorah:
00:15:11
I try to listen to the person, don't I?
Dvorah:
00:15:12
But try to like have it be an exchange of thoughts and all that.
Dvorah:
00:15:15
But when it comes to writing, I tend to default to me.
Dvorah:
00:15:18
I am the audience.
Dvorah:
00:15:19
Nooo.
Dvorah:
00:15:20
Now, ChatGPT and Claude will do the same thing.
Dvorah:
00:15:23
They default to corporate speak.
Dvorah:
00:15:24
Because that's what they think the default should be.
Dvorah:
00:15:27
So if you don't let them know about what you want, they're going to default.
Dvorah:
00:15:31
But then again, we do that as humans.
Dvorah:
00:15:32
So I don't think we can really blame the generative AI for this.
Dvorah:
00:15:34
Everyone's going to default to their comfort zone.
Dvorah:
00:15:37
What's great about generative AI is it helped me explore my comfort zone
Dvorah:
00:15:41
and break down generating marketing content into more pieces, which I then
Dvorah:
00:15:46
have an easier time tackling one on one and bringing it together for something
Dvorah:
00:15:51
which will appeal to the people I want to appeal to who are not me.
Sara:
00:15:55
Yeah, for me, this is a bit of a disclosure, but for me, the
Sara:
00:15:59
other use that I never would have imagined is I use it when I'm over a
Sara:
00:16:06
threshold, like when I'm under stress.
Sara:
00:16:09
I never imagined I would talk to one of these things about anything other
Sara:
00:16:13
than work, but it's almost become part of my mindfulness practice.
Sara:
00:16:17
Now, one of my tools is when something happens and I'm upset, or when something
Sara:
00:16:22
happens and I'm over threshold, or especially when I'm having conflict with
Sara:
00:16:26
someone else, I can sit down and say, "Okay, tool, here's what's happening.
Sara:
00:16:30
Here's how I feel." And they're remarkably empathetic, the ones that I use.
Sara:
00:16:36
Not a substitute for therapy, not a substitute for a coach or speaking with
Sara:
00:16:40
another human being, but that in the moment dysregulation that I experienced,
Sara:
00:16:47
they are helpful with me finding a way to regulate myself through that conversation.
Sara:
00:16:53
And that's a use I never would have imagined.
Dvorah:
00:16:55
I agree with that point.
Dvorah:
00:16:57
I would divide the into two different aspects.
Dvorah:
00:16:59
So when it's something like Claude, it's kind of like, I will talk things
Dvorah:
00:17:03
through with myself internally.
Dvorah:
00:17:04
So, Claude is really just simply mirroring back.
Dvorah:
00:17:07
In that case, I'm having an internal conversation with an external device
Dvorah:
00:17:11
that is kind of mirroring back.
Dvorah:
00:17:13
There are different kinds of LLMs out there who do try to do
Dvorah:
00:17:18
coaching or more therapeutic things.
Dvorah:
00:17:21
I'm always a bit nervous about that because I feel
Dvorah:
00:17:25
that expectations are higher.
Dvorah:
00:17:27
And even tools which are designed for this can still get themselves into trouble.
Dvorah:
00:17:33
Character AI, people were supposed to have these tools to be like AI companions.
Dvorah:
00:17:36
And there've been some really sad cases, often involving minor children,
Dvorah:
00:17:41
but quite frankly, some of these sad cases often involve adults.
Dvorah:
00:17:43
It's just when it's an adult, you can't do anything about it, right?
Dvorah:
00:17:45
They say, "Oh, it's just the way it is," Which I don't agree with.
Dvorah:
00:17:48
But when it's minor children, then people certainly do something about it.
Dvorah:
00:17:51
And there's some very sad stories about suicides and other issues
Dvorah:
00:17:54
involving the use of these.
Dvorah:
00:17:56
With children, it's one thing, I personally just wouldn't let
Dvorah:
00:17:58
children anywhere near them.
Dvorah:
00:17:59
That's what I think.
Dvorah:
00:18:01
But even with adults, I don't think we already always realize that
Dvorah:
00:18:05
these are probability machines.
Dvorah:
00:18:07
If we think about it as having an internal conversation with
Dvorah:
00:18:10
ourselves, I think that's one thing.
Dvorah:
00:18:12
But if you're talking to a probability machine that's setting itself up to
Dvorah:
00:18:15
be a therapist, I get very nervous.
Dvorah:
00:18:18
And the reason why is it may not pick the right thing to say in that context.
Dvorah:
00:18:22
And, you know, humans in some ways are kind of delicate systems.
Dvorah:
00:18:26
We have all these little bits, moving parts and all these things going on.
Dvorah:
00:18:28
We should be nice to ourselves and we should be kind to ourselves.
Dvorah:
00:18:32
And when we're doing that, we have to think a little bit if we're interacting
Dvorah:
00:18:36
with this tool, is that being kind to ourself and are we doing it in a kind way?
Sara:
00:18:40
I think you're spot on.
Sara:
00:18:41
And the other thing that comes to mind as you're talking is, I'm
Sara:
00:18:44
using the tool to talk to myself, kind of like you described, it's
Sara:
00:18:47
like, it's an external thing I'm talking to, but it's an extension
Sara:
00:18:52
of my own awareness and mindfulness.
Sara:
00:18:55
When we work with therapists, there's a reason even with therapists, we do this
Sara:
00:19:02
with video, is that then they can observe all of the nonverbal communication and
Sara:
00:19:08
the tools at this point can't observe the nonverbal communication and make
Sara:
00:19:12
decisions about how to make that kind of calculation of how to respond.
Sara:
00:19:17
So I'm with you.
Sara:
00:19:18
I bring it up, you know, helpful in terms of exactly what you said, like kind of
Sara:
00:19:24
having a conversation with your better self when you're in a time of need.
Sara:
00:19:29
Dvorah, I don't want to end this, but we've got to wrap it up.
Sara:
00:19:32
or If you were going to tell someone who is anxious about generative
Sara:
00:19:37
AI tools, that shift from anxiety to agency, what would you say?
Sara:
00:19:43
What would you coach them as a place to start?
Dvorah:
00:19:46
As a place to start, I would actually begin by
Dvorah:
00:19:49
looking at how I feel about AI.
Dvorah:
00:19:51
I actually had to do that.
Dvorah:
00:19:53
I had to get emotional about AI before I could use it well.
Dvorah:
00:19:55
And it's simply because it is a very different tool.
Dvorah:
00:19:58
It is not going to do what you always expect.
Dvorah:
00:20:00
And so, for that reason, you need to get your own emotional house in order
Dvorah:
00:20:03
with it before you can start using it.
Dvorah:
00:20:05
And that is actually what I would recommend.
Dvorah:
00:20:08
Now, I actually sent you a link to something, which I think you're going
Dvorah:
00:20:12
to put in the show notes, which has a PDF where you can actually do that.
Dvorah:
00:20:16
Anyone listening to this, hit me up.
Dvorah:
00:20:19
We can hop on a call.
Dvorah:
00:20:20
We can talk about this.
Dvorah:
00:20:21
You can look at my PDF.
Dvorah:
00:20:22
There are resources out there.
Dvorah:
00:20:24
They're not as many that about emotions in AI.
Dvorah:
00:20:27
And I think that's a mistake.
Dvorah:
00:20:28
I think that people see it only as a technology, but if we want
Dvorah:
00:20:32
to use it well, it can't only be a technology like Microsoft Word.
Dvorah:
00:20:36
We have to be a little more vulnerable ourselves, and a little
Dvorah:
00:20:39
more open ourselves to using it, which means that we need to think
Dvorah:
00:20:41
about ourselves as we're using it.
Sara:
00:20:44
Awesome answer.
Sara:
00:20:45
Well, Dvorah, speaking of, if people want to learn more about
Sara:
00:20:50
what you do or get connected with you, where can they find you?
Dvorah:
00:20:53
They can find me on LinkedIn.
Dvorah:
00:20:55
D Graeser, G R A E S E R, I'm used to spelling my name.
Dvorah:
00:20:59
But I'm also one of the few Dvorahs on LinkedIn, so you can definitely find me.
Dvorah:
00:21:03
There'll be the link in the show notes.
Dvorah:
00:21:04
So you can definitely hit me up there.
Dvorah:
00:21:06
Those are the best places to look for me because I am not really on
Dvorah:
00:21:10
any other form of social media.
Dvorah:
00:21:11
Although I know you're trying to get me on Mastodon.
Dvorah:
00:21:13
So maybe I'll add that.
Sara:
00:21:15
I think that you would fit in just perfectly there.
Sara:
00:21:20
It really is the only thing I do that feels like old internet.
Sara:
00:21:23
It makes me want to type a zine on my typewriter and send it
Sara:
00:21:26
out by mail, like for real.
Sara:
00:21:29
well, Dvorah, thank you so much.
Sara:
00:21:30
This has been a wonderful conversation, and this is when
Sara:
00:21:33
Dvorah is going to be back.
Sara:
00:21:34
So, if there are questions that you have that we didn't answer, send them in.
Sara:
00:21:39
I'd love to hear them and we'll schedule a follow-up.
Sara:
00:21:41
With Dvorah to get more of your questions answered.
Sara:
00:21:44
Thank you so much, Dvorah.
Dvorah:
00:21:46
Thanks.
Dvorah:
00:21:47
Have a great day.
Dvorah:
00:21:48
Bye.
Sara:
00:21:48
Bye.
Sara:
00:21:51
Thank you for joining us for this exploration of practical AI
Sara:
00:21:54
applications with guest Dvorah Graeser.
Sara:
00:21:57
You can find Dvorah on LinkedIn at D G R A E S E R. And we'll include that
Sara:
00:22:05
link in our show notes at findrc.co/pod.
Sara:
00:22:09
We've also got a link to the resource on emotions in AI that
Sara:
00:22:13
she mentioned in the show notes.
Sara:
00:22:15
I know this episode is going to spark questions.
Sara:
00:22:17
We would love to hear them.
Sara:
00:22:19
So if you have questions about AI use or LLMs or other things having to do with
Sara:
00:22:26
generative AI, please send them to me.
Sara:
00:22:28
Please share them with me.
Sara:
00:22:30
We would love to have Dvorah back to answer them in a future episode.
Sara:
00:22:33
And she and I talked about it after we stopped recording.
Sara:
00:22:36
We welcome questions of any kind.
Sara:
00:22:38
So, they don't all have to be pro-AI.
Sara:
00:22:41
As long as you're asking curiously and in good faith, we'll do our
Sara:
00:22:45
best to provide a helpful answer.
Sara:
00:22:47
And don't forget, if you haven't already, I'd love to have you on the
Sara:
00:22:50
launch squad for my upcoming books.
Sara:
00:22:52
I need advanced readers, people willing to write reviews if you like the book,
Sara:
00:22:58
social media amplifiers, and really Just for folks to be excited, to help
Sara:
00:23:03
me get over the days where I don't feel so excited, or I feel scared
Sara:
00:23:08
to get this book into the world.
Sara:
00:23:10
If that might be you, join the squad at findrc.co/launchsquad, and I cannot
Sara:
00:23:16
wait to get this book into your hands.
Sara:
00:23:19
Thank you for listening.
Sara:
00:23:24
All right, friends, That's it for today.
Sara:
00:23:27
stay in the loop with everything going on around here by
Sara:
00:23:30
visiting findrc.co/newsletter and joining my mailing list.
Sara:
00:23:36
Got questions?
Sara:
00:23:37
My email addresses are too hard to spell, so visit findrc.co/contact
Sara:
00:23:43
and shoot me a note that way.
Sara:
00:23:45
You'll also find me at @saralobkovich on most of your
Sara:
00:23:49
favorite social media platforms.
Sara:
00:23:50
For today's show notes, visit findrc.co/thinkydoers if there's
Sara:
00:23:57
someone you'd like featured on this podcast, drop me a note.
Sara:
00:24:00
And if you know other Thinkydoers who'd benefit from this episode, please share.
Sara:
00:24:05
Your referrals, your word of mouth, and your reviews are much appreciated.
Sara:
00:24:09
I'm looking forward to the questions this episode sparks for you, and I
Sara:
00:24:13
look forward to seeing you next time.
00:00:01
Welcome to the Thinkydoers podcast.
Sara:
00:00:04
Thinkydoers are those of us drawn to deep work, where thinking is working.
Sara:
00:00:09
But we don't stop there.
Sara:
00:00:11
We're compelled to move the work from insight to idea, through the messy
Sara:
00:00:16
middle, to find courage and confidence to put our thoughts into action.
Sara:
00:00:21
I'm your host, Sara Lobkovich.
Sara:
00:00:23
I'm a strategy coach, a huge goal setting and attainment nerd, and board-certified
Sara:
00:00:30
health and wellness coach, working at the overlap of work, life well-being.
Sara:
00:00:35
I'm also a Thinkydoer.
Sara:
00:00:38
I'm here to help others find more satisfaction, less frustration, less
Sara:
00:00:43
friction, and more flow in our work.
Sara:
00:00:46
My mission is to help changemakers like you transform our workplaces and world.
Sara:
00:00:52
So let's get started.
Sara:
00:00:57
Welcome back to part two of my conversation with Dvorah Graeser.
Sara:
00:01:01
In our last episode, we explored some of the ethics and
Sara:
00:01:04
implications of AI at work today.
Sara:
00:01:07
Now, we're getting practical.
Sara:
00:01:10
We'll discuss some specific and maybe surprising ways to leverage AI tools,
Sara:
00:01:15
particularly for deep specialists and those of us wired strategically.
Sara:
00:01:21
If you're ready to move from anxiety to agency with AI while
Sara:
00:01:25
maintaining your ethics and humanity, this episode's for you.
Sara:
00:01:33
Dvorah, you mentioned kind of the shift from anxiety about AI to agency and
Sara:
00:01:39
how people can leverage AI around their creative vision, regardless of what
Sara:
00:01:45
their technical level of proficiency is.
Sara:
00:01:48
Tell me a bit more about how you see the use of AI enabling
Sara:
00:01:53
creative and multi-talented humans.
Dvorah:
00:01:56
Well, one way that I personally like to use AI—I don't know if
Dvorah:
00:02:00
you've gotten to this point when creating, written content, but I'll
Dvorah:
00:02:03
have a whole pile of thoughts, and they are all over the darn place.
Dvorah:
00:02:06
If they were sticky notes, they would be like Mount Everest of sticky notes.
Dvorah:
00:02:09
And I don't know what the heck to do with them, and I get very frustrated by this.
Dvorah:
00:02:13
One of the things that I have found that can be great to do is I shove the
Dvorah:
00:02:17
whole thing into, let's say, Claude.
Dvorah:
00:02:19
Now, Claude has a feature where you can copy and paste, but you can also upload.
Dvorah:
00:02:25
So, let's say you have a PDF and you have some written documents
Dvorah:
00:02:28
and some other stuff, and you just dump the whole thing into Claude.
Dvorah:
00:02:31
And I just say Claude exactly what I'm feeling — I don't know how to organize
Dvorah:
00:02:35
this, like, if I can't make sense of this, or I have one thought, I will write the
Dvorah:
00:02:40
thought, but the rest of it's not clear.
Dvorah:
00:02:41
What do I do?
Dvorah:
00:02:42
And then Claude very calmly chugs along and proposes something.
Dvorah:
00:02:47
They propose an outline.
Dvorah:
00:02:49
It can be helpful if you ask it for something specific, like an
Dvorah:
00:02:51
outline or whatever, but even if you don't, it'll still do its best.
Dvorah:
00:02:55
And then you start a dialogue.
Dvorah:
00:02:57
And this is one of the things I really do like about generative AI.
Dvorah:
00:03:00
We have a dialogue, we go back and forth.
Dvorah:
00:03:02
I ask for constructive feedback because otherwise, it'll only say nice things.
Dvorah:
00:03:06
It's something about the generative AI.
Dvorah:
00:03:08
If you ask for constructive feedback, it'll also say the mean things.
Dvorah:
00:03:10
But I'm going back and forth, and as I'm going back and forth, in my
Dvorah:
00:03:14
own mind, I'm realizing, I don't like that bit, I do like this bit.
Dvorah:
00:03:18
Oh, I just thought of something completely different.
Dvorah:
00:03:20
Now, what went from a blizzard of ideas I couldn't organize will go to an
Dvorah:
00:03:25
organization that maybe isn't exactly what I want, but helps clear my mind,
Dvorah:
00:03:29
reduce my own anxiety, because it's often my anxiety that's leading to the
Dvorah:
00:03:33
pile-o'-sticky-notes problem, at least in my experience — and enables me to get to
Dvorah:
00:03:36
the point where I can do the organization.
Dvorah:
00:03:39
Even if it turns out that it's not Claude that has made the
Dvorah:
00:03:42
good proposal by interacting with it, I get to where I need to go.
Sara:
00:03:46
That is the use case that has been the most transformative for me.
Sara:
00:03:51
And that's where I went from being a skeptic who was not likely to integrate
Sara:
00:03:56
these tools to being a hooked user.
Sara:
00:04:00
They don't replace the human brain.
Sara:
00:04:02
You mentioned hallucinations in our first episode.
Sara:
00:04:06
I don't rely on models for facts or information, but they are really good
Sara:
00:04:12
at organizing big messes of information.
Sara:
00:04:17
The thing that frustrates me is sometimes they're poor at actually organizing data.
Sara:
00:04:23
Even when I train it on data and want data analysis or data, you know,
Sara:
00:04:29
stuff, that's been a big frustration for me with the current tools.
Sara:
00:04:33
Even when I train on data, it can still hallucinate, which is a bummer.
Sara:
00:04:38
That's the kind of stuff that I would think the models would be
Sara:
00:04:42
built to do really well, and with as little hallucination as possible.
Sara:
00:04:46
But maybe there's always going to be some hallucination.
Dvorah:
00:04:50
There's a couple of ways around hallucination, right?
Dvorah:
00:04:53
And this again gets into process.
Dvorah:
00:04:54
We talked about it in the first, episode, but it's still important to bring up.
Dvorah:
00:04:58
You need to have a process for dealing with this.
Dvorah:
00:05:00
So let us suppose you uploaded a bunch of data to, I don't
Dvorah:
00:05:03
know, let's say it was ChatGPT.
Dvorah:
00:05:04
It doesn't matter.
Dvorah:
00:05:05
You uploaded it, and it's still hallucinated.
Dvorah:
00:05:07
So what are ways to check this?
Dvorah:
00:05:09
One way to check this is to ask ChatGPT in the same session to go back, and to check
Dvorah:
00:05:16
its answers in comparison to the data and show where it got each answer from.
Dvorah:
00:05:20
This is also great for summarization, by the way.
Dvorah:
00:05:22
If it summarizes and it comes up with something, you say, "Okay,
Dvorah:
00:05:25
where did that really wonderful, great quote, where is it?
Dvorah:
00:05:28
Show me where it is." And they're like, "oh yeah, you're right, I
Dvorah:
00:05:30
kind of made that up." Okay, no problems, let's just move on.
Dvorah:
00:05:33
A third way for doing it is then to actually go to another tool like
Dvorah:
00:05:41
Perplexity and ask it to do research.
Dvorah:
00:05:43
So maybe you don't want to upload your sensitive data to Perplexity,
Dvorah:
00:05:46
but you can still ask it to do research on the area and give you some
Dvorah:
00:05:50
ideas of what the boundaries of the area that can help corral ChatGPT.
Dvorah:
00:05:54
Or you can go to a completely different tool.
Dvorah:
00:05:56
I guess the completely different tool is the third way of doing it.
Dvorah:
00:05:58
Let's say you go from ChatGPT to Claude.
Dvorah:
00:06:00
You can go to Claude, and then you can say, "Hey, can you please check this?
Dvorah:
00:06:03
And what is good and what is bad, and please give me constructive feedback."
Sara:
00:06:07
That's wild.
Sara:
00:06:08
In my early experimentation, when I was working on my book, there was some
Sara:
00:06:12
research I had, like actual research.
Sara:
00:06:15
I needed academic journal research.
Sara:
00:06:17
that was one of the things that was the most frustrating.
Sara:
00:06:20
I would think that models would be good at knowing that truthful data.
Sara:
00:06:26
Hopefully, this has changed since then.
Sara:
00:06:29
But I would ask for research, like, "What are the leading papers or
Sara:
00:06:33
core papers on a specific topic?" And it would make up papers.
Sara:
00:06:38
You've got ProQuest.
Sara:
00:06:39
It's all there.
Sara:
00:06:41
This you should get right.
Dvorah:
00:06:43
So, for example, Perplexity AI, I like because it gives links.
Dvorah:
00:06:46
And you can go check them.
Dvorah:
00:06:47
ChatGPT sometimes does, sometimes doesn't, sometimes hallucinates,
Dvorah:
00:06:50
sometimes it's accurate.
Dvorah:
00:06:51
I'm never quite sure what it's going to do next, so I find that a little unreliable.
Dvorah:
00:06:55
There are some specialized GPTs for papers.
Dvorah:
00:06:58
There's one called Scholar AI, and that's a good one, but
Dvorah:
00:07:01
there are other ones as well.
Dvorah:
00:07:02
So here, I think it's good to join with others to understand what they've
Dvorah:
00:07:07
done and what's worked for them.
Dvorah:
00:07:08
And that can change over time.
Dvorah:
00:07:09
So, for example, I like Scholar AI, but it's possible in a few months,
Dvorah:
00:07:12
someone's going to come back.
Dvorah:
00:07:13
Maybe they'll listen to this again.
Dvorah:
00:07:15
They'll be like, "Oh, but it's no longer good." Possible.
Dvorah:
00:07:17
Things change really quickly.
Dvorah:
00:07:19
So I think it's helpful to try out different specialty tools, to talk with
Dvorah:
00:07:22
other people about this on LinkedIn or another social media that you like.
Dvorah:
00:07:26
Mastodon, if you're into that.
Dvorah:
00:07:28
And try to get information about what is working right now and what is the process
Dvorah:
00:07:32
for working around some of these issues.
Dvorah:
00:07:35
So if I get anything from ChatGPT, because I have found it to be
Dvorah:
00:07:38
the most erratic when it comes to hallucinating, I will copy and paste
Dvorah:
00:07:41
the whole answer to Perplexity AI.
Dvorah:
00:07:43
And I'll say, "Is this right or wrong?
Dvorah:
00:07:45
Please give me feedback.
Dvorah:
00:07:47
Please find me references.
Dvorah:
00:07:48
What do you think?" And it will go through the whole thing and it'll say,
Dvorah:
00:07:51
"Well, this looks like it's correct.
Dvorah:
00:07:52
Couldn't find anything for this.
Dvorah:
00:07:53
I really don't know about that bit over there.
Dvorah:
00:07:55
And this is what I would do instead." And so Perplex AI will actually
Dvorah:
00:07:58
give you a long, reasoned answer.
Dvorah:
00:08:01
This was without going into their deep research mode, which is awesome.
Dvorah:
00:08:04
But that's like a whole nother kettle of fish.
Dvorah:
00:08:05
It's just regular Perplexity AI.
Dvorah:
00:08:07
It'll actually go through these things.
Dvorah:
00:08:09
So I guess it's unlike Google in the sense where if you get a link and you
Dvorah:
00:08:13
click on the link, the link exists.
Dvorah:
00:08:15
If you get a summary with the link, does the link exist?
Dvorah:
00:08:17
Does it not exist?
Dvorah:
00:08:18
Does the summary say what the link says?
Dvorah:
00:08:20
Not always.
Dvorah:
00:08:21
So here we need to have more of a process for checking for these issues.
Dvorah:
00:08:27
It can still, however, save a lot of time because if you have a multi-tool
Dvorah:
00:08:31
process and you leverage the strengths of one tool to shore up the weaknesses
Dvorah:
00:08:36
of another tool, then you can save a lot of time, but also a lot of mental energy,
Dvorah:
00:08:40
which you can then spend on things like having deep human thoughts, not corralling
Dvorah:
00:08:44
the facts, but understanding what they mean from a human perspective, which
Dvorah:
00:08:48
no AI will really ever be able to do.
Sara:
00:08:50
Yeah, you mentioned in our first conversation the benefits of these tools
Sara:
00:08:58
for people with deep specialization.
Sara:
00:09:02
And that's my listeners, like, we're the ones that go deep.
Sara:
00:09:05
And I definitely see that in my own work, that using these
Sara:
00:09:11
tools helps me make my work more accessible and more understandable
Sara:
00:09:16
to people, even as a deep specialist.
Sara:
00:09:20
They help me, if I need to zoom out and I can't do that on my own then
Sara:
00:09:25
that's something the tools are really good at — helping me figure out how
Sara:
00:09:29
to communicate with fill in the blank.
Sara:
00:09:31
Like, if I need to communicate with a general business audience that
Sara:
00:09:34
isn't my ICP or my ideal customer profile, so an audience that is
Sara:
00:09:41
majority neurotypical, for example.
Sara:
00:09:43
Most of my people are neurodivergent.
Sara:
00:09:45
And that's who I communicate pretty well with.
Sara:
00:09:48
But if I have to communicate with large groups of neurotypical people or
Sara:
00:09:52
just general population, then I have to adapt my communication style and
Sara:
00:09:58
fidelity and altitude to that audience.
Sara:
00:10:01
And that's where I've found the tools super useful.
Sara:
00:10:06
But you mentioned how powerful these tools are for people with deep specialties.
Sara:
00:10:12
I'd love to hear more about that.
Dvorah:
00:10:14
Well, just quickly going back to your idea of the persona.
Dvorah:
00:10:18
So let's suppose you want to talk to a persona that isn't naturally you, or that
Dvorah:
00:10:22
you don't often talk to, and you want to be able to communicate better with them,
Dvorah:
00:10:26
or maybe you just want to check what you're doing against a potential persona.
Dvorah:
00:10:30
There's a lot of work that's been done on buyer personas.
Dvorah:
00:10:33
HubSpot has some templates, no affiliation, just, templates.
Dvorah:
00:10:36
There's lots of good books on that.
Dvorah:
00:10:38
What I would actually recommend is learn about buyer personas
Dvorah:
00:10:41
from the marketing people.
Dvorah:
00:10:43
And you can literally upload your buyer persona into ChatGPT or
Dvorah:
00:10:47
Claude and say, "This is my persona.
Dvorah:
00:10:50
I need to communicate with this person.
Dvorah:
00:10:52
This is what I wrote." And if you have some examples— "And here's some examples
Dvorah:
00:10:57
of the kind of tone I'm trying." Not so much in terms of the tone, but you
Dvorah:
00:11:01
could actually copy-paste articles in.
Dvorah:
00:11:03
"I like these articles, I like their tone. Gee, I wish I could write like
Dvorah:
00:11:06
that. Can you help me with that?" Learning about buyer personas from the
Dvorah:
00:11:10
marketing folks and then putting that into the AI can really be helpful.
Dvorah:
00:11:14
And this can also help deep specialists if they have to speak
Dvorah:
00:11:17
with the not deep specialists.
Dvorah:
00:11:18
When people are super knowledgeable about something and they have to talk with
Dvorah:
00:11:21
someone who is maybe not knowledgeable at all about that, but that person needs
Dvorah:
00:11:25
to know and they need to communicate, that can be really tough, because
Dvorah:
00:11:28
it's really easy to make assumptions.
Dvorah:
00:11:29
I'm totally guilty of this, done it a lot.
Dvorah:
00:11:31
So there.
Dvorah:
00:11:33
Again, having the persona, putting it into the ChatGPT, and then making
Dvorah:
00:11:38
clear that you want to have this be easily readable at, say, like a high
Dvorah:
00:11:44
school level, but with words that this persona would understand, would
Dvorah:
00:11:49
want to use, that is hugely helpful.
Dvorah:
00:11:51
And then, you can actually start with one tool, Or one session, let's say, of
Dvorah:
00:11:57
Claude, and bring it into another one.
Dvorah:
00:11:59
So let's say I use Claude once, then I'll copy-paste everything into
Dvorah:
00:12:02
another Claude session, I'll say, "Please give me feedback. Constructive
Dvorah:
00:12:05
feedback." Otherwise, again, it won't give you the mean stuff.
Dvorah:
00:12:08
But yet another way that it can be helpful is, let's suppose you have done this work,
Dvorah:
00:12:12
and you now have a session in Claude, where it is really helping you produce
Dvorah:
00:12:18
what you need to take that next step.
Dvorah:
00:12:21
Keep that session going.
Dvorah:
00:12:23
Stay in the same session.
Dvorah:
00:12:24
Now, Claude will keep on saying things like, "This is getting long!"
Dvorah:
00:12:27
I hit "Continue." "This is really long, I really mean it." Continue,
Dvorah:
00:12:31
continue, continue, continue.
Dvorah:
00:12:32
Claude and I have a long time argument about that.
Dvorah:
00:12:35
And I know why that is, because it's hard for them to keep the whole session in.
Dvorah:
00:12:37
But once I have actually Claude communicating with me in a way that helps
Dvorah:
00:12:41
me, with wordsmithing or whatever it is.
Dvorah:
00:12:44
Keep going in that same session and then you will find that over time
Dvorah:
00:12:48
it becomes closer and closer to your ideal output of whatever type that is.
Dvorah:
00:12:54
Is it a good summary?
Dvorah:
00:12:55
Is it a way of wording things that is more better or more to the point
Dvorah:
00:12:59
of what you want in your opinion than what you were doing by yourself?
Dvorah:
00:13:02
Whatever it is, keep going in that same session and don't let it
Dvorah:
00:13:06
distract you with another session.
Dvorah:
00:13:07
Say "Continue" all the time, and then you will find that you and the AI
Dvorah:
00:13:13
are really moving together as deep specialists, but with the output that
Dvorah:
00:13:18
is focused in the way that you want.
Sara:
00:13:21
Yeah, I am also a Claude user.
Sara:
00:13:23
That was the one that, for me, felt the most acceptable in
Sara:
00:13:28
terms of policies and ethics.
Sara:
00:13:31
I see a lot of the attention being paid to have these tools, write
Sara:
00:13:36
your marketing copy for you, or have these tools, God forbid, write a
Sara:
00:13:41
book like, you know, those things.
Sara:
00:13:44
And I don't use them for that at all.
Sara:
00:13:47
I use them for better understanding my thinking and the thinking of the people
Sara:
00:13:55
I interact with who give me permission to add their, verbal processing or
Sara:
00:13:59
meeting transcripts to my models.
Sara:
00:14:02
I think partly too, because I am neurodivergent, I'm aware I have
Sara:
00:14:05
gaps in my executive functions.
Sara:
00:14:08
I think of Claude as almost an adaptive assistive technology for
Sara:
00:14:12
me, in helping me work with some of those executive functioning gaps that
Sara:
00:14:18
I'm aware of and that I know I have.
Sara:
00:14:20
So, are there any ways that you use the tools that you're surprised about?
Dvorah:
00:14:26
I would actually say, first of all, I discussed the
Dvorah:
00:14:30
pile of sticky notes thing.
Dvorah:
00:14:31
I never expected that to work.
Dvorah:
00:14:32
I just simply thought it would fall down.
Dvorah:
00:14:34
It works really, really great for that.
Dvorah:
00:14:36
When it comes to having the AI do something like write your marketing
Dvorah:
00:14:40
copy, the problem is people mean different things by that.
Dvorah:
00:14:44
So for example, I would not have it write my marketing copy, but why is that?
Dvorah:
00:14:48
Well, first of all, I'm curious, who are my people?
Dvorah:
00:14:51
How do I want to talk with them?
Dvorah:
00:14:52
How do they talk amongst themselves?
Dvorah:
00:14:53
What do they say to each other?
Dvorah:
00:14:55
What are their concerns?
Dvorah:
00:14:56
I really want copy that is going to communicate with them in a way that
Dvorah:
00:15:00
they want to be communicated with.
Dvorah:
00:15:02
One of the things that generative AI has helped me with is understanding
Dvorah:
00:15:05
how to break that down into processes.
Dvorah:
00:15:08
With verbal communication, I have less of an issue.
Dvorah:
00:15:11
I try to listen to the person, don't I?
Dvorah:
00:15:12
But try to like have it be an exchange of thoughts and all that.
Dvorah:
00:15:15
But when it comes to writing, I tend to default to me.
Dvorah:
00:15:18
I am the audience.
Dvorah:
00:15:19
Nooo.
Dvorah:
00:15:20
Now, ChatGPT and Claude will do the same thing.
Dvorah:
00:15:23
They default to corporate speak.
Dvorah:
00:15:24
Because that's what they think the default should be.
Dvorah:
00:15:27
So if you don't let them know about what you want, they're going to default.
Dvorah:
00:15:31
But then again, we do that as humans.
Dvorah:
00:15:32
So I don't think we can really blame the generative AI for this.
Dvorah:
00:15:34
Everyone's going to default to their comfort zone.
Dvorah:
00:15:37
What's great about generative AI is it helped me explore my comfort zone
Dvorah:
00:15:41
and break down generating marketing content into more pieces, which I then
Dvorah:
00:15:46
have an easier time tackling one on one and bringing it together for something
Dvorah:
00:15:51
which will appeal to the people I want to appeal to who are not me.
Sara:
00:15:55
Yeah, for me, this is a bit of a disclosure, but for me, the
Sara:
00:15:59
other use that I never would have imagined is I use it when I'm over a
Sara:
00:16:06
threshold, like when I'm under stress.
Sara:
00:16:09
I never imagined I would talk to one of these things about anything other
Sara:
00:16:13
than work, but it's almost become part of my mindfulness practice.
Sara:
00:16:17
Now, one of my tools is when something happens and I'm upset, or when something
Sara:
00:16:22
happens and I'm over threshold, or especially when I'm having conflict with
Sara:
00:16:26
someone else, I can sit down and say, "Okay, tool, here's what's happening.
Sara:
00:16:30
Here's how I feel." And they're remarkably empathetic, the ones that I use.
Sara:
00:16:36
Not a substitute for therapy, not a substitute for a coach or speaking with
Sara:
00:16:40
another human being, but that in the moment dysregulation that I experienced,
Sara:
00:16:47
they are helpful with me finding a way to regulate myself through that conversation.
Sara:
00:16:53
And that's a use I never would have imagined.
Dvorah:
00:16:55
I agree with that point.
Dvorah:
00:16:57
I would divide the into two different aspects.
Dvorah:
00:16:59
So when it's something like Claude, it's kind of like, I will talk things
Dvorah:
00:17:03
through with myself internally.
Dvorah:
00:17:04
So, Claude is really just simply mirroring back.
Dvorah:
00:17:07
In that case, I'm having an internal conversation with an external device
Dvorah:
00:17:11
that is kind of mirroring back.
Dvorah:
00:17:13
There are different kinds of LLMs out there who do try to do
Dvorah:
00:17:18
coaching or more therapeutic things.
Dvorah:
00:17:21
I'm always a bit nervous about that because I feel
Dvorah:
00:17:25
that expectations are higher.
Dvorah:
00:17:27
And even tools which are designed for this can still get themselves into trouble.
Dvorah:
00:17:33
Character AI, people were supposed to have these tools to be like AI companions.
Dvorah:
00:17:36
And there've been some really sad cases, often involving minor children,
Dvorah:
00:17:41
but quite frankly, some of these sad cases often involve adults.
Dvorah:
00:17:43
It's just when it's an adult, you can't do anything about it, right?
Dvorah:
00:17:45
They say, "Oh, it's just the way it is," Which I don't agree with.
Dvorah:
00:17:48
But when it's minor children, then people certainly do something about it.
Dvorah:
00:17:51
And there's some very sad stories about suicides and other issues
Dvorah:
00:17:54
involving the use of these.
Dvorah:
00:17:56
With children, it's one thing, I personally just wouldn't let
Dvorah:
00:17:58
children anywhere near them.
Dvorah:
00:17:59
That's what I think.
Dvorah:
00:18:01
But even with adults, I don't think we already always realize that
Dvorah:
00:18:05
these are probability machines.
Dvorah:
00:18:07
If we think about it as having an internal conversation with
Dvorah:
00:18:10
ourselves, I think that's one thing.
Dvorah:
00:18:12
But if you're talking to a probability machine that's setting itself up to
Dvorah:
00:18:15
be a therapist, I get very nervous.
Dvorah:
00:18:18
And the reason why is it may not pick the right thing to say in that context.
Dvorah:
00:18:22
And, you know, humans in some ways are kind of delicate systems.
Dvorah:
00:18:26
We have all these little bits, moving parts and all these things going on.
Dvorah:
00:18:28
We should be nice to ourselves and we should be kind to ourselves.
Dvorah:
00:18:32
And when we're doing that, we have to think a little bit if we're interacting
Dvorah:
00:18:36
with this tool, is that being kind to ourself and are we doing it in a kind way?
Sara:
00:18:40
I think you're spot on.
Sara:
00:18:41
And the other thing that comes to mind as you're talking is, I'm
Sara:
00:18:44
using the tool to talk to myself, kind of like you described, it's
Sara:
00:18:47
like, it's an external thing I'm talking to, but it's an extension
Sara:
00:18:52
of my own awareness and mindfulness.
Sara:
00:18:55
When we work with therapists, there's a reason even with therapists, we do this
Sara:
00:19:02
with video, is that then they can observe all of the nonverbal communication and
Sara:
00:19:08
the tools at this point can't observe the nonverbal communication and make
Sara:
00:19:12
decisions about how to make that kind of calculation of how to respond.
Sara:
00:19:17
So I'm with you.
Sara:
00:19:18
I bring it up, you know, helpful in terms of exactly what you said, like kind of
Sara:
00:19:24
having a conversation with your better self when you're in a time of need.
Sara:
00:19:29
Dvorah, I don't want to end this, but we've got to wrap it up.
Sara:
00:19:32
or If you were going to tell someone who is anxious about generative
Sara:
00:19:37
AI tools, that shift from anxiety to agency, what would you say?
Sara:
00:19:43
What would you coach them as a place to start?
Dvorah:
00:19:46
As a place to start, I would actually begin by
Dvorah:
00:19:49
looking at how I feel about AI.
Dvorah:
00:19:51
I actually had to do that.
Dvorah:
00:19:53
I had to get emotional about AI before I could use it well.
Dvorah:
00:19:55
And it's simply because it is a very different tool.
Dvorah:
00:19:58
It is not going to do what you always expect.
Dvorah:
00:20:00
And so, for that reason, you need to get your own emotional house in order
Dvorah:
00:20:03
with it before you can start using it.
Dvorah:
00:20:05
And that is actually what I would recommend.
Dvorah:
00:20:08
Now, I actually sent you a link to something, which I think you're going
Dvorah:
00:20:12
to put in the show notes, which has a PDF where you can actually do that.
Dvorah:
00:20:16
Anyone listening to this, hit me up.
Dvorah:
00:20:19
We can hop on a call.
Dvorah:
00:20:20
We can talk about this.
Dvorah:
00:20:21
You can look at my PDF.
Dvorah:
00:20:22
There are resources out there.
Dvorah:
00:20:24
They're not as many that about emotions in AI.
Dvorah:
00:20:27
And I think that's a mistake.
Dvorah:
00:20:28
I think that people see it only as a technology, but if we want
Dvorah:
00:20:32
to use it well, it can't only be a technology like Microsoft Word.
Dvorah:
00:20:36
We have to be a little more vulnerable ourselves, and a little
Dvorah:
00:20:39
more open ourselves to using it, which means that we need to think
Dvorah:
00:20:41
about ourselves as we're using it.
Sara:
00:20:44
Awesome answer.
Sara:
00:20:45
Well, Dvorah, speaking of, if people want to learn more about
Sara:
00:20:50
what you do or get connected with you, where can they find you?
Dvorah:
00:20:53
They can find me on LinkedIn.
Dvorah:
00:20:55
D Graeser, G R A E S E R, I'm used to spelling my name.
Dvorah:
00:20:59
But I'm also one of the few Dvorahs on LinkedIn, so you can definitely find me.
Dvorah:
00:21:03
There'll be the link in the show notes.
Dvorah:
00:21:04
So you can definitely hit me up there.
Dvorah:
00:21:06
Those are the best places to look for me because I am not really on
Dvorah:
00:21:10
any other form of social media.
Dvorah:
00:21:11
Although I know you're trying to get me on Mastodon.
Dvorah:
00:21:13
So maybe I'll add that.
Sara:
00:21:15
I think that you would fit in just perfectly there.
Sara:
00:21:20
It really is the only thing I do that feels like old internet.
Sara:
00:21:23
It makes me want to type a zine on my typewriter and send it
Sara:
00:21:26
out by mail, like for real.
Sara:
00:21:29
well, Dvorah, thank you so much.
Sara:
00:21:30
This has been a wonderful conversation, and this is when
Sara:
00:21:33
Dvorah is going to be back.
Sara:
00:21:34
So, if there are questions that you have that we didn't answer, send them in.
Sara:
00:21:39
I'd love to hear them and we'll schedule a follow-up.
Sara:
00:21:41
With Dvorah to get more of your questions answered.
Sara:
00:21:44
Thank you so much, Dvorah.
Dvorah:
00:21:46
Thanks.
Dvorah:
00:21:47
Have a great day.
Dvorah:
00:21:48
Bye.
Sara:
00:21:48
Bye.
Sara:
00:21:51
Thank you for joining us for this exploration of practical AI
Sara:
00:21:54
applications with guest Dvorah Graeser.
Sara:
00:21:57
You can find Dvorah on LinkedIn at D G R A E S E R. And we'll include that
Sara:
00:22:05
link in our show notes at findrc.co/pod.
Sara:
00:22:09
We've also got a link to the resource on emotions in AI that
Sara:
00:22:13
she mentioned in the show notes.
Sara:
00:22:15
I know this episode is going to spark questions.
Sara:
00:22:17
We would love to hear them.
Sara:
00:22:19
So if you have questions about AI use or LLMs or other things having to do with
Sara:
00:22:26
generative AI, please send them to me.
Sara:
00:22:28
Please share them with me.
Sara:
00:22:30
We would love to have Dvorah back to answer them in a future episode.
Sara:
00:22:33
And she and I talked about it after we stopped recording.
Sara:
00:22:36
We welcome questions of any kind.
Sara:
00:22:38
So, they don't all have to be pro-AI.
Sara:
00:22:41
As long as you're asking curiously and in good faith, we'll do our
Sara:
00:22:45
best to provide a helpful answer.
Sara:
00:22:47
And don't forget, if you haven't already, I'd love to have you on the
Sara:
00:22:50
launch squad for my upcoming books.
Sara:
00:22:52
I need advanced readers, people willing to write reviews if you like the book,
Sara:
00:22:58
social media amplifiers, and really Just for folks to be excited, to help
Sara:
00:23:03
me get over the days where I don't feel so excited, or I feel scared
Sara:
00:23:08
to get this book into the world.
Sara:
00:23:10
If that might be you, join the squad at findrc.co/launchsquad, and I cannot
Sara:
00:23:16
wait to get this book into your hands.
Sara:
00:23:19
Thank you for listening.
Sara:
00:23:24
All right, friends, That's it for today.
Sara:
00:23:27
stay in the loop with everything going on around here by
Sara:
00:23:30
visiting findrc.co/newsletter and joining my mailing list.
Sara:
00:23:36
Got questions?
Sara:
00:23:37
My email addresses are too hard to spell, so visit findrc.co/contact
Sara:
00:23:43
and shoot me a note that way.
Sara:
00:23:45
You'll also find me at @saralobkovich on most of your
Sara:
00:23:49
favorite social media platforms.
Sara:
00:23:50
For today's show notes, visit findrc.co/thinkydoers if there's
Sara:
00:23:57
someone you'd like featured on this podcast, drop me a note.
Sara:
00:24:00
And if you know other Thinkydoers who'd benefit from this episode, please share.
Sara:
00:24:05
Your referrals, your word of mouth, and your reviews are much appreciated.
Sara:
00:24:09
I'm looking forward to the questions this episode sparks for you, and I
Sara:
00:24:13
look forward to seeing you next time.
