Listen
Elsewhere Online
Apple Podcasts | Spotify | Amazon Music | YouTube | iHeartRadio | RSS Feed
Today, we have a special guest, Robert Csizmar, also known as Joker, who is the CTO and co-founder of PublishDrive. We’re thrilled to discuss the innovative AI metadata tool that PublishDrive has recently launched.
Meet Joker: The Technical Mind Behind PublishDrive
Robert Csizmar, affectionately known as Joker, is a co-founder and the Chief Technology Officer at PublishDrive. With a background in databases and artificial intelligence, Joker has spent years working in various sectors including government, banks, and educational institutes. Though he doesn’t consider himself an author, his technical expertise has greatly influenced the publishing industry. Interestingly, Joker also has a pet project involving writing books, a fact that came as a surprise even to his colleagues!
The AI Metadata Tool: A Game-Changer for Authors
One of the most exciting developments at PublishDrive is their new AI metadata tool. This tool is designed to help authors and publishers create high-quality metadata, which is important for the success of any book. Metadata includes elements like titles, subtitles, descriptions, and keywords, all of which play a vital role in a book’s discoverability and sales performance.
Why Metadata Matters
Joker emphasized the importance of metadata in the marketing and organic reach of a book. Poorly crafted metadata can hinder a book’s success, no matter how good the content is. The AI metadata tool aims to save authors time and effort by providing high-quality recommendations for metadata based on the book’s content.
How It Works
The tool analyzes the content of the book and generates recommendations for titles, subtitles, series names, keywords, and blurbs. It uses a database of high-performing keywords and market intelligence to ensure that the recommendations are not only relevant but also optimized for current market trends. The tool allows authors to choose from multiple options, giving them control over the final metadata.
No Training Policy
One of the key concerns for authors is the security of their content. Joker assured that PublishDrive has a strict no-training policy, meaning the content uploaded to the metadata tool is not used for training any models. This ensures that authors’ works remain secure and private.
Practical Applications and Real-Time Data
The AI metadata tool is not just about creating metadata; it’s about creating effective metadata. The tool connects directly with Amazon to get real-time data on trending keywords and categories. This ensures that the recommendations are always up-to-date and relevant. Authors can also use the tool to experiment with different pen names and see how they perform in terms of keyword optimization.
The Power of AI in Everyday Tasks
Joker shared how AI has transformed his workflow, not just in publishing but in various other tasks. For instance, he uses AI to summarize legal documents, making complex information easier to understand. This capability to simplify and clarify information is one of the many ways AI can be a valuable tool for authors.
Tips for Effective AI Use
One of the essential tips Joker shared is the importance of crafting good prompts. The quality of the output from AI tools like ChatGPT depends heavily on the input. Authors should be explicit about what they want and provide as much context as possible. Additionally, understanding how AI interprets prompts can significantly improve the results.
Future Prospects and Feedback
PublishDrive is continuously looking to improve their tools and welcomes feedback from users. The AI metadata tool is still new, and user experiences and suggestions are invaluable for its development. Joker encouraged authors to try out the tool and share their feedback to help refine and enhance its capabilities.
The intersection of AI and authorship is an exciting frontier, and tools like PublishDrive’s AI metadata tool are leading the way. By simplifying complex tasks and providing high-quality recommendations, AI is empowering authors to focus more on their creative work. We encourage you to explore this innovative tool and see how it can benefit your publishing journey.
For more information and to try out the AI metadata tool, visit PublishDrive Assistant.
Full Transcript
[00:00:00] Welcome to Brave New Bookshelf, a podcast that explores the fascinating intersection of AI and authorship. Join hosts, Steph Pajonas and Danica Favorite, as they dive into thought provoking discussions, debunk myths, and highlight the transformative role of AI in the publishing industry.
Steph Pajonas: Hello, everyone. And welcome back to another episode of Brave New Bookshelf. I’m one of your co hosts, Steph Pajonas. I’m the CTO and COO of the Future Fiction Academy, where we teach authors how to use AI in any part of their business. We’re changing things up a little bit this morning. We usually do most of our podcast interviews in the afternoon, my time.
And this morning, I got everything up and ready to go because my lovely co host is in Europe. And she’s with us today in the offices at Publish Drive. So I’m going to hand it over to Danica Favorite.
How are you today, Danica?
Danica Favorite: Thank you. It’s very fun to be here and recording at Publish Drive in our office. And uh, fun to actually have my guests right [00:01:00] next to me too. So of you who don’t know me, I’m Danica Favorite. I am the Community Manager at Publish Drive. And PublishDrive is a company that helps authors publishers distribute their books to a worldwide audience.
What’s really cool is some of you have heard me teasing about our new AI metadata tool, which so fun. So today we’re doing something a little differently. We’re talking about the AI metadata tool and I have right here, this is Robert Csizmar, otherwise known as Joker. To tell us about it uh, is our CTO and the co founder of PublishDrive.
And has spent many hours working on the development of this tool. And really excited for him to tell us about that.
But before I do that, I have to tell you something hilarious because was playing on ChatGPT today, and uh, want to share this. I hope everyone does this as an experiment because it’s so fun.
So I asked ChatGPT, I said, who is Danica’s Favorite? And uh, GPT said, Danica Favorite is an author known for her inspirational romance [00:02:00] novels, featuring themes of faith, hope, and love. Her books typically focus on historical and contemporary settings, offering heartwarming stories that explore the complexities of relationships and personal growth.
Danica Favorite’s works are well received by readers who enjoy clean, uplifting romance. a strong sense of community and moral values. also known for her engagement with readers through social media and writing workshops, she shares insights and advice on writing and storytelling. So then, of course, I asked it, who is Steph Pajonas?
Pajonas is an author known for her diverse range of fiction, including contemporary romance, fiction, and cozy mysteries. She’s particularly recognized for her Nogiku Series, I hope I said that right, which blends Japanese culture and dystopian science fiction. Pajonas writing often features strong female protagonists, intricate world building, and elements of Japanese traditions and language.
In addition to her novels, she is active in the writing community, sharing her experience and advice on writing, self publishing, and [00:03:00] marketing through her blog and social media channels. Pajonas work appeals to readers who enjoy genre crossing stories with rich cultural elements and dynamic characters.
So I think Chat GPT knows us pretty well.
Steph Pajonas: Yeah, I think it’s got us nailed. It’s good.
Danica Favorite: Yeah. Yeah. Right. I was just asking Chat GPT.
So, um, did not ask it about Joker. I should have actually, now that I think about it, but I’m going to go ahead and introduce Joker. Like I said uh, Csizmar is his name, but we all call him Joker. is the co founder of Publish Drive and also the CTO of Publish Drive.
So Joker, why don’t you tell us a little bit about yourself, what you do, and we really want to hear about the metadata tool.
Robert Csizmar (Joker): Okay. No problem. hello everybody. I’m Robert. Nickname is Joker. I’m the one of the three Co-founder of Publish Drive and I’m the, CTO of the technology officer of the company, which basically means two roles here.
First of all, I lead the product team creating made any technicals within the [00:04:00] company. So if you use our platform is the thing that we’re constantly working on. Important role here to research the new things that we will talk about later, probably. If any new product is coming, the research part definitely responsibility here the company. The metadata tool and I also right now also in the research part of next thing. So, few months, probably we can talk that also.
I’ve worked a database intelligence expert, worked for government, banks also for educational institutes.
So mindset is basically about databases, about artificial intelligence, technology part. basically I’m not an author. I’m a technical, coming from a technical work.
However, the recent 10 years, I have a lot of experience also in this publishing world. talking about, I also have my pet project writing books.
Danica Favorite: Wait, wait, stop there. Your pet project writing books? Yes. Are you writing a book and you didn’t tell me about this?
Robert Csizmar (Joker): Well, everybody has secrets.
Danica Favorite: Yeah, we just got done with lunch and I didn’t hear this. So now I’m going to [00:05:00] be with him the rest of the week. So I’m going to bug him about it. All right, please continue.
Robert Csizmar (Joker): Okay. About the metadata crafting tool. It’s important to highlight that as Danica introduced ourself this time, probably, you know, the aggregation service that we have. However we try to, Put together services not just in for distribution, but other aspects other stages of the journey .And this metadata creator is a very very good example for that because this one is not chained to the distribution one.
So metadata product is also for those customers who don’t use our distribution product. So it’s much, much wider audience for that.
So very, very officially, we trying to cover all the aspects the stages of the journey also with education session, just as you have and also the strategic partnerships, which can be now tracking the distribution platform but also in other services we currently have or will have in the near future.
So let’s talk about the very briefly about the [00:06:00] metadata tool We definitely see that being an author is a very hard stuff. It has a lot of things to do beside the writing itself. However, getting enough income from that needs a lot of other steps that need to be done and this is how the metadata creator came to the picture.
As we’ve seen this metadata and a basically part of the marketing and the authorship is creating the content and the marketing for that.
Basically, these are two steps for that. And the metadata is a very crucial part as we see of the marketing as the organic reach. at Amazon or any other storefront that you use to sell your books is the very basic thing that defines the success is the metadata itself.
And we see a lot of books uploaded to our platform, and it’s very, very common issue. Of course, they know what should be the title, even the subtitle, however, we also see a lot of books without subtitles. But when we arrive to the description it, basically it’s purely done. I totally understand [00:07:00] and get it that the main thing, the biggest why you hear the content itself. I totally admit that. However, that also still a product that needs to be sold. And in that part, the metadata pays a very, very important role.
However, a lot of other marketing activities needed for that. But the metadata is the basis, as this is the link between the content and the customers, the buyers, and the storefronts. If it’s done poorly, it can even ruin all other activities that may need it for a successful book launch and book selling.
What we want to, first of all, save up a lot of time for authors and publishers to focus on that part, which are the most joyful and what you authors can do best, creating the content itself. Save up time for that and make the metadata crafting on quality that you don’t need to invest time and money, effort For that just choose from very final options.
A very important aspect of this tool that it creates recommendations for metadata and you are the one who [00:08:00] finally choose which one you want to use and how you want to use is totally up to you.
And the most important point here is that purely works on the content of the book. So you don’t need to know the categories of Amazon’s. You don’t need to know how keywords working. What are the best practices on the market for crafting a description.
All of these things are developed in the tool itself and using AI is working in your content. So it’s not templates, not predefined stuff. It’s everything working particularly on your content. It reads your book from the beginning to the end and craft this kind of data that needed at KDP or any other publishing platform.
You get recommendation even for the titles for the subtitles for the series on why it’s important because this kind of metadata also very important search terms as well. So this is why we give recommendation also for the titles also for the subtitle, series name. Even we provide pen names.
[00:09:00] About the blurb, it’s very important. I did this research for that and check our books that uploaded to our platform or checking your Amazon pages, I’ve seen that authors make a lot of mistakes when creating the blurbs because they know too much the content. The content is your baby basically, and you cannot tell it as an outsider, you’re totally in that.
However, the blurb is a sales pitch. It’s not just purely about the content. It’s not just the plot, but it’s a sales pitch. It has to sell as a product, as a simple product. And this is how we made this process to generate all the parts that catchy, that sexy and has a high probability to actually sell your book.
Danica Favorite: Yes. So one of the things I wanted to say, cause I think this is fun. I would have never thought, Oh, I should. See what a good pen name for this would be for keywords like that is funny. Like I said, I mean, most of us aren’t going to use that, but I think that’s fun to think about.
One of the things I want to point out and just share with you I know our [00:10:00] authors very concerned about this.
When you upload your book to our metadata tool to create the metadata,do we use that in any other way to train data or anything like that?
Robert Csizmar (Joker): No, no, it’s very, very important. Very important part of the concept that we have a very, very hard, No Training Policy. So the content that is uploaded , it’s solely used only for creating those metadata that you request for. It’s not used in any other purpose And how we can do that, it’s very simple. You know this sentence if you do not pay for a product that you are the product. . Facebook knows that the best as they’re doing a very big business from that.
However, this product is a payable product. So whenever you upload a book, you have to pay for that to process that and get those recommendations.
And as we have our money on that service fee, basically we don’t use the content for any other purpose, just to fulfill your request and nothing else.
Steph Pajonas: Yeah, I think that’s important. A lot of authors want to know that their books are not being used for training, so I’m glad you pointed that out.
Robert Csizmar (Joker): It’s [00:11:00] a very important note. And, it’s a very, very basic right that we totally respect and we don’t use any other purpose.
However also wanted to highlight, I said, this is a payable product and will be. Right now, as we in the introduction part of the product who starts using this one, you receive a trial, a bonus of credits that you can spend within the product without any obligation.
You just start to use and you have enough credits. So if you try it out, you can do it without any further application.
Go there try it out.
Steph Pajonas: So you’re giving away some credits during a beta testing process, basically?
Robert Csizmar (Joker): I wouldn’t say it’s a beta testing process It’s much more to give you a chance to try it out without paying anything. So we’re trying to gain visibility for that. I can be very honest about that.
However, it’s part of the introduction campaign. We do not have a deadline for that when we draw that, but it depends on the success of the product. So after a while, sooner or later, it will be redrawn.
Steph Pajonas: [00:12:00] Excellent. Excellent. Danica, have you tried this yet?
Danica Favorite: Yes. So I’ve had a lot of fun with it. I got to play with some of the beta versions as he was building it and testing it. So it is a lot of fun. Like I said, I’m not going to use the author pen name thing, although now I’m thinking about it. Now that I’ve got some ideas of future books I might write, like it might be fun to test that out and do a pen name. I actually I’m doing an experiment haven’t looked yet to see the results, but I have a series of books. I only did it for the one book, but in this series of books, the blurb stinks and I have revised it and revised it and revised it.
Originally when I did the first blurb, I paid someone to write it for me. And it was not great. And so I’ve been tweaking it ever since. So I actually revamped the blurb based on what our metadata collector told us. I’m eager to see, like, I’m letting it run for a little while under the new blurb to see what that description change does for my sales.
what I thought was really interesting with that is. [00:13:00] It’s still very much like you and I talk about with the Future Fiction Academy, what type of AI user are you because you put your book in, it analyzes your story. And then it’s going to give you a couple of different options. What I did, it’s the same thing, you know, I’m still a weaver according to the FFA terminology.
And so I took bits and pieces of each blurb that it gave me, and I combined it into what I liked the best. And so it was still that hands on author experience of, okay, this works for me. This doesn’t work for me. Yeah, no. And really making it my own. But the AI did the heavy lifting of saying, here are the really important points of your book. This is what’s selling. This is what’s important.
And that’s the other thing I wanted Joker to talk about is that where are we getting this data? How are we hooked up to get this A. I. Data to analyze the books?
Robert Csizmar (Joker): Yes, this tool is far more just prompting GPT with different prompts and the content.
As I described before, we built up and maintaining a database of [00:14:00] keywords we know how they performing, which are better, which are worse. So we definitely have a complete list of that. And when you upload the content to us, what we do we find the most fitting by content and best ranking keywords to your book. And this is what we give you as the options for keywords.
You are in charge, you are the boss. So you decide finally which one you want to use. However, from technical point of view, these are the best keyword that fits most for your content and also very good performing on the market.
We also do one more, which is the market intelligence, because right now we connect the content to the market knowledge, then our market knowledge in that sense, for example, the keywords, but also market knowledge are the categories as well.
We recommend categories for Amazon and also for BISAC one. The Amazon is using KDP the BISAC used with Wide stores. Amazon has more than 8, 000 different categories, BISAC has [00:15:00] 5, 000. So it can be tough to find the best one. However, if you find the most fitting, the most niche category, then you have the better chance to find your readers and find people who are interested with your best audience.
Danica expressed that the hardest part, the 80, 90 percent of the work is done by the AI. My database is an AI and however, in the end, you are the one who choose .
Danica Favorite: And what I like, from talking to you during the development process is that some of this is directly talking to Amazon. The AI is looking right at Amazon. So it’s more of a real time other than, you know, googling or whatever to find what the categories are . He’s got it hooked up to talk with Amazon. I mean, not have a conversation with Amazon.
Robert Csizmar (Joker): Yes, yes. But it’s something like that. It has the most trending data about the keywords. So, different trends coming and going. The keywords that are working one year ago are not the best currently and won’t in one year. And the [00:16:00] data that is working is constantly updated and all the most up to date keywords are there. So it means if you run for the same content the assistant with even a month or a year difference, probably you will get different keywords because every time the most trending, the best keywords will be provided for your content. It saves a lot of time here also with the research. A lot of knowledge lot of technical tools, you can do some similar. However, it’s take a lot of time to do it manually with our assistant is done in seconds.
Steph Pajonas: I like that. I want efficient help .We just want to get back to writing, right? We just
Robert Csizmar (Joker): yeah. This is the concept here to save as much as time as possible to give you back. And you can use it for writing for family, whatever, you know, what is the most valuable part of you? If we can do it in high quality and save up time for you, this is good for you and also good for us.
Danica Favorite: Yeah. I thought it was really cool and that’s why I did that experiment with this blurb. Like I said, I’ve left it alone. [00:17:00] I haven’t looked at it. I’m resisting the temptation to check every day to see what the sales on this book are.
But, the blurb was written years ago. And so how is that going to impact and improve sales? We’ll, we’ll see how it goes, but yeah, I thought this was a perfect time to bring someone in to chat about that.
Now we’ll go into some of our questions that we usually ask unless there’s anything else you wanted to share about the metadata tool.
Robert Csizmar (Joker): If you have any feedback on that. It’s quite new. So we released it nearly two weeks ago. If you have any feedback on that, it’s more than welcome. Please share with us your experience.
Danica Favorite: Obviously, you’ve developed an AI tool. You’ve got experience with AI. I don’t think people know this, but PublishDrive has had some behind the scenes AI for quite some time. So this isn’t out of the blue.
Robert Csizmar (Joker): Yes. It’s not the first time we’re playing with AI. Even years ago, still much, much before the whole ChatGPT revolution that happened one and a half year ago. Also had an assistance in the [00:18:00] distributions book upload phase, when a pre trained model also suggested categories for the uploaded book. It was awesome tool, however, it It was quite slow. It work for only English content. So if you have even not English title, but French, German, Chinese, whatever, it will also take care of that, which is important and if you use our platform, there is a chat support opportunity there where you can ask your questions and it’s answered immediately. That’s also an AI engine working behind that. An AI can give you a good response and move you on.
Danica Favorite: I think it’s cool because again, I know you’ve been in AI a lot and we always ask our guests, how are you approaching AI and publishing? So, love to hear, especially now that I know he’s writing a book, seriously, we’re going to chat about this.
But, what is your approach to AI and publishing?
Robert Csizmar (Joker): Well, I’m writing a book that was actually came from that curiosity the AI came to the [00:19:00] picture is it really useful for creating content. I was playing with chat GPT, also other dedicated services.
And I just tried out. I had plots in my mind for years, but now it was the chance to actually make content of that. And I haven’t finished any of them, but I could use that to basically turn my ideas into four or five chapters I could create basically by ChatGPT and other tools.
So this is what I use for publishing. I would highlight here, that AI is not an alternative for Google. The measurement of the AI is not how much he directly knows without any context.
So, every time I use the AI, I always describe where I am. What information I have, where I want to arrive, what output I expect and what format I expect.
For example, what helped a lot that I do a lot of research and also with coding, it speeds up. Well, I have an [00:20:00] idea that I want to try out. It can be work or not. Previously, it took days to create such code that be a proof of concept. Is it working or not? Now it takes a lot of time. 10 minutes to create a prompt and in two, three rounds, fix it until it actually runs and creates example that is it working or not?
Steph Pajonas: I love the fact that ChatGPT can really take a huge task and condense it down into something much simpler and easier, but you really do need to know how to prompt, right? This is one of those things where if you don’t know what you’re asking for, if you don’t know how to ask for it, you definitely need to get those skills rolling so that you can use these tools more effectively.
And like you said, it’s not, it’s not the best at search.
Robert Csizmar (Joker): No, definitely, definitely. Instead, I actually said that do not use external sources. Use my data that I provide you.
Danica Favorite: Yeah, I like that. And that’s a really good tip. I think sometimes the frustration with writers is we don’t give it enough data or we don’t explicitly say use only what I provide and then [00:21:00] suddenly it goes off the rails. You know, I know you’re sitting here saying, I don’t know how applicable this is for writing, but I think it is. I think a lot of what you’re saying here is stuff writers really need to think about when they’re crafting their prompts and what they’re asking the AI to do, because it’s really only as smart as you are telling it to be.
So I think that’s super cool. And it also goes back to what you and I talk about a lot with our guests is. Enabling people who haven’t been writing or been able to write before to write. Like you were saying, you have all these ideas in your head and now you’re finally able to put it on page. And I think that’s, what’s really, really important as well.
Robert Csizmar (Joker): Yeah. And the other stuff I commonly use and it’s part of a business but really don’t like to deal with contracts, legal documents, so read through legal documents. 10 15 pages. I also use GPT for that. Summarize me, highlight which part is important from this point of view or that or that. And that speeds up, it saves me hours a day. [00:22:00] It’s, saves a lot of time.
Steph Pajonas: I like for it to tell me things in plain language. I like the Explain it to me like I’m five right? , like I’m five years old. Explain it to me like I’m five years old because I don’t necessarily understand like all of this legal mumbo jumbo, which is what we call it.
So I like the fact that AI gives you a chance to learn and understand things that maybe are just written in ways that are meant to be for the legal profession or maybe even the medical profession that we don’t necessarily use in our everyday lives.
Robert Csizmar (Joker): And also the opposite direction. If I know what I want to express, but I cannot make in a form that is good for a legal document. I also use that purpose that this is what I want to tell and for me to be eligible for a contract.
Steph Pajonas: Yeah. It’s a translator.
Danica Favorite: Yeah. Yeah. Exactly. So the next thing we’ve talked a little bit about this, but what do you use AI in your [00:23:00] workflow for? Like we’ve talked about some different suggestions of how you use AI, but do you have any particular workflows that you find AI helps you with? Or what does that look like?
Robert Csizmar (Joker): mean in general in my work?
Danica Favorite: Yeah, just in general in your work, or if you’ve got like a writing tip you want to give. You can, I know you’re not like our traditional author guests, so you can play with this question and do with it a little bit what you think would be beneficial to authors.
Robert Csizmar (Joker): Well, I definitely use for, that when I have an idea, even it’s a small plot of, a possible novel that I definitely use it to enrich that. So if I have a very small list what I want and ask it to enrich, add more suggestion for that, describe that. What is important that I use it not once, but multiple times. So if I get something ChatGPT As a result, I review that, replace it with what I want, and I ask it again, okay, now review that, find the holes there, and do it a lot of times, three or four times, until I have the response for that.[00:24:00]
However I very amateur level, but I play the guitar and I totally learned it myself. And ChatGPT and AI is quite similar. You cannot tell it explicitly how to use that. You need to learn by experience how it actually works and every, hour you spend on that playing with that, even with real or not real tasks, the more you understand the details which cannot be explained you can learn that only by experience.
And, not big thing, but here in Publish Drive, we have dedicated channels we send jokes and memes and now I can create images with jokes, by prompts and sending to my team , and we laughing on that. I want to create the joke myself, it would do an hour at least, which I definitely won’t invest for that. But if it’s just a prompt, it can be done in 10 seconds that at the same time as I Google it.
Danica Favorite: I think there’s just so many fun things to do. So I know we’ve talked a lot and you may have already [00:25:00] technically answered this. What is your favorite AI tool right now?
Robert Csizmar (Joker): Well, basically chat GPT and also the OpenAI API which has more functionality, but it’s more technical how you can use that. However, it gives more opportunity to utilize yourself, but it’s more technical level.
If you know that, that opens more dimensions for utilizing AI. But once more, that’s much more the technical part of that.
Steph Pajonas: Yeah, we at the Future Fiction Academy, we teach all of our students how to use the playground so that they get access to all the hyper parameters and they can change the temperature and the top P and all of those things. And we teach them what those are and how they work because when you’re writing, you don’t necessarily want the voice that Chat GPT is going to be infusing into your work.
You want to be able to say, use new words or use the same words that I’ve been using over and over again, and then follow the prompt or feel free to be a little bit more creative and go a little bit [00:26:00] further afield.
Robert Csizmar (Joker): Yes. The playground. Yes. You’re totally right. The playground on its own is already a big step from the original GPT. However, you probably know that it’s working on the other pricing structure. As for Playground, you need to actually pay the API fee for every request you submit.
Steph Pajonas: Yeah. My little secret that I tell authors is that you actually will probably pay less by using the API on the back end. I canceled my chat GPT $20 a month and my Claude $20 a month because by using the API, I’m actually, I’m paying so much less, I’m saving so much money.
Robert Csizmar (Joker): yes, and, I think it’s important detail that in Playground you can switch between the different engines. And if you need a simple summary for, a text, for example, you can use the much, much cheaper 3.5 and for summarization, it gives wonderful results for that. You don’t need to use the higher ones.
For more sophisticated problems that the [00:27:00] most up to date models are more useful. However, my personal opinion is the newest model, the 4. 0, is worse than the previous simple four. I don’t like that. It starts too much for me and didn’t give that the same quality of the results, especially for more sophisticated. This is my opinion.
Steph Pajonas: Oh, those are fighting words.
Robert Csizmar (Joker): but, but, but in my experience, when also when we were building the assistant at the quite very late stage of the development OpenAI announced that Four Oh, and we tried for that. And for many tasks, the four produced worse results than the Four, simple Four. Probably it’s also a question of the prompts. If we invest a lot of effort to restructure the prompt for that, maybe in that case that works the same level or even better. But using the very same prompt, don’t know how you are. I cannot scroll LinkedIn or Facebook without getting a lot of prompt collections that want to, sold to me.
From that point of view, that can be [00:28:00] interesting that the same prompt that worked for 3. 5 or 4, maybe they don’t work the very same efficient. But again, it’s my personal opinion
Steph Pajonas: I understand. I had to change a lot of our prompts. I use a lot of prompts when I’m editing this podcast, you know, grab the transcript, give me summaries find, parts of the podcast that I should bring into a blog post all that kind of stuff and I switched over to 4 Oh at some point and I had to redo all of my prompts in order to get it to work correctly.
And I, so I agree. They just seem to be too different. Almost like two different animals, and you really have to adjust between them. I agree.
Robert Csizmar (Joker): Yeah, I found out the difference between 4 and 4. 0 is that when was using the simple chat GPT 4 the results, if we asked to create different versions for the same input, it was actually different. And really the response was dynamic. However, the 4. 0, even if we ask that create different version, [00:29:00] they will mostly the same. So this dynamics somehow missing from the 4. oh.
Danica Favorite: We don’t always talk about that. And that’s why we have such a variety of guests is that we want different perspectives. We want people to share different experiences. And explain things which I really think is good because I hadn’t thought about that actually until you just said that, that it makes perfect sense that of course, when the model changes, You’re going to have to change your prompts and it, I don’t know why it just didn’t click with me until now.
I was like, man, these prompts aren’t working as well. And uh, know I’ve discussed my love affair with Gemini and I have said that we sort of are on a break right now because Gemini isn’t working as well for me. And now I’m like, oh, I wonder because Gemini also went through a little bit of a change. And I didn’t change my prompts and I’m like, Oh, duh.
So I think it’s always good for us to learn different perspectives and points of view from people with different levels of [00:30:00] experience, because you’ve worked with AI for so long, even before it was a buzzword. And the stuff that you’re talking about and sharing. Is that meat deep down inside that we don’t always get to hear about. So I appreciate that. And thank you.
Robert Csizmar (Joker): Yeah. And one more tip that may be useful that when ChatGPT interprets the prompt you provide them, there is a U, pattern for that, which means it gets the beginning and the end of the prompt in much higher scale.
So it’s focused more what you express in the beginning and the end, not it’s a middle. When you just switch the order of the data I provided and put the very explicit comment in the very end, it radically changed the output.
Steph Pajonas: That’s something that humans have problems with too. We’ll get a list of items and you’ll generally remember the first thing and the last thing, and you don’t necessarily remember what’s inside. And my tip to authors for that is if you have like a mystery or something like that, and you need to hide a piece of evidence.
You hide it right in the middle because [00:31:00] the people don’t remember it. They remember the first things and the last things. And I guess the same is true for AI.
Robert Csizmar (Joker): Yeah.
Danica Favorite: Yeah. And what’s funny is that when he was talking about that, I was thinking how many of you have heard of Margie Lawson or taken any of her classes, but she used to talk a lot about front loading or back loading. stuff in your writing. Again, to make it more powerful is putting it at that beginning and end so that that’s where the more powerful words are.
And so it’s really interesting that all of these concepts, they all work together. And I think that it’s really cool that we can apply the same important lesson that we do for everything. It isn’t like you have to learn a completely new skill set, as long as you remember how these skills work and why.
Robert Csizmar (Joker): I describe it as like learning an instrument or a new language that you have to do it constantly. The only way to use it well is to learn.
Danica Favorite: Yeah, I think that is a perfect place to end it is that idea of the only way you’re going to get good at things like AI, chat [00:32:00] GPT, whatever program you’re using is to use it often because practice is what is going to make you better at it. So, for those who are saying, I tried that. That didn’t work for me.
Well. You tried it or you’ve been practicing at it because that makes a huge difference.
Steph Pajonas: Practice makes perfect. Excellent. So thank you so much for joining us, Joker. I’m glad that we got a chance to learn about the AI metadata tool, which I’m hoping people will go check out.
Do we have a specific URL to send people to, Danica? Yeah.
Danica Favorite: Now it’s just on the main PublishDrive website. Working on some changes to the site itself, but as long as you go to PublishDrive. com.
Robert Csizmar (Joker): PublishDrive. com slash assistant. You will be redirected to the landing page of that .
Steph Pajonas: Excellent. I’ll make sure that they’re in the notes. So, anybody who wants to come check it out should definitely come to a brave new bookshelf. com. Check out the show notes for this episode. We’ll make sure that they have all the information about what Joker was talking about and the link to the tool so they can go check it out.
Thank you guys for joining [00:33:00] us. I’m hoping you have a great time on your trip through Europe, Danica.
Danica Favorite: I know. I’m excited. This will be good. know you and I will get a little bit of a break from recording for a couple of weeks. still will have episodes going out because we’ve recorded a bunch ahead of time. So thank you to our guests who accommodated that for us and we can have some time off to enjoy.
And uh, guys won’t get time off, but we do.
Robert Csizmar (Joker): Thank you very much for inviting me.
Steph Pajonas: Thank you so much. We’ll see everybody soon. Bye!
Danica Favorite: All right. Bye!
Thanks for joining us on the Brave New Bookshelf. Be sure to like and subscribe to us on YouTube and your favorite podcast app. You can also visit us at BraveNewBookshelf. com, sign up for our newsletter, and get all the show notes.