Ep 82: AI Turned a Phase 10 Question Into 7.5 Million Simulations
Watch the YouTube video version above or listen to the podcast below!
Episode Summary
Alex shares how a casual Phase 10 question on vacation turned into an unexpected research project. After asking AI to compare the probabilities of different phases in the game, the experiment grew from a few thousand Monte Carlo simulations into 7.5 million simulations covering different player counts, wild cards, and skip cards. The project ultimately became a research paper and published dataset, prompting a larger discussion about how AI can help ordinary curiosity develop into serious research.
The hosts then compare their experiences with tools such as Claude, ChatGPT, Gemini, Fable, NotebookLM, and Copilot. They describe Claude as a thoughtful partner that often challenges assumptions, ChatGPT as useful for quick answers, Gemini as especially helpful for everyday and visual tasks, and Fable as feeling more like an on-demand researcher. Their experiences show that different tools are better suited to different kinds of work.
Ruthi raises concerns about the information bubbles that can form when an AI system remembers a user’s previous conversations and builds on a shared set of assumptions. While this context can make AI more helpful, it can also narrow the range of ideas it produces. Alex suggests testing important prompts in a fresh conversation to see whether prior context is improving the response or quietly limiting it.
The discussion then turns to authorship, attribution, and accountability. The group agrees that even when AI generates part of an email, presentation, research paper, or creative project, the person sharing the work remains responsible for its quality and accuracy. They criticize the habit of sending large amounts of unedited AI-generated content, arguing that this simply transfers the work of reviewing and filtering to the recipient.
The episode closes with a broader look at creativity, AI agents, and workplace productivity. Dave explains how he uses AI to support research, outlining, editing, and character development while keeping the final writing in his own voice. The hosts also question whether autonomous agents will truly reduce workloads or simply encourage employees to manage more tasks at once, turning AI-powered productivity into another source of burnout.
Ep 82: AI Turned a Phase 10 Question Into 7.5 Million Simulations Podcast and Video Transcript
Dave Dougherty: Hello, and welcome to the latest episode of Enterprising Minds. We got the whole crew here, which is awesome. We are, um, gonna have an interesting episode today, I think. I'm looking forward to this discussion. Uh, you know, a little bit of, little bit of news, a little bit of personal projects, a little bit of personal conflict, and, uh, therapy therein.
So if that doesn't get you excited, um, let's talk about accidentally writing a research paper. Alex, off to you.
Alex Pokorny: Okay. We can start with that one. All right, go for that one. All right.
Phase 10 Probability Quest
Alex Pokorny: Um, all right, long story short, basically I was playing a card game on vacation with my brother-in-law, and he was wondering the probability of one thing versus the other, and I was like, "Oh, that's an interesting question.
I don't know." We talked about it for, like, two seconds, and then later that night I was like, "Hey, Claude, what's the difference? Like, tell me the probabilities between all these different things for this given card game, like the outcomes." And it reported back two things that I thought were interesting, and the first was there is no existing research on this.
Um, and the game is called Phase 10. Turns out it's the second most popular card game for sales after Uno. So- Hm ... I never knew about it before my in-laws started playing it, so I thought it was some random old game. Turns out it's not. No, I
Ruthi Corcoran: love
Alex Pokorny: Phase
Ruthi Corcoran: 10.
Alex Pokorny: That's what they talked about. It's great. Oh. I've never done it.
It's, it's... At least a David said, "All right. Uh, hopefully to some people." They're all staring at their phones being like, "Of course, Alex. Of course, we know this." Um, but trying to figure out the probability of the different phases is tricky. Um, and apparently that's what Claude said is basically is that there is no existing data set, which I thought was interesting.
And the second thing is it said, because I was using Fable, it ran 3,500 Monte Carlo simulations, which basically means it tried to, you know, virtually deal the deck and see what happened. Um, and then it gave me probabilities, and I was like, "Oh, that's kind of cool." Um, so the next day on the long drive home, I was wondering like, hey, you know, there's a lot of like, you know, non-available data online from random papers and the lo- the rest.
There's scientific databases that you can post stuff- You know, should this be posted? Like, maybe that's useful.
Publishing the Simulations
Alex Pokorny: Um, I said it could, but I was like, "Well, can I, can I write a research paper on this?" So I have a research paper. It's, it's out there. Um, it ran 7.5 million simulations for two, four, and six players, along with some other things, including the wild card, skip cards and stuff.
So it ends up being a lot of simulations. 7.5 million in total. Um, but the data is live. I have a DIO, like a, um, number on it, which is the same thing if you had a dissertation. You'd get one of those too. I have a research ID now, so people can find my ORC ID, my researcher ID as well now. Um, and I submitted it to a publication, um, who I guess they're taking a look at it now, but then I'll see in a couple of weeks.
If not, I'll go to the next publication. So crazy. Um, just kind of a fun, random creative project.
Ruthi Corcoran: Mm-hmm.
Alex Pokorny: And yeah, something a little bit different. Um, I'm doing other stuff like that too. In the last week, that was one thing, but the other one was I found an integer set, so a set of random numbers that hasn't been found before.
So I submitted that one too, so I got credit for that one too. So I am now, versus last week, marketer. Now I am apparently a mathematician and researcher. So all thanks to Fable. So that brings us to our next question of attribution.
Ruthi Corcoran: Wait, hang on. This is so cool, Alex.
Alex Pokorny: Yeah. No, it's kind of fun. I
Ruthi Corcoran: think this is awesome.
I, I wanna hear more about your interactions with Fable. Like, how much probing did you have to do, or was it, was it similar- Oh, it's not much ... to the experience you have with other ones where it's just like, "Oh, here's an idea. You wanna run with this?" And you're like, "Yeah, let's run with that." Like, what, what did you find distinct about Fable versus some of the other models you've been using?
Yeah.
Um, tell us about how you found a new integer. Um- Sure ... and also just a side note that you should definitely send your Phase 10 research to a, a friend of ours who's very excited about statistics- Oh, that's true ... to see what he has to say. Yes.
Alex Pokorny: Oh, that'll be fun. Considering his AI stances as well, it'll be fun to talk about that one too.
Um.
Dave Dougherty: Is this
Alex Pokorny: the camera?
Fable vs Other Models
Alex Pokorny: Yeah, so I, I, um, use all the different models that are out there, uh, including the various ones from Gemini, the various ones from, um, ChatGPT, uh, all the stuff that basically that's available through Copilot. I haven't played around with the Meta ones yet or any of the Chinese-based ones yet, so I'm a little gapped on some of those.
I've tried some of them in the past. They just never were much really in comparison to the other kind of frontier models. Um, so maybe that's changed. Meta, I know, has been doing some really aggressive work on theirs, so maybe that's gonna change soon too. But so that, that's kind of like just a quick that's what I know kind of basis.
Yeah. Um, Fable is fantastic, to be honest. Um, it has a lot of safeguards pre-built into it, so if you ask it even just the barest of a biology question like, I don't know, like identify this, what kind of bird this feather could have come from, it will be like, "Nope, safeguard. Can't do it," and it'll switch you down to Opus.
Um, it, it is very, very finicky on that, but that's because there are always concerns of bio-terrorism, like ingredient creation and stuff like that. So, um, they've d- they've put a lot of safeguards on that one. So that, that model can be a little particular to use because you get kicked down to Opus 4.8 pretty frequently, um, depending on what you're working on.
With Phase 10, that was totally fine. It was able to make the whole paper. It was good with that. Um, I started trying to do some other projects, and it absolutely was like, "Won't do it." The big difference, in short, that I found with Fable versus prior models was I now feel like I have an online access to a person who is a researcher at my beck and call basically.
That is a huge step change difference than Opus 4.8, which is good, and it can think of things more complete. Um, a lot of the ChatGPT models don't. They'll give you what I consider, like, the fast answer. Like, if someone researched it for five minutes or clicked on the first three listings on Google, and they summarized an answer, that's the kind of answer that ChatGPT gives you frequently.
To the point even where I found ChatGPT in the last week, using the thinking model, the kind of higher level models and stuff like that, I asked it for a, a list of federal departments by employee size, and it gave me a list, and one of them was listed twice Like, so it's just an amateur mistake. Like, it's small.
It's, you know, easy to be like, "Whatever," and keep on going. But it was just ... It was missing basic things. Like, there were commonalities between a bunch of them that should have been stated considering my goal that I had already stated to it.
Ruthi Corcoran: Mm-hmm.
Alex Pokorny: Um, there were ... It's just ChatGPT is one that I use if I run out of credits with, uh, Claude, basically.
Yeah. And then I always copy and paste the work back to Claude once I have credits again and saying, "Critique this from ChatGPT," which if you, by the way, if you say to any of these models, there's a fun paper out there on this. If you say to Claude that this is something from Claude, it will be a lot softer.
If you say to Claude this is from ChatGPT, it will be much more critical. Um, so depending on how you want it to take that information in, um, you know, you can always phrase it however you like. Um, it's, it's, it's a funny paper. There's a lot of weird things that AI agents do. Um-
Ruthi Corcoran: I would love to see the difference of, you know, I write something and get Claude's feedback on it with Claude thinking it's me.
Alex Pokorny: Right.
Ruthi Corcoran: And then I write something and say, "This is ChatGPT writing it. Critique this." Yeah. And, like, what, what the difference is.
Alex Pokorny: Okay. That, that's mean. That's my writing.
Yeah, but Fable was amazing because it was ... Instead of ChatGPT, which is kind of giving you that quick research summary.
Ruthi Corcoran: Yeah.
Alex Pokorny: Um, the Opus models were good in that they'll give you an answer plus a bit more, so it will kinda understand a little bit more of your goal. It might push back even before it- Yes
starts in saying, like, "Do you ... You're, uh, talking about two different things here. Like, do you want this or do you want that?"
Ruthi Corcoran: Which I so appreciate, yes.
Alex Pokorny: I do too. Mm-hmm. Because there's a lot of these kind of questions, especially projects like this, that are very open-ended, and, you know, it's not really clear what the end goal will be.
I mean, I went through- Right ... multiple iterations, even the short story that I talked about, of, you know, what my end point was supposed to be. Um, so that, that kind of back and forth is really, really helpful. But also a completeness of answer that not just says, "Here is the answer to your question," the factoid kind of thing, which is the ChatGPT version.
Instead it's a bit more of, "Okay, here's what you wanted, and then here's where it can go, and tho- those are the pieces of information you need to know about that step." Um, what I found with Gemini is it, it loves to answer a follow-up quest- or it would love to ask a follow-up question. Um, and as you ask it things, it will come up with a follow-up question, which sometimes not completely relevant at all.
Ruthi Corcoran: Yes.
Alex Pokorny: Um, but it's just trying to continue a conversation. It's just like a- Yeah ... chatbot, you know, which there's been chatbots for decades now at this point. Um, and that just try to get you to keep interacting. And that- Yeah ... seems to be its goal. Um, ChatGPT seems to be, like, speed, and then- Mm ... uh, Claude seems to be completeness, so.
Ruthi Corcoran: And when I have, I've mostly been using Opus over the last, we'll call it two weeks in particular.
Alex Pokorny: Yeah.
Ruthi Corcoran: And I have also noticed that where I will give it... I'll ask a question or say, "Help me think through this," and it will start off by first clarifying what it is I'm asking or perhaps saying, like, "What if we think about it this different way?"
and then ask- answer that question. And, and by and large, the way that it's rephrasing or clarifying does indeed get to what I was trying to get to, but perhaps couldn't. Mm-hmm. And I'm often using it in the context of a project. And so-
Sure ...
I fo- I found it super useful, uh, where it'll be like, "You were exploring this thing."
It's assuming we're exploring a similar train of thought. Let's build off of that and reference it.
AI Bias and Bubbles
Ruthi Corcoran: And then the thing I worry about with that, though, is how much of an information bubble am I putting myself in.
Yes.
And how do I check against that, right? Because if it's based- Yeah ... off for previous conversations and it's like we created a shared understanding here and we're building off of that, how much am I in my own very solid bubble, so that when I talk to somebody else, they're like, "What on earth are you talking about?"
Because I only have the shared understanding with Claude, I don't have it with other people. And this might be very speculative, and especially when I'm working in, sort of trying to better understand an area I don't know, it's hard to know, like, where, what's speculation, what are the probabilities of different outcomes versus, yes, this is sort of known and proven.
Alex Pokorny: Yeah. I, I've run into a similar thing with projects of it takes prior chats and it, it starts to really narrow down too much, to the point where- Yeah ... I'll copy and paste a prompt to an incognito chat and ask it- Mm ... again via that just to- Mm ... compare and contrast the responses to see if, like, the creativity basically is being reduced with it.
Um, or restating the goals of, like, okay, for this response, I don't want the prior goals. My goals now are these things in this order. Or, like, rank it by this, critique it with this in mind or something like that. Mm. Just to kind of keep trying to break it out of those, those cycles where it starts to really narrow in a lot.
As for information that you know versus, you know, IRL, in real life, I don't know. That's, that's a tough one. I haven't run into that too much lately just because I've been- ... staying to the stuff I know or, you know, barely outside of it. But-
Dave Dougherty: Yeah, I've, I've been finding, at least for me, with between the three of them, Alex, you're right on point with Claude is a great kind of thought partner.
You know, Ruthi, same thing. Yeah. I, I really like the fact that it pushes back, and I can just have kind of more of a general conversation with it.
Ruthi Corcoran: Mm-hmm.
Dave Dougherty: You know? Um, 'cause I find my outputs are better when I don't type, actually. 'Cause then I can just, you know, go the speed of my brain.
Everyday AI Use Cases
Dave Dougherty: I'm using Gemini in more of the day-to-day random questions.
Alex Pokorny: Yeah.
Dave Dougherty: You know?
Ruthi Corcoran: Yeah.
Dave Dougherty: Like, I had to measure my front door for a storm door the other day, and you know, it's one of those things where it's like, I'm a close enough kind of person, but I know that that doesn't help when you're trying to build something. So I took, I took the tape measure, and I put on Gemini Live, and I just said, "This is what I think it is for this, you know, storm door measure.
Is that right, or am I reading this wrong?"
Ruthi Corcoran: Yes. Like,
Dave Dougherty: here's
Ruthi Corcoran: the
Dave Dougherty: model
Ruthi Corcoran: that- All the- It was funny ... all the DIY house stuff, I usually start with Gemini. Absolutely. I'll check with Claude, but I'll usually start with Gemini.
Alex Pokorny: Do you do the same thing- Yeah ... where it's Gemini Live, or- Yeah.
Ruthi Corcoran: I have not used Gemini Live.
I just say- Okay ... "This is the project I'm trying to do. I know nothing." Mm. "Tell me."
Alex Pokorny: Okay. Yeah. Yeah. I, I use the chat functionality, like, really frequently. The voice functionality, um, with ChatGPT was actually pretty good, but the way I was recording it for projects and stuff was messy in the past. Mm-hmm. And then Claude, I even had entire sessions get, like, deleted or missed.
Ruthi Corcoran: Yeah.
Alex Pokorny: It would type it all out, and then it would just, like, not know about it, and if you refresh, it was gone. Like, it was... There were some weird bugs. So I've- Mm ... which literally is like, you know, a handful of uses over time. So I gotta get more time on it to really get a good opinion. But that's awesome, Dave.
Like, that's a really cool idea, like using the live camera, being like, "This." Yeah. "What is this?"
Dave Dougherty: Read this tape measure. This is the door I wanna buy.
Alex Pokorny: Yeah.
Dave Dougherty: Tell me if I'm okay, because I think- Yeah ... I am, but I don't know, and I don't enjoy this kind of work, so let's just- Yeah ... get it done.
Alex Pokorny: Yeah. No, that, that's awesome.
I mean, that's, that's really smart too.
Dave Dougherty: So, um-
Alex Pokorny: Get those second opinions.
Dave Dougherty: Yeah. Well, and then they sent somebody out to measure it anyway, so I'm like, "Cool. Glad I bothered."
Alex Pokorny: Oh, man. Better than a window shade that I ordered a little bit too precise. It's, like, hard to pull down now because it literally fills side to side- ... perfectly. So if you don't, if you angle it at all, it's, like, hard to pull. It's like, I should have known, which I didn't, that window shades are, are a little bit different the way that they're measured because of that.
There's a, there's a-
Ruthi Corcoran: Yes ...
Alex Pokorny: difference there. Yes. But I didn't know that, so when it said, you know, measure it, I measured it, which it's just... Dave, that's a really smart usage because if I had used a tool like that, it would've said this, but in this case you don't do that. Or was it- Yes ... or you do? I don't know the difference.
All
Dave Dougherty: right. Right. Yeah,
Ruthi Corcoran: I'm not about to know-
Dave Dougherty: This reminds me- ... the
Alex Pokorny: difference. Mm-hmm ...
AI Goes Mainstream
Ruthi Corcoran: signs of the sort of adoption curve. My aunt, um, is currently using ChatGPT to help her redesign her fireplace fa- like, overlay. Oh. So she's got sort of a brick fireplace, and she has taken a picture of it, and then she gets five a day she tells me for ChatGPT to give her new alternates of, like, here's manipulation of the image, of how it would with this different effect, this different effect.
And I go, "Okay, th- that's, that's an important part of the adoption curve-"
Alex Pokorny: Yeah. Mm-hmm ... "
Ruthi Corcoran: that my aunt is using it for something like this." And, you know, she's... I wouldn't... She's not particularly technically savvy, but she's not als- also behind the times. Mm. And it goes, okay, this is now part of, part of more and more people's lives.
Dave Dougherty: So along those lines, my dad, who's, you know, 71 retiree, um, he is part of the MG, North American MG Club, right? So old British sports cars. Teeny, tiny little things like- Yeah ... you know, basically real life Hot Wheel cars. Um, they... He is the editor of the MG Gazette as part of his, uh, retirement. So he, he writes little articles for any of the MG enthusiasts, and he was recently going to research, um, MG as a company a- in the context of innovation.
Like how innovative has MG been over the years, which most people in the US probably don't know MG unless you're like a total, you know, grease head, right? Um, but he started using ChatGPT, and he started paying for it, and he only had to write his little article, but because he's an English teacher and an academic, he started doing deep research on it, and now he has, um, basically an outline for a 500-page book that he's very strongly considering writing.
Alex Pokorny: 500 page. Okay.
Dave Dougherty: Yeah. And he's like, "Yeah, I just spun this up in an afternoon. This is amazing." And I'm like, "Wow, okay. Yeah."
Alex Pokorny: Yeah. "
Dave Dougherty: Cool. You went from kind of putzing around with it to totally next level. That's pretty cool." You know? Yeah.
Alex Pokorny: Should
Ruthi Corcoran: publish it. That's really cool.
Dave Dougherty: Mm-hmm.
Ruthi Corcoran: He should go talk to Fable, it sounds like, because I feel like- Yeah
he's not gonna get dinged by Fable for talking about cars, and sounds like Alex-
Dave Dougherty: has had great success publishing already, so. Yeah. Yeah, and then once I told him about Google's, uh, NotebookLM, the academic in him got so excited because he's like- Oh ... "Yeah, it outlined it in Chicago style with footnotes, and I didn't have to putz around with that. This is amazing."
Alex Pokorny: Sir, your standards are above things I would even n- not even know about.
Dave Dougherty: Yeah. So-
Ruthi Corcoran: Ooh, that's so cool ... it is cool to use these things.
How to Cite AI
Ruthi Corcoran: Which brings us to AI citations.
Alex Pokorny: Yeah.
Ruthi Corcoran: That is a great, yeah. How will you... You might be an expert to bring on the show. The question before the court is- ... how do you cite AI?
Alex Pokorny: Yeah. This is a really, I don't know. I've been struggling with this one, and I, I know this has come up, um, kind of as a mini topic a few times during prior episodes.
Dave Dougherty: Mm-hmm.
Alex Pokorny: So it'll be cool to get the three of us, kind of our opinion on it.
Ruthi Corcoran: Yeah.
Alex Pokorny: Uh, basically is how do you reference the work that you've created with AI? Are you giving credit to it? Are you kinda hiding behind it saying like, "AI created this, I don't know if it's any good." Like, kind of like this is an early draft kind of caveat.
Um, or are you like, "This isn't something that I feel, you know, responsible for. This thing created it, I didn't." Um, in the case of, it was interesting the way that I had to put a disclaimer in that research paper, of course, because- Mm ... it ran the data. And while I was able to run it a few times and have the Python and the raw data and everything else files, it still, it, it ran it, right?
Dave Dougherty: Right.
Alex Pokorny: So at the end, there is a, a disclaimer paragraph that says that basically that I directed it, and I'm responsible for it, and that it did the research work, right? And I thought that was kind of an interesting thing because we've talked in the past of like feeling almost like a curator where you're commit- or commissioning, uh, an artwork that it created, right?
'Cause you put in a prompt, you did, you know, .1% of the work and it did 99% of the work, and you got this cool-looking image at the end of it. And then it's like, yeah, I'm, I've maybe commissioning it because you went to, uh, you know, an artist and asked for it to be done because you didn't create it. But at the same time, on the reverse, if you take a photo, do you credit your camera?
Or do you say, "I took this picture when I was on vacation." Right? So why does the tool now get more credit than the tool before? '
Dave Dougherty: Cause people anthropomorphize AI way stronger than a camera.
Alex Pokorny: Keep going, Dave. Like...
Ownership vs Outsourcing
Dave Dougherty: So I've struggled with this too, because having been someone who has outsourced work to people, on the one hand I'm responsible for creating the project, outlining it, scoping it so that they're capable of doing it.
So I, you know, I definitely take ownership of that. But then once you hand it off to them, I mean, it's kind of their thing, right? I mean, if you were... Let's just assume a human right now. You hand it off to a human, they go, they do it. They're responsible for the output, and then as the person who commissioned it, you have to check whether or not it meets standards, right?
Prompting Like A Brief
Dave Dougherty: Uh, especially in the agency days, right? Like, all right, the client wants this. Did you execute that? Or, you know, do we have to quick turn this around to make it way better? Um, you know, those are the conversations that we're used to having. Um, when it comes to AI, I think it's, it's a very similar thing where if you're prompting in the, it's not necessarily the right way, but the way that allows you to have it be more personalized to you, you know, and it's, and it's really methodical, and here's the scope of what I'm trying to do.
Here are the things you should know. Here are the players. Here's what would be considered good output. Here's something that would be considered bad output, yada, yada, yada, yada, right? I mean, you get into those really long prompts.
Ruthi Corcoran: Yes.
Dave Dougherty: So you're definitely responsible for that. But that output, yeah, I mean, that's, that is where it's weird.
Why AI Output Feels Generic
Dave Dougherty: But then, Ruthi, based on what you said to me the other week, this is what had me in this thought where I'm not interested in reading somebody else's AI output, right? But how much different is that from the actual... If, if I, if somebody else put together a research paper, how much different is that really, honestly?
If you're just curating sources and putting together some spin on it of these five sources have said this, therefore the logical conclusion is, you know, X. I, I don't, I don't... That's not very far off from that intern thing for me.
Alex Pokorny: I definitely see some distance there in terms of, um, the type of output that you'll get is much more of, like, if I- So the given industry that I, or niche industry that I work in- Mm-hmm ... um, there's a lot of purchasable papers that I can get on market research, market demand- Mm-hmm ... stuff like that, right? And I kinda know when I'm getting a paper like that, that it's going to be generic.
Um, it's gonna say kind of their own use of terms and the like- Right ... that they've defined themselves. Um, it's going to cover a lot of topics which are outside the scope of necessarily what I'm interested in. Um, and it's gonna mean that I'm gonna sort through, you know, 30 to 50 pages to kind of find the stuff that I'm interested in and get the pieces that I want out of it.
Dave Dougherty: Right.
Alex Pokorny: Um, if I ask a person about a specific goal in mind and after it, what they present back typically is more of an argument. You know, instead of just a, "Here's a reference material piece," instead it's much more, "Here's a defined piece that answers your question yes or no," or- Mm-hmm ... aggressively yes, aggressively no, or weakly yes, weakly no.
You know, that kind of a thing, right? That's more of that kind of the, the, the output gets created much more personal. Mm-hmm. And then it's also a little bit shorter, more concise, and kind of hits the points that matter, right? If you're looking for general reference, though, once you hit that level, maybe there isn't a difference.
If you're doing that in-between piece, which it's much more of like the market research piece, then it hits a line between this is gonna be more generalized or this is gonna be more specific. And typically with the AI outputs, they're very generic. Um-
Dave Dougherty: Right ...
Alex Pokorny: so it's not as useful reading as if someone else wrote it, because someone else would've put their, their spin on it after consuming the material, cr- you know, their opinion shows up, and then they wrote down their opinion plus the research, right?
And that opinion is the thing that I care about. The research is-
Ruthi Corcoran: Yes ...
Alex Pokorny: I'll find it later. Uh-
Ruthi Corcoran: Yeah, I think This is such a, a interesting line of thought.
Accountability And Editing
Ruthi Corcoran: So a few thoughts. The first one is, I think there- we're currently in the world in which
Agents and assistants, chatbots are acting on my behalf. We're not quite in a space where we've got autonomous acting ones, so I'm gonna set that aside 'cause that might change how I think about these things-
Mm-hmm ...
um, in the future. But for now, like, to me, if an AI writes something and I email it to you, or I create a Jira story, or, um, I, I create a PowerPoint presentation, like, I am accountable to everything that goes in to that communication presentation.
It's coming from me, even if an AI wrote it. I started out putting in, like, um, in certain communications, like, part of this was AI generated, just because it was sort of new, it was emerging, and I wanted to be transparent about it. I think at least in, in my work culture, we've reached the point where it's so ubiquitous that it's just assumed, whereas before it wasn't assumed.
Um, and so now you don't even have to put it in because the presumption is like, yes, you worked with AI to write it. And with that, I think there's a dis- difference between I worked with AI to write this thing, and by sending it, I'm putting my stamp on approval, which means I'm gonna do the work ahead of time to make sure that it's good and I stand by the things that are in it.
Right.
And what I see happening and, and what, um, I've seen in my own workplace is- Not doing the editing, not doing the, "I'm gonna read through and make sure I agree with this." Instead it's, "I'm gonna move as quick as I possibly can. I'm gonna send stuff out."
Yeah.
And I'm not even gonna worry- Yeah ... about the contents of it, and I'm gonna...
Where I really take issue is when I, I see this with coworkers who put, who put the time and work on the receiver. They say, "I couldn't be bothered to take this 70-page presentation and down it to seven that you actually care about or that would, that are getting to the relevant argument. I'm gonna make you do all the effort to sift through the garbage that my AI created."
Mm-hmm. Like that, that- Is not taking ownership of the work that you're doing with AI. Um, so I think there's a cultural component to it.
Share Prompts Not Dumps
Ruthi Corcoran: Um, and I was thinking about this research point that Alex, you brought up too, of like the opinion is what you care about, less so the research. Mm-hmm. And I wonder if in those cases where you're like, "I want other people to be on the same page," or to be able to have a little background information, if it's better to be like, "Here's the prompt."
Or, or Alex, you've shared, uh, and Dave, I think you have as well, "Hey, here's the, the starter conversation I had with ChatGPT-" Mm-hmm. "... but now you can go continue that conversation," which then puts the onus in my court to be able to have the conversation. It gives me agency versus having to read your conversation- Right
which frankly isn't that interesting.
Dave Dougherty: Right.
Different Communication Styles
Alex Pokorny: Yeah, and the different learning styles there too. That's an interesting point. Um, I kind of found that with quite a few people, like written down information, and maybe also in the method that, you know, that I prefer reading things, is significantly different than the way that they'll ever take in the information.
Mm-hmm. Like I had an old boss that it was a slide with three bullet points. If you hit five, you probably were needing another meeting with her. Like that was it. You wanna write an email? That's pointless. You wanna write a Teams message? That's pointless. She wanted a slide. One slide, three bullet points, that's the most, and they better be short.
Like, but I found with her, um, what really resonated was slides that did builds. So even if I had a graph, I would present, you know, just the X Y axis and talk about that, and then we would add in the bars or the line, and then we'd talk about- Mm ... where the point is, and then we'd put like a arrow or a star at the point that we care about, and then talk about why this matters.
Um, because I knew also if I threw a chart, she would gloss over it, and she usually got distracted by something else on her computer or phone or something like that real quick. So different people. That's kind of an interesting point though with the AI tools of being able to hand off. That's a really cool thing.
Because I'm sure what she would've preferred was instead of me writing a lengthy email, which to me makes sense, um, she would've preferred, uh, a summary of it or an image-based version of it that she can consume a lot faster than, versus me. In text I can consume a lot faster, right?
Dave Dougherty: Right.
Alex Pokorny: Which is totally fair that we have different styles.
So interesting if the tools can do it.
Reading Between The Lines
Dave Dougherty: Well, and there's the interesting point too with a, a lot of the communication studies is, you know, if you take the same message but it's delivered by two different people, three different people- There will be vastly different takeaways
Alex Pokorny: Oh, true
Dave Dougherty: You know? So I mean, this is the same thing where, okay, do you want it the short and snappy?
Do you want the video? Do you want, you know... Is it a heavier subject, so do you need more authority? So then do you lean toward, you know, some biases towards authority you know? Sure. Or do you not, right? Like, one of my favorite assignments... Now granted, I know nobody's gonna go do this, but it would be cool if people did.
Um, go read the CEO letters for the annual reports for publicly traded companies, and then read between the lines and write down everything they're not saying All right?
Alex Pokorny: Yeah.
Dave Dougherty: Then go take that and compare it to their most recent Glassdoor ratings.
And you can start getting a picture of what the culture's like, what's actually happening, what's not actually happening, what the investors might be missing when, you know, you get those rosy colored, you know, 10-Ks that are released. Um, it's a fun little experiment.
AI And Music Royalties
Alex Pokorny: So Dave, with your... You've done some poetry writing, you've done some- Mm-hmm.
Have you tried around any of the music tools with AI yet? Where do you see- I have not ... kind of like from the musician standpoint of where credit is due or not due with regards to kind of more art creative work?
Dave Dougherty: Well, the music thing is interesting. Uh, the
The music industry, industry has always been unkind to the artist, and that's fine. You know that. You're getting into it. Like, that's... It is what it is. What's interesting now, having started my music career in the aftermath of Napster, when everybody stopped paying for music, right? Um, to where we are now with the streaming services, which is like, you know, I'd kinda rather you steal it instead of, you know, saying you're paying me when it's, you know, a tenth of a penny per stream.
Like, so what? You know? Um, now Spotify says it doesn't have to pay you until you reach at least 20,000 streams. Why? Because they own their own platform. Um, so you're just not gonna be paid because you're not fitting the algorithm, right? Um, and does anybody bother to, like, look into this when they're streaming music?
No, they just wanna stream music. Um, which, great. Consume music. Everybody should, you know? I mean, there, there's a reason it's there. Um, the business of it is crap, and has gotten more crap, um, because the AI tools have essentially scraped YouTube without permission, or other digital libraries, um, and then created their own thing.
And so now, um, essentially the largest organizations like Universal Music Group or, um, you know, other record companies like that are cutting deals with the AI so that they get licensures with the AI companies. But all the independent artists don't get that same treatment, right? The independent artists get invited to upload your catalog to the AI tools so that they can then leverage the creativity, but then do you get a royalty for being included in the library?
No, you don't. Um, so for me, I have not done anything with AI for my music aside from "I have this weird chord progression, give me some options." Totally in the chatbot. Because, you know, at least with music, the AI can't listen to what you're doing and then give you suggestions, right? And it can't really look at what you're playing.
I mean, maybe if you're doing live mode, I've never tested this, but maybe it could, but it defaults to really standard voicings, because the vast majority of content is for, you know, beginning, beginner players, right? So if I'm using a jazz chord in a weird way, it's gonna not, it's not really gonna know what to do with that.
So, um Yeah. I, I keep music for me, so I don't really use it other than to problem solve the composition stuff if I'm looking for an idea. Um, and then of course I'll try it out, but then I'm like, "Oh, that was weird. I don't really like that." Um, where I have used it is doing deep research on particular artists and their composition styles, and their particular techniques.
Um, so that if I wanted to make something, like, in that style, I now have kind of a blueprint of, you know, how to think about that.
Craft Layer For Writing
Dave Dougherty: Um, with creative writing it's a lot easier, right? 'Cause, I mean, chatbots are built on text and whatever else. But, um, like any of the, the prompting for any of these things, it's been interesting to really take the creative process and chunk it out into its, like, individual parts, right?
So it's like, okay, I'm in kind of the discovery phase right here. I don't really know what I'm doing, I just need, like, all the ideas to narrow down to what this thing actually is, right? Then you have the editor role, you have these, you know, other things. Like, with, with writing, I've found really... what's really helpful is a, is what I'm calling, like, a craft layer, where I just had a conversation with Claude and said, "Hey, for all of my writing projects, if I tell you I want my name on it, you are not allowed to write a single thing.
You're allowed to be an editor, generate an idea for me, but because of the copyright standards, you're not allowed to do anything."
Ruthi Corcoran: Because I- That's a cool layer.
Dave Dougherty: Yeah. Well, and then on top of that, 'cause then I can say, "Hey, you know what? Like, I want this idea out into the world, but I don't really care to go through the nine months of writing something.
So go ahead and do it under a pen name. Write the hell out of it. We'll release, self-publish it as an AI thing, right?" Now, that one, I haven't done that one yet, but it's at least set up in the, the layer that way. But then I also went through and said, "Here's the list of my favorite movies. Here's the kinda list of my favorite books.
Here's my favorite authors, my favorite poets. Here are my favorite rappers and how they use words and wordplay, and, you know, yada, yada." And just, like, just an exploration of the craft of language, right? Um, you know, it's like, I like Hemingway, but I hate his over-masculinity. So, like, kill that part of it, but recognize the, you know, the iceberg structure that everybody talks about with Hemingway.
Um- I love Billy Collins and, um, Allen Ginsberg and, and all those guys. Um-
Ruthi Corcoran: Did you put all that in a NotebookLM, or is that, like, in a markdown file somewhere that it references from? No,
Dave Dougherty: that's in a-
Ruthi Corcoran: Or is that...
Dave Dougherty: Yeah, so I had this whole conversation about style and reference and, like, you know, dos and don'ts- Yeah
based on my voice and my preferences. Yeah. And then so that became a markdown file that goes into the instructions, and then just in case, because, because of the music producer background of mine, I don't trust file storage, so I keep it in, like, seven different places, you know? So I have a Word doc version of it.
I have it, you know, in Claude, so then it's a reference document every time it goes to set... you know, go through the stuff to say, "Okay, I'm doing poetry, therefore I should be doing this with his writing for, you know, for this particular thing." Then I have one for prose, which is, okay, we're working on, um, we're working on a creative writing more, you know, novel type situation.
So all right, great, blah, blah, blah, blah, blah. Um, and so I... within... with these layers and chunking out the working process into the kind of, like, five master prompts-
Ruthi Corcoran: Mm-hmm ...
Dave Dougherty: um, which was an idea I picked up off of, uh, a webinar I went to. I think Alex and I, we talked about this in a previous episode.
Character Cards And Outlines
Dave Dougherty: But I leveraged that plus this craft layer- To then take my creative writing capstone from university, which was eight chapters of a book.
Um, and I put the, put the first couple prompts in it, uploaded the thing, and I said, "Critique it. I know it's young writer, so, like, you can ignore that stuff, 'cause yeah, I know, I get it." But then based on what my... you know, what I prefer now and how I try to write now, let's do this, let's build out an outline, like, you know, it...
the way James Patterson does his outlines, which are these ridiculous 70, 80-page outlines on a legal pad. I said, let's write a scene-by-scene kinda hit list for the overall, you know, novel structure, and then it took the 50,000 words that I had for the first draft from, you know, 20 years ago, and then started putting all the sections and all the different things.
And so now I went and did a four-hour writing session last weekend, and I was able to rewrite the entire first chapter and most of the second chapter in the new style with these sorta like developed character cards too, to be like, "This person would do this, this person would do that. They wouldn't ever do this, so make sure that you look out for that."
Um, so it's been... That part of it's been very cool. It's a lot easier with writing than, than with music, at least in the way that I've, I've been using.
Ruthi Corcoran: And it occurs to me, like, those character cards, maybe you already do this, those could be, like, a marked-on file themselves. They are, yeah. So that y- you've got this constant reference back.
Yep. Which a- at least for me, I find that super helpful, both because that's one less thing I have to keep track of, but also- Mm-hmm ... it just keeps the, the, the consistency and accuracy of the responses I'm getting just a, a little bit more on track.
Dave Dougherty: Yeah, you'll appreciate this, Ruthi. I s- I specifically told it, "Build out RPG-style character cards-"
Ruthi Corcoran: Love
Dave Dougherty: it
"for all of these things." So it gave me, like, psychological attributes, it gave me physical attributes. It gave me, you know, um, all these different things. But then in terms of, like, troubleshooting, it was interesting 'cause I said, you know, like based on my favorite movies, my favorite books, my favorite whatever, like, I'm really struggling with what's at stake in this story, and it's really hard to get going on a new draft for this when I don't understand what, what the problem is.
And that's a really boring book if there's no problems.
And so ended up having this whole long conversation of like, okay, yeah, if you, you know, you mention Chekhov, and if you took that approach, then this. Or like Dostoevsky, it'll go that. Or David Foster Wallace, it'll go this way. And I'm like, okay, but I don't wanna be preachy like a lot of Russian literature is.
Also, I don't wanna be depressing like a lot of Russian literature is. But I do like the high-mindedness of it, so let's, you know, explore that. And, you know, um, so yeah, that craft layer has been really fun because then it's like you're actually talking to a peer-
Ruthi Corcoran: Yes ...
Dave Dougherty: who has the same kind of context and, and, um, you know, you could make a, a random reference to, you know, Vonnegut's unpublished short story from '52, I like this element of it.
And it's like, oh yeah, that's a good thing, and we can include it this way. And like, dang, okay. You know?
Ruthi Corcoran: Yes.
Dave Dougherty: It's like total nerding, and it's awesome. Um, but it is, to your point, like this is why there's an editing process at the end. But that is a very distinct part of the process, right? Uh,
Ruthi Corcoran: so- Distinct and critical.
That- Yeah ... that's so... That's such a cool applic- application, how you've taken your process and sort of figured out, okay, how do I apply this? Or how do I make this work for me? And now out of it, to your point, like you get, you get a better collaboration partner. And like on its own, off the shelf, you pay...
And this might be true of the free version, you pay 20 bucks for Claude. Like you already have- Mm-hmm ... a collaboration partner, and now you're just making it even more bespoke to work for your process.
Dave Dougherty: Mm-hmm. And I now have them set up in particular projects so that I can just- Yeah ... go in and, and, you know, add to these things, and it has the specific instructions to go through.
And, you know, it's like I could do co-work for some of this stuff, but that seems like-
Ruthi Corcoran: Yeah ...
Dave Dougherty: it's too much, you know?
Agents Cowork And Copilot
Dave Dougherty: Like I, I think- Yeah ... I don't know if you guys have gotten to this point yet, but like the capabilities that are there now, I don't necessarily have a use for all of them yet. You know? Like Claude released Design, and then they just released this new scientific thing.
It's like, okay, well, I'm not ever gonna touch that one. But like the Design one could be interesting. Um, but again, I don't have use for that right now, so we'll wait to play with that later. And I don't- Yeah ... I'm interested to play with Fable, but to your point, Alex, earlier, like I'm a little wary of the things that they had to do in order to release it.
You know? It's like how much of this is actually usable? Like I'm c- I'm okay with 48- Yeah ... Opus 48, like- Do I need to go putz around and learn a new system? Probably not for what I'm needing right now, you know?
Alex Pokorny: Yeah. And, uh, uh, there's a push right now, um, probably haven't seen it yet because it just came out literally this morning, but, uh, ChatGPT is basically pushing chat into basically ChatGPT Classic, it's gonna be called, and then basically trying to make their version of CoWork kind of a, a computer takeover method, basically, that kind of agent.
Ruthi Corcoran: Mm-hmm.
Alex Pokorny: Yeah. Um, as one of the main agents, and it can run for hours now instead of minutes. And a lot of people have, you know, tested in the past. It was pretty slow and wasn't that great. Um, but apparently some big improvements to it. So they're really pushing towards that, which, um, a lot of the online response seems to be, you know, that's a great way to use up a lot of tokens.
Yeah. Mm. Um, and chat is probably, uh, kind of a lost leader. So if they're trying to get more of that enterprise money, which-
Dave Dougherty: Right ...
Alex Pokorny: I mean, they're financially very, very consumer-based, unlike Claude, which is very enterprise-based.
Ruthi Corcoran: Mm-hmm.
Alex Pokorny: Um, they could be trying to push off some of that, that chat piece and try to get people more into the CoWork version.
And- Yeah. Mm ... Claude just had an update today that also pushes them more closely together, CoWork and chat. So I think-
Ruthi Corcoran: Which is showing up in Copilot, guys. I'm so excited. I asked for help rewriting an email, like just give me some formatting. And then yesterday, yesterday, this is hot off the presses.
Copilot- Put this little module in the chat where I could send the email directly from Copilot, and I was like, "Yes, push the button. This is so great."
Alex Pokorny: If your IT team has allowed it, it can also take over your screen or browser, um, same way that Claude can or-
Ruthi Corcoran: Yeah, there's a little toggle that's like you can have chat or you can have Cowork on, and I actually haven't toggled the Cowork.
This was just built into the chat functionality.
Alex Pokorny: Oh, that's cool.
Ruthi Corcoran: And so I gotta check out the Cowork. Maybe that'll be a topic for our next section of like how are we using Cowork? Do we like it? H- like, how does it compare across all these different things?
Burnout And Wrap Up
Alex Pokorny: I wonder the day that we will soon have two laptops assigned to everybody or virtual machines assigned to everybody, which is probably more likely, so that you can have your little bot do its thing, 'cause it takes over your screen and your mouse.
I mean, your mouse is moving around, so you can't really do anything except go to lunch.
Ruthi Corcoran: You can go get a coffee. You can change your laundry, 'cause we're all working from home.
Alex Pokorny: But productivity.
Ruthi Corcoran: Stays constant.
Dave Dougherty: See previous comments on focusing on the wrong thing.
Alex Pokorny: Also previous comments on the rise of burnout for some
Ruthi Corcoran: reason.
Alex Pokorny: I don't know why. It seems like tech workers are burning out really fast. I don't know why. We're just making them do seven things at once versus one.
Dave Dougherty: Yeah. It's like the Lemmings game where they just walk off the ledge into the lava- Yeah ... and they don't know why. They just keep doing it. Yeah. Yeah. Anyway, what a cheery way to end on
this last episode. Thank you for making it this far. As promised, this was a really interesting conversation. Um, thanks for sticking around and listening. Like, subscribe, share. Go check out Pathways. Lots of new stuff there. Um, yeah, thanks, and we'll see you in the next two weeks in the next episode of Enterprising Minds.
Take care.
Alex Pokorny: Cheers, all.
Ruthi Corcoran: Cheers.
