Blake Rayfield: the things that I'm asking of them today were unimaginable five, six, seven years ago, right? Because for a lot of students, they're able to go from intro material to really advanced, really cool things in a in a little time, right? And In Python, think about how lots of us learn Python, right? In particular, we learn it because we want to build something, right? Whether or not that's an app or even some kind of data science tool or something like that. I mean, LLMs help you build something. That's undisputed.
Jess: Welcome to the Tech PubCast, a behind the scenes look at what it really takes to write and publish a technical book. I'm your host, Jess Haberman. I'm a literary agent, publishing consultant, and former acquisitions editor with nearly twenty years of experience publishing books. Today I'm thrilled to welcome Blake Rayfield to the podcast. Hello, Blake. Do you want to share a brief introduction?
Blake Rayfield: Thank you, Jess. My name is Blake Rayfield. I'm currently an assistant professor of finance and fintech at the University of North Florida. And of course, the most recent author of Applied Math with Python, solving real-world problems with Python. So thank you.
Jess: Co yeah, of course. Welcome. Congrats on your new book, published by Wiley. Blake and I got to work together on this project, which was really fun. you know, there's just like a good foundation in math that I think a lot of folks who kind of make their way into kind of data science and AI and machine learning, adjacent work, realize that they they they need to have. so I th I think that that would stuff
Blake Rayfield: Thank you.
Jess: you know, it's a just a great area to kind of dive in a little bit further. It's a good kind of amount to cover in a book. you know, how was the process for you? This was your, you know, your first book with a major publisher and you know, what did you learn?
Blake Rayfield: Wow, it was eye-opening. So as I said in my intro, I'm I'm an academic, so typically I'm used to publishing. I have lots of published papers. I've written book chapters for academic type books or some kind of papers that are mixing between academia and business. but this was just a totally different ballgame, right? So for me, I had to learn a lot along the way, and we can go into all those details, but I learned a lot along the way and it was just totally different than anything I'd done before.
Jess: Yeah. I mean,
Blake Rayfield: Yeah.
Jess: it was it did you feel like the writing like process was different or kind of like the expectations in terms of like how the like publishing and editing like that process goes?
Blake Rayfield: little bit of both. typically if we go into the academic like writing process, right? So I write some paper, maybe it's quantitative in nature, right? It has some results in there. mostly the audience I'm writing for is other academics, right? And so when they go read the paper, I can use all the terms and phrases that they're gonna understand. and so when I send it to a journal, right? for to hopefully to publish, I know that that's gonna be the audience that reads it and gives me feedback. Okay, and so if you can imagine that process, it's it gets to kind of be a little bit of a game, like repetitive in the sense, right? Like I I've gotten really good at that process over the years. So when I went to go write this book, which was of course not at all for that audience, right, there's just a totally new level of expectation, right? I had to make sure that I was connecting with the potential audience that may read my book. I had to make sure that I had all the, you know, of course, when you write a book for a general population or at least people who are interested in tech books, there's like a whole different expectation on making it nice to read, right? Look a lot nicer, right? In academia, we really care how it's formatted so much, right? So we had to make sure that this formatting's right, the sections are right. I mean it was just like all of that was just a great learning process though.
Jess: Yeah, that's great. There's the sort of vernacular, like there's the vocabulary that you that you should and shouldn't be using. Like you're trying to simplify things because right, you're speaking to like a more general audience who, may or may not understand some of the more technical concepts and jargon. and And it does have to look different on the page, you know, 'cause you are focused on like a print you know, print books are still selling, are still very popular. I do that too. I did a lot of like sports and nonfiction and history and outdoor recreation books like early in my career and
Blake Rayfield: Mm-hmm.
Jess: And we had like s different like trim sizes of books. So for example, like some that were really small. So like we were really focused on like where line breaks happened and like you would mark up like we're what were what are called widows and orphans, which like I think even in fiction a lot of fiction publishers don't even worry about anymore. but it was like it was like so specific to like have it look a certain way on the page or fit or like really specific trim size. So I you know, I don't think Wiley is gonna is really going too far in the weeds there, but but yeah it's totally different from academic publishing and it's interesting because you're you are getting like the kind of the review period. Maybe talk a little bit about like technical review, you know, how did that work in your process? Like what did you learn out of the feedback that you're getting from from your technical reviewers.
Blake Rayfield: Sure, so of course like the techno reviewers, the ones who are looking over the code or the math for the book, it it was a different process as well, right? So someone's constantly checking even small snippets of code that I'm putting in there, to large snippets of code. And I think for me, some of the really quick feedback that I got was making sure I'm connecting everything. So again, as someone who writes it either for academia or maybe you you're even used to writing code. We know comments are a thing, but even with comments, it doesn't leave enough breadcrumbs for a reader to be able to connect all the sections of code throughout a chapter, or even throughout the entire book, right? And so a lot of times in in chapters, I would write snippets, small little snippets of code, but maybe that snippet was dependent upon the previous snippet or a section. And that section may have been dependent on something I loaded in a previous section. So that was some of the initial feedback. that I received. Otherwise the code I mean if you know how to write code, the code is that. so I think that was the that was the biggest thing.
Jess: for folks who are starting, like their writing or working with a publisher is like what are the expectations for your code? Is it expected to be like, you know, you're div you're delivering a full like Jupiter notebook of code that y can run alongside the book, maybe as like a a downloadable for the reader? Or is it like, yeah, you're illustrating certain concepts with snippets that don't necessarily have to connect to each other. But I think it is really important to think about like how your code like builds or does it build on on prior snippets, things like that.
Blake Rayfield: to add to that, I mean you mentioned Jupyter Notebook, right? So that's how a lot of people are used to writing these this type of code, right? So I mean, of course you have the app type, which is not in a Jupyter notebook, but for math and science and you know statistics, we're used to Jupyter notebooks. And so with Jupyter notebooks, everything runs sequentially and it's all right there, right? So you don't really need to connect it. But and and honestly, when I was writing the chapters, that's how I wrote it. It was in Jupyter Notebooks, and so for me it all made sense. But for the Readers, you're seeing it presented as snippet of code with excl you know, exclamations and and all sorts of other details in there. So it it doesn't come like that free-flowing.
Jess: Yeah. Yeah, it's interesting the way that the format of Jupiter notebooks have helped us to think about like coding differently and help us to think about writing books differently. Like when I was at O'Reilly, we did have a lot of of authors submit books via Jupyter notebook or at least use that kind of style, that linear style of like everything sort of building on in that you can mix sort of, you know, markup and and code. Jupyter notebooks have not been around forever, you know. So it's it's feels like a kind of interesting format, you know, that really aligned itself really well with books. but yeah, it sounds like you found that format pretty useful for for your You're drafting.
Blake Rayfield: Yeah, absolutely. And I mean to put in a plug for the book too, I put my Jupyter notebooks on GitHub, so they're available. I mean, as you read the chapters, you could just go reference those. And of course the code is run sequentially there. So even if you missed the footnote that says you needed to run a previous section, it's all there.
Jess: Yeah, exactly. I guess in terms of like expectations when you're going into working with with a publisher on this kind of book, like what were some of the like the biggest misconceptions that you know you sort of were were clarified for you during this process?
Blake Rayfield: I think having edited a technical book before I wrote the technical book. You know, you learn pretty quickly that editors all have different expectations of what that technical book will both look like and also what they expect, for example, the chapter links to be or the sections to look like, or it you know, it's really from person to person. So I think that was one of the biggest misconceptions that I was going in thinking it was gonna be like, the other technical book that I had worked on, or maybe one that I had read. whereas again, it's kind of a dance between you and your other editors on on the team to see like what is gonna work for you and them and everybody. So I think it's really you know collective work. So I'm I think that was one of the biggest misconceptions.
Jess: Yeah. I think to your earlier point about reviewing content in sections, like that's another real benefit of having a technical reviewer who is with you for the whole book. because like they can help you see like where you need to kind of make connections across different chapters and and code snippets and things like that. They see kind of like the long game. so that's something to keep in mind when you're like often a publisher. It will ask authors for recommendations on who they think would be a good a good tech reviewer. So someone who, you know, has the has the time and attention to support you throughout the whole process is is great. Doesn't have to work that way, but but it does I think really benefit the sort of the long tail of writing, which like it is like a long process.
Blake Rayfield: and absolutely. I can say, I mean, your team is the one kind of that covers for you too. So I mean as you start to get in the weeds deep in the book, they may have shortened deadlines, you may have shortened deadlines. So it's it's gotta be a great relationship. And luckily I had a great team. I loved everybody that was there. So but yeah, it's it's definitely important to have a good team.
Jess: Yeah, yeah, absolutely. Was there anything that you found like to be harder than expected? Like what was hardest for you while while you're working on the book?
Blake Rayfield: Just writing enough. I mean, I'll go back to my academic roots. So going back to academic roots, we kind of get to the point, right? So meaning like when when there's a point we're trying to make, it's it's a pretty short and condensed, like here's the point, right? But here I had to I think one of the harder things for me, and this is one of those like learn bad behaviors, is that I had to make sure that I kind of gave it enough detail, and kind of
Jess: Yeah.
Blake Rayfield: talked about it in more than just here's the point. So that was one of the most difficult things for me to get over. But that's not gonna be everybody.
Jess: Yeah, right. a lot of folks have the opposite problem. They write too much and then you you gotta edit it down. But it is easier to edit down than to like build up more and more. So so I definitely hear you on that. was there any sort of like habit or system for you that helped you kind of with either that part of the process or just helped you kind of like in your regular writing, you know, helped you sit down and get a chunk of writing done at a time?
Blake Rayfield: Yeah. Yeah. Yeah, I'm a kind of old school in the sense of making outlines before I write. So I love to make outlines and really that's kinda my writing process from the beginning and that's kind of how I Expanded it as well, right? So that I found that when I made the outline, I had an idea of about how long each section was going to be. and so the more details, the more, you know, sub-points, sub-bullets I can put inside of the outline, I could expand certain chapters and and continue on and on. And honestly, you talk about the long, the bigger picture, the the outlines really help with the bigger picture as well. so for example, I think I have a chapter early on that's about visualization. And then I have a later chapter that's about more detailed visualization. So the outlines kind of helped me build on, you know, more advanced topics of visualization without covering everything again and making sure that I got enough content in there.
Jess: Yeah, did your outline pretty much stay the same throughout the process or did you find that you had to make adjustments like as you went through?
Blake Rayfield: Ironically there was only one big change and that was just to make more content. but otherwise my outline remained pretty stable. you know, I had the outline and after some of the first feedback it was just making sure we had enough content in there.
Jess: Yeah, I don't think it's bad to have to amend your outline your table of contents while you're in progress, but I think if it does stay the same, it's uncommon and it just shows that you've done a lot of prep before, you know. Like, I mean we worked together on putting together a really strong proposal for Wiley. I know we went back and forth on the proposal and sort of like kind of work to find the right niche, work to find the right place for it on Wiley's list, which is something that I think a lot of authors don't really think about is like, what other books has Wiley published? What are they working on right now that we, you know, might not be clear in the market yet and like where do they want to like where do they want their list to go? so I think just like the publisher's list is something that's like kind of a foreign concept to a lot of authors, which which like is where I think I I'm able to help as an agent to start to anticipate like where p you know, what publishers are looking to do. I think, Wiley in particular, you know, unlike O'Reilly, who has been publishing in this space for a very long time and has a really robust list in terms of AI, machine learning, data science, math. I know because I published a bunch of them while I was there. Wiley was sort of diving into this this topic area a little later. so they had a lot of a lot of like kind of gaps in their in their lists where there were like bigger opportunities for us to fill. So so that was fun to be able to kind of like work with them to find the right like hone in on the right proposal for them. How was that process, do you think, for you in terms of crafting the proposal and tweaking it to kind of get it to an acceptable stage?
Blake Rayfield: Man, I mean to your point, Jess, you're a lifesaver there. So I mean you obviously as you just pitched to us here, I mean you did a great job. it's a crowded space, right? So you have all the technical books that are there. You mentioned O'Reilly, I mean all their content, Wiley, and and really self-publishing too, right? You have all of this stuff out there. Finding the right niche was really difficult. And that's definitely where you came and helped. and I think that was one of the things for me is that I just wanted to make sure I was writing a book that was not like I'm not just doing version two of you know math with Python. So yeah thank you.
Jess: Yeah, of course, of course. Yeah, I think it's for folks who you know, I think like in your case, like you're good at a lot of different things. Like you and you can kinda you're good at honing your material in a certain direction. and and not everyone is either good at that or really wants to do that. Like they might have a really clear idea of what they want to publish and they're not really looking to like adjust it based on the publisher's list. So for the for folks who are listening who they feel like maybe they fall into that category. be looking at the full spectrum of publishers who you might want to work with. and self-publishing is an is an option as well. I mean, I'm a traditional publisher at heart, so I don't always recommend self publishing, but it it is a suitable direction for a lot of folks. but certainly like look at O'Reilly, look at Wiley, Pearson, Manning, no starch press, packed. Like there's quite a lot who publish in this space who and you can pitch multiple publishers at a time. So, and if you'd like help with that, that's where I can help you. but that's that's the idea is like don't feel like you have to get you kind of trained in on one publisher and what they're looking for. You have the option to to pitch elsewhere. and you know, I th like every publisher is really different. I reference Wiley and O'Reilly a lot. I just think that their their process is really good. and I haven't like sold anything to Manning or no starch just yet. So so I have like less less knowledge there. But but yeah, I think it was really fun working with Wiley. The they've been really easy, I think. It sounds like, you know, you had a you really enjoyed your time working with them too.
Blake Rayfield: Yeah. I think at first too there was a little bit of a learning curve, right? So when you're trying to make sure that you fit it into their typesetting type things, you know, there's always a learning curve. But once you get used to it, I am I my word is still in their typesetting like a template. So yeah, it's once you get into it, it gets going and you're you're going off to the races.
Jess: Yeah. That's great. Yeah. Yeah. different publishers will have different expectations in terms of how they want you to craft your content. So that's another thing to keep in mind too. It's an and sort of what process works best for you. So let's talk a little bit about your like teaching. and you know, how I mean we're just in a we're in a new era now with like all these LLMs and you know, Gen AI tools and things like that. I imagine that's affected your teaching in a lot of ways, both from kind of what you're expecting your students to be able to do and and how, and then as well, like how you potentially you're using these tools in your in crafting your your program and the structure of your courses. But l I'd love to hear a little bit of like where do you think there's you've seen the biggest changes and adjustments.
Blake Rayfield: man, yeah, I'll just casually bust out the biggest thing to happen to education. And in many years I'll be able to wrap up in about two minutes. no, there's there's so much that has changed, to be honest. I mean I can't even I don't want to say I can't remember, but it's just it's just amazing how far we've come. And and I was thinking about this really deeply just the other day. I think we're just at the surface. so let me start with
Jess: Ha ha ha. Sure.
Blake Rayfield: How LLMs have changed education, right? So, first off, it it seems pretty obvious, right? I mean, a lot of these LLM tools, one of their first applications was like, How can I as a student cheat on an exam, right? how can I how can I have it write my essay? How can I have it answer my multiple choice? So, I mean, a lot of faculty, a lot of professors, a lot of teachers in general, their first knee-jerk reaction is like, this is such a bad tool because you know my students are using it to cheat, and I need to give them a stamp. approval that says like this student knows this material and if they're cheating then that must mean that LLMs are bad. However, kind of as time has evolved, A lot of teachers, including myself, have realized that you know this could be a really good opportunity, right? I mean, a lot of my students, the things that I'm asking of them today were unimaginable five, six, seven years ago, right? Because for a lot of students, they're able to go from intro material to really advanced, really cool things in a in a little time, right? And In Python, think about how lots of us learn Python, right? In particular, we learn it because we want to build something, right? Whether or not that's an app or even some kind of data science tool or something like that. I mean, LLMs help you build something. That's undisputed. So it's changed education. I think it's changed education for the better. I think it's changed probably everybody's lives a little bit for the better at this point. and there's just so many different things we can talk about.
Jess: Yeah, I love that perspective. It's n it's one I don't often hear from from teachers. So that's very optimistic so yeah, I'd love to talk a little bit about like how do you handle the kind of like cheating aspect? You know, do you just allow folks to use you know, LLMs or do you kind of craft a different kind of assessment? that is not so easily sort of solved with an LLM.
Blake Rayfield: Yeah, that's a good question. I kinda have a personal responsibility approach to the cheating market, whether it's LMs or something else. you know, when I give my students assignments, I take it kind of as a personal responsibility to make sure that not all assignments can be cheated on, if that makes sense. Right? And if you think about even how you might have done school however many years ago, right? I mean, I had some assignments that I was able to go use a Google for. I had some assignments that I did not, I had some assignments I had in class. It it's the same approach. Right. I just have to make sure that I, for example, if I give an essay or if I give a short written assignment, I I have to realize that either one that's probably going to come back with LLM written content or have that expectation, or I just need to adjust the expectations for the assignment, maybe, or maybe just change the assignment overall. Right? so those are some of the things that have changed on the cheating front. I don't think It's it's all bad, right? I mean, yes, there's a lot of things that LMs can help you cheat on, but you have to remember, and this is especially for the other educators, it's it's been kind of a dark couple of years, right? So when we had COVID and we shift most everything to online, it it was kind of a dramatic shift, right? Because we started giving a lot more leeway to students. Because they're going through this pandemic, they're going alone, maybe they're working at their parents' house and and it's just difficult for them. LMs came out at the same time as most of us were locked down or at least working from home. But as soon as we come out of that, boom, there's this tool that helps everybody cheat or on things. And so It's been kind of a tug and pull for a lot of educators. But I'll say in general, if you take the same approach as as we've always taken, right? A mix of assignments. Some assignments I do allow them to use LLMs on, others I don't. just trying to see what they actually know, and it's not always about what the LLM knows.
Jess: Yeah. Do you use any kind of like AI detection kind of software? Would that even be helpful for you? or is it like I you can tell. You can tell when AI is being used.
Blake Rayfield: okay, I we of course most universities have this ability, right? They have something, but it it's my personal kind of thing that I don't think it works very well, right? It's it's kind of statistically motivated and so detecting whether or not an LM wrote something is difficult, right? And so I can tell. Most of the times you can tell, right? Bullet points are a big giveaway. of course emojis. Hopefully must in some put emojis in there. But if there's emojis, I know that that's probably not your writing. and of course no one has spelling or grammatical mistakes anymore, which is actually better. In some senses. but yeah, so you can tell. I don't like to use the other the software because I I don't think one, I don't think it works very well, and two, I think it's hard to accuse a student of cheating without actually see seeing them in the act, right? I mean, just like any crime. We can't accuse them of cheating if they have a right answer if I didn't actually see them using Elam. So yeah.
Jess: Cool, it makes me happy I'm not an educator these days. That's hard. how do you think about teaching like Python now? is that different than you might you know than you would have pr you know, pre LLM days? you know, in it sounds like, you know, if if students can get from kind of like square one to more advanced very quickly, then like using LLMs in this process is helping them helping your students get
Blake Rayfield: Yeah.
Jess: Further along faster, helping them build stuff with Python faster.
Blake Rayfield: Yes. so what a loaded question again. I mean, these are topics we could do whole hours on, but learning Python. So when I learned Python, it was like the old way, right? I had my syntax guide, I learned like the for loops and while, and then I had to go learn what packages were relevant, mostly through Anaconda, right? Seeing which ones were like important for scientific learning and things like that. and of course LMs have changed that, right? The students that even students that don't learn programming necessarily, they can
Jess: Totally.
Blake Rayfield: just go in and plug something in and get a python script and and run it and see if it works. So Python learning Python has changed. for me personally it was really difficult because like I said I had learned python the older way learning the syntax first and so I almost had to relearn how to teach and learn Python if that makes sense right because yeah because everybody who's learning python today like
Jess: Yeah, it does.
Blake Rayfield: Yeah, maybe they go learn the syntax, maybe they go learn the ecosystem, but most of it's just gonna kind of be serendipitous, right? They just Type something in and look at this, it says import, you know, numpy. Well, I don't know what that is, but it's it does this matrix algebra. So yeah, it it it was quite a challenge, I think, at first. And so the my the way my students learn Python, I still go through syntax, but the time frame is significantly condensed, right? Because they don't have to memorize it anymore. but we're able to build things a lot faster and then it it excites them more because they see the the cooler things they can build. Build a website, they can build an app, whatever it might be. And so I think it again, just like everything, it's been positive.
Jess: That's awesome. Yeah, and then correct me if my assumptions are wrong here, but I imagine like LLMs were not like really built to do like good math. So like have has like the emergence of LLMs like changed the way you teach math at all? Or you know, or maybe less less of a change than the way you might teach like a code you know, a a programming language.
Blake Rayfield: that's a good question. I think the programming language has definitely dramatically changed. I think math is one of those foundational things that yeah, it's changed a little bit. but yeah, like you said, LMs aren't the best at math. Of course they keep getting better, and who knows by the time this podcast comes out they may be really great, right? but Yeah, you still you still need to know math, still gotta make sure it's accurate. When we talk about math, of course there's many different branches, right? And the way LLMs approach the different branches, like just say algebra versus statistics, that has dramatically changed, right? Algebra and LLMs, it doesn't really mix as well, right? But of course, LLMs and statistics, that has actually changed, right? So my students ability to do statistics in a meaningful way with LLMs is gone way above and beyond, right? so yeah it depends on the subcategory of math as well.
Jess: Yeah. Gli like tell me a little bit more about that. Is that because L LMs are predictive that it like aligns better with statistics and less with algebra? Or how does that work?
Blake Rayfield: yeah, I'd say predictive, but I also say there's a little bit of a logical element to statistics, right? So statistics, yeah, it does have some formulas, but there's a lot of like When you see this data, here's some prescriptive type methodology you can use on this type of data. And a lot of things with statistics, which drives you know the regular math people crazy, is that there are multiple ways to get multiple different answers, right? There's not just one specific number. And so with take a data set, right? so if you just have a rudimentary data set and you're trying to find housing prices based on characteristics of homes, for example, it sounds very simple, but there are many different ways you can approach that problem. Right. So whether or not you're using just linear aggression or or some other kind of model or whether or not you're just doing t-tests or and then how you interpret all of those variables, whether or not there's multicollinearity or you know, all these different things, those are the things that LLMs are actually really good at flagging for my students. So like when they go to let's just say they're trying to predict these housing prices with this data set, the LM is able to tell them, hey, you should look out for multicollinearity or hey, warning. You know, bedroom and bathrooms are related, don't put them in the same predictive model as the housing price. So those are the kind of things I think are getting a lot better actually with math and LMs.
Jess: That's great. That's really great. as you said, we could go on forever about a lot of these topics, but we are running out of time. what what's next for you? You've you've wrapped up your first book, sort of what are you planning to work on next or sort of are you kind of really just focused on on your teaching?
Blake Rayfield: I took some time after writing the first book and and I think this is common for authors, like you write the first book and you say I'll never do that again. but as I look back, I think I'd like to start a new book. maybe more in the LLM education space or something like we've been discussing here today. and so Well, I don't know. I guess the next topic. I I think I have a lot to say and I'm ready to say it. So yeah, let's figure that out, Jess. Yeah.
Jess: Let's get another book going. I love love to hear it. I know I it is it is an addictive process. And it is true. Everyone's always like, no, I'm done. I'm done for a while. And then you're like, well, I do have other ideas. Like I could do it again. So I love it. Well, let's make sure folks you know, they can go get your book now, applied math with Python.
Blake Rayfield: No. Yeah.
Jess: by Wiley and available wherever books are sold. Is there anything else? you know, anywhere people should find you or anything else that you want to plug?
Blake Rayfield: People can find me on LinkedIn, they can find me on my website, blakerayfield.com, which also has links to the book. I love to connect with any of the people who read the book. I've already gotten some feedback from individuals. And so hopefully get out there and get your coffee today.
Jess: Brilliant. Love it. Blake, thanks so much for joining the podcast today. It's been a pleasure. Looking forward to working with you on your next book. And if you enjoyed conversations like this, stay tuned for more. we've got some other really exciting guests coming on the podcast in the weeks to come. So thanks and Blake, have a good one. We'll see you again soon.
Blake Rayfield: Thank you, Jess. Bye.
Jess: Welcome to the Tech PubCast, a behind the scenes look at what it really takes to write and publish a technical book. I'm your host, Jess Haberman. I'm a literary agent, publishing consultant, and former acquisitions editor with nearly twenty years of experience publishing books. Today I'm thrilled to welcome Blake Rayfield to the podcast. Hello, Blake. Do you want to share a brief introduction?
Blake Rayfield: Thank you, Jess. My name is Blake Rayfield. I'm currently an assistant professor of finance and fintech at the University of North Florida. And of course, the most recent author of Applied Math with Python, solving real-world problems with Python. So thank you.
Jess: Co yeah, of course. Welcome. Congrats on your new book, published by Wiley. Blake and I got to work together on this project, which was really fun. you know, there's just like a good foundation in math that I think a lot of folks who kind of make their way into kind of data science and AI and machine learning, adjacent work, realize that they they they need to have. so I th I think that that would stuff
Blake Rayfield: Thank you.
Jess: you know, it's a just a great area to kind of dive in a little bit further. It's a good kind of amount to cover in a book. you know, how was the process for you? This was your, you know, your first book with a major publisher and you know, what did you learn?
Blake Rayfield: Wow, it was eye-opening. So as I said in my intro, I'm I'm an academic, so typically I'm used to publishing. I have lots of published papers. I've written book chapters for academic type books or some kind of papers that are mixing between academia and business. but this was just a totally different ballgame, right? So for me, I had to learn a lot along the way, and we can go into all those details, but I learned a lot along the way and it was just totally different than anything I'd done before.
Jess: Yeah. I mean,
Blake Rayfield: Yeah.
Jess: it was it did you feel like the writing like process was different or kind of like the expectations in terms of like how the like publishing and editing like that process goes?
Blake Rayfield: little bit of both. typically if we go into the academic like writing process, right? So I write some paper, maybe it's quantitative in nature, right? It has some results in there. mostly the audience I'm writing for is other academics, right? And so when they go read the paper, I can use all the terms and phrases that they're gonna understand. and so when I send it to a journal, right? for to hopefully to publish, I know that that's gonna be the audience that reads it and gives me feedback. Okay, and so if you can imagine that process, it's it gets to kind of be a little bit of a game, like repetitive in the sense, right? Like I I've gotten really good at that process over the years. So when I went to go write this book, which was of course not at all for that audience, right, there's just a totally new level of expectation, right? I had to make sure that I was connecting with the potential audience that may read my book. I had to make sure that I had all the, you know, of course, when you write a book for a general population or at least people who are interested in tech books, there's like a whole different expectation on making it nice to read, right? Look a lot nicer, right? In academia, we really care how it's formatted so much, right? So we had to make sure that this formatting's right, the sections are right. I mean it was just like all of that was just a great learning process though.
Jess: Yeah, that's great. There's the sort of vernacular, like there's the vocabulary that you that you should and shouldn't be using. Like you're trying to simplify things because right, you're speaking to like a more general audience who, may or may not understand some of the more technical concepts and jargon. and And it does have to look different on the page, you know, 'cause you are focused on like a print you know, print books are still selling, are still very popular. I do that too. I did a lot of like sports and nonfiction and history and outdoor recreation books like early in my career and
Blake Rayfield: Mm-hmm.
Jess: And we had like s different like trim sizes of books. So for example, like some that were really small. So like we were really focused on like where line breaks happened and like you would mark up like we're what were what are called widows and orphans, which like I think even in fiction a lot of fiction publishers don't even worry about anymore. but it was like it was like so specific to like have it look a certain way on the page or fit or like really specific trim size. So I you know, I don't think Wiley is gonna is really going too far in the weeds there, but but yeah it's totally different from academic publishing and it's interesting because you're you are getting like the kind of the review period. Maybe talk a little bit about like technical review, you know, how did that work in your process? Like what did you learn out of the feedback that you're getting from from your technical reviewers.
Blake Rayfield: Sure, so of course like the techno reviewers, the ones who are looking over the code or the math for the book, it it was a different process as well, right? So someone's constantly checking even small snippets of code that I'm putting in there, to large snippets of code. And I think for me, some of the really quick feedback that I got was making sure I'm connecting everything. So again, as someone who writes it either for academia or maybe you you're even used to writing code. We know comments are a thing, but even with comments, it doesn't leave enough breadcrumbs for a reader to be able to connect all the sections of code throughout a chapter, or even throughout the entire book, right? And so a lot of times in in chapters, I would write snippets, small little snippets of code, but maybe that snippet was dependent upon the previous snippet or a section. And that section may have been dependent on something I loaded in a previous section. So that was some of the initial feedback. that I received. Otherwise the code I mean if you know how to write code, the code is that. so I think that was the that was the biggest thing.
Jess: for folks who are starting, like their writing or working with a publisher is like what are the expectations for your code? Is it expected to be like, you know, you're div you're delivering a full like Jupiter notebook of code that y can run alongside the book, maybe as like a a downloadable for the reader? Or is it like, yeah, you're illustrating certain concepts with snippets that don't necessarily have to connect to each other. But I think it is really important to think about like how your code like builds or does it build on on prior snippets, things like that.
Blake Rayfield: to add to that, I mean you mentioned Jupyter Notebook, right? So that's how a lot of people are used to writing these this type of code, right? So I mean, of course you have the app type, which is not in a Jupyter notebook, but for math and science and you know statistics, we're used to Jupyter notebooks. And so with Jupyter notebooks, everything runs sequentially and it's all right there, right? So you don't really need to connect it. But and and honestly, when I was writing the chapters, that's how I wrote it. It was in Jupyter Notebooks, and so for me it all made sense. But for the Readers, you're seeing it presented as snippet of code with excl you know, exclamations and and all sorts of other details in there. So it it doesn't come like that free-flowing.
Jess: Yeah. Yeah, it's interesting the way that the format of Jupiter notebooks have helped us to think about like coding differently and help us to think about writing books differently. Like when I was at O'Reilly, we did have a lot of of authors submit books via Jupyter notebook or at least use that kind of style, that linear style of like everything sort of building on in that you can mix sort of, you know, markup and and code. Jupyter notebooks have not been around forever, you know. So it's it's feels like a kind of interesting format, you know, that really aligned itself really well with books. but yeah, it sounds like you found that format pretty useful for for your You're drafting.
Blake Rayfield: Yeah, absolutely. And I mean to put in a plug for the book too, I put my Jupyter notebooks on GitHub, so they're available. I mean, as you read the chapters, you could just go reference those. And of course the code is run sequentially there. So even if you missed the footnote that says you needed to run a previous section, it's all there.
Jess: Yeah, exactly. I guess in terms of like expectations when you're going into working with with a publisher on this kind of book, like what were some of the like the biggest misconceptions that you know you sort of were were clarified for you during this process?
Blake Rayfield: I think having edited a technical book before I wrote the technical book. You know, you learn pretty quickly that editors all have different expectations of what that technical book will both look like and also what they expect, for example, the chapter links to be or the sections to look like, or it you know, it's really from person to person. So I think that was one of the biggest misconceptions that I was going in thinking it was gonna be like, the other technical book that I had worked on, or maybe one that I had read. whereas again, it's kind of a dance between you and your other editors on on the team to see like what is gonna work for you and them and everybody. So I think it's really you know collective work. So I'm I think that was one of the biggest misconceptions.
Jess: Yeah. I think to your earlier point about reviewing content in sections, like that's another real benefit of having a technical reviewer who is with you for the whole book. because like they can help you see like where you need to kind of make connections across different chapters and and code snippets and things like that. They see kind of like the long game. so that's something to keep in mind when you're like often a publisher. It will ask authors for recommendations on who they think would be a good a good tech reviewer. So someone who, you know, has the has the time and attention to support you throughout the whole process is is great. Doesn't have to work that way, but but it does I think really benefit the sort of the long tail of writing, which like it is like a long process.
Blake Rayfield: and absolutely. I can say, I mean, your team is the one kind of that covers for you too. So I mean as you start to get in the weeds deep in the book, they may have shortened deadlines, you may have shortened deadlines. So it's it's gotta be a great relationship. And luckily I had a great team. I loved everybody that was there. So but yeah, it's it's definitely important to have a good team.
Jess: Yeah, yeah, absolutely. Was there anything that you found like to be harder than expected? Like what was hardest for you while while you're working on the book?
Blake Rayfield: Just writing enough. I mean, I'll go back to my academic roots. So going back to academic roots, we kind of get to the point, right? So meaning like when when there's a point we're trying to make, it's it's a pretty short and condensed, like here's the point, right? But here I had to I think one of the harder things for me, and this is one of those like learn bad behaviors, is that I had to make sure that I kind of gave it enough detail, and kind of
Jess: Yeah.
Blake Rayfield: talked about it in more than just here's the point. So that was one of the most difficult things for me to get over. But that's not gonna be everybody.
Jess: Yeah, right. a lot of folks have the opposite problem. They write too much and then you you gotta edit it down. But it is easier to edit down than to like build up more and more. So so I definitely hear you on that. was there any sort of like habit or system for you that helped you kind of with either that part of the process or just helped you kind of like in your regular writing, you know, helped you sit down and get a chunk of writing done at a time?
Blake Rayfield: Yeah. Yeah. Yeah, I'm a kind of old school in the sense of making outlines before I write. So I love to make outlines and really that's kinda my writing process from the beginning and that's kind of how I Expanded it as well, right? So that I found that when I made the outline, I had an idea of about how long each section was going to be. and so the more details, the more, you know, sub-points, sub-bullets I can put inside of the outline, I could expand certain chapters and and continue on and on. And honestly, you talk about the long, the bigger picture, the the outlines really help with the bigger picture as well. so for example, I think I have a chapter early on that's about visualization. And then I have a later chapter that's about more detailed visualization. So the outlines kind of helped me build on, you know, more advanced topics of visualization without covering everything again and making sure that I got enough content in there.
Jess: Yeah, did your outline pretty much stay the same throughout the process or did you find that you had to make adjustments like as you went through?
Blake Rayfield: Ironically there was only one big change and that was just to make more content. but otherwise my outline remained pretty stable. you know, I had the outline and after some of the first feedback it was just making sure we had enough content in there.
Jess: Yeah, I don't think it's bad to have to amend your outline your table of contents while you're in progress, but I think if it does stay the same, it's uncommon and it just shows that you've done a lot of prep before, you know. Like, I mean we worked together on putting together a really strong proposal for Wiley. I know we went back and forth on the proposal and sort of like kind of work to find the right niche, work to find the right place for it on Wiley's list, which is something that I think a lot of authors don't really think about is like, what other books has Wiley published? What are they working on right now that we, you know, might not be clear in the market yet and like where do they want to like where do they want their list to go? so I think just like the publisher's list is something that's like kind of a foreign concept to a lot of authors, which which like is where I think I I'm able to help as an agent to start to anticipate like where p you know, what publishers are looking to do. I think, Wiley in particular, you know, unlike O'Reilly, who has been publishing in this space for a very long time and has a really robust list in terms of AI, machine learning, data science, math. I know because I published a bunch of them while I was there. Wiley was sort of diving into this this topic area a little later. so they had a lot of a lot of like kind of gaps in their in their lists where there were like bigger opportunities for us to fill. So so that was fun to be able to kind of like work with them to find the right like hone in on the right proposal for them. How was that process, do you think, for you in terms of crafting the proposal and tweaking it to kind of get it to an acceptable stage?
Blake Rayfield: Man, I mean to your point, Jess, you're a lifesaver there. So I mean you obviously as you just pitched to us here, I mean you did a great job. it's a crowded space, right? So you have all the technical books that are there. You mentioned O'Reilly, I mean all their content, Wiley, and and really self-publishing too, right? You have all of this stuff out there. Finding the right niche was really difficult. And that's definitely where you came and helped. and I think that was one of the things for me is that I just wanted to make sure I was writing a book that was not like I'm not just doing version two of you know math with Python. So yeah thank you.
Jess: Yeah, of course, of course. Yeah, I think it's for folks who you know, I think like in your case, like you're good at a lot of different things. Like you and you can kinda you're good at honing your material in a certain direction. and and not everyone is either good at that or really wants to do that. Like they might have a really clear idea of what they want to publish and they're not really looking to like adjust it based on the publisher's list. So for the for folks who are listening who they feel like maybe they fall into that category. be looking at the full spectrum of publishers who you might want to work with. and self-publishing is an is an option as well. I mean, I'm a traditional publisher at heart, so I don't always recommend self publishing, but it it is a suitable direction for a lot of folks. but certainly like look at O'Reilly, look at Wiley, Pearson, Manning, no starch press, packed. Like there's quite a lot who publish in this space who and you can pitch multiple publishers at a time. So, and if you'd like help with that, that's where I can help you. but that's that's the idea is like don't feel like you have to get you kind of trained in on one publisher and what they're looking for. You have the option to to pitch elsewhere. and you know, I th like every publisher is really different. I reference Wiley and O'Reilly a lot. I just think that their their process is really good. and I haven't like sold anything to Manning or no starch just yet. So so I have like less less knowledge there. But but yeah, I think it was really fun working with Wiley. The they've been really easy, I think. It sounds like, you know, you had a you really enjoyed your time working with them too.
Blake Rayfield: Yeah. I think at first too there was a little bit of a learning curve, right? So when you're trying to make sure that you fit it into their typesetting type things, you know, there's always a learning curve. But once you get used to it, I am I my word is still in their typesetting like a template. So yeah, it's once you get into it, it gets going and you're you're going off to the races.
Jess: Yeah. That's great. Yeah. Yeah. different publishers will have different expectations in terms of how they want you to craft your content. So that's another thing to keep in mind too. It's an and sort of what process works best for you. So let's talk a little bit about your like teaching. and you know, how I mean we're just in a we're in a new era now with like all these LLMs and you know, Gen AI tools and things like that. I imagine that's affected your teaching in a lot of ways, both from kind of what you're expecting your students to be able to do and and how, and then as well, like how you potentially you're using these tools in your in crafting your your program and the structure of your courses. But l I'd love to hear a little bit of like where do you think there's you've seen the biggest changes and adjustments.
Blake Rayfield: man, yeah, I'll just casually bust out the biggest thing to happen to education. And in many years I'll be able to wrap up in about two minutes. no, there's there's so much that has changed, to be honest. I mean I can't even I don't want to say I can't remember, but it's just it's just amazing how far we've come. And and I was thinking about this really deeply just the other day. I think we're just at the surface. so let me start with
Jess: Ha ha ha. Sure.
Blake Rayfield: How LLMs have changed education, right? So, first off, it it seems pretty obvious, right? I mean, a lot of these LLM tools, one of their first applications was like, How can I as a student cheat on an exam, right? how can I how can I have it write my essay? How can I have it answer my multiple choice? So, I mean, a lot of faculty, a lot of professors, a lot of teachers in general, their first knee-jerk reaction is like, this is such a bad tool because you know my students are using it to cheat, and I need to give them a stamp. approval that says like this student knows this material and if they're cheating then that must mean that LLMs are bad. However, kind of as time has evolved, A lot of teachers, including myself, have realized that you know this could be a really good opportunity, right? I mean, a lot of my students, the things that I'm asking of them today were unimaginable five, six, seven years ago, right? Because for a lot of students, they're able to go from intro material to really advanced, really cool things in a in a little time, right? And In Python, think about how lots of us learn Python, right? In particular, we learn it because we want to build something, right? Whether or not that's an app or even some kind of data science tool or something like that. I mean, LLMs help you build something. That's undisputed. So it's changed education. I think it's changed education for the better. I think it's changed probably everybody's lives a little bit for the better at this point. and there's just so many different things we can talk about.
Jess: Yeah, I love that perspective. It's n it's one I don't often hear from from teachers. So that's very optimistic so yeah, I'd love to talk a little bit about like how do you handle the kind of like cheating aspect? You know, do you just allow folks to use you know, LLMs or do you kind of craft a different kind of assessment? that is not so easily sort of solved with an LLM.
Blake Rayfield: Yeah, that's a good question. I kinda have a personal responsibility approach to the cheating market, whether it's LMs or something else. you know, when I give my students assignments, I take it kind of as a personal responsibility to make sure that not all assignments can be cheated on, if that makes sense. Right? And if you think about even how you might have done school however many years ago, right? I mean, I had some assignments that I was able to go use a Google for. I had some assignments that I did not, I had some assignments I had in class. It it's the same approach. Right. I just have to make sure that I, for example, if I give an essay or if I give a short written assignment, I I have to realize that either one that's probably going to come back with LLM written content or have that expectation, or I just need to adjust the expectations for the assignment, maybe, or maybe just change the assignment overall. Right? so those are some of the things that have changed on the cheating front. I don't think It's it's all bad, right? I mean, yes, there's a lot of things that LMs can help you cheat on, but you have to remember, and this is especially for the other educators, it's it's been kind of a dark couple of years, right? So when we had COVID and we shift most everything to online, it it was kind of a dramatic shift, right? Because we started giving a lot more leeway to students. Because they're going through this pandemic, they're going alone, maybe they're working at their parents' house and and it's just difficult for them. LMs came out at the same time as most of us were locked down or at least working from home. But as soon as we come out of that, boom, there's this tool that helps everybody cheat or on things. And so It's been kind of a tug and pull for a lot of educators. But I'll say in general, if you take the same approach as as we've always taken, right? A mix of assignments. Some assignments I do allow them to use LLMs on, others I don't. just trying to see what they actually know, and it's not always about what the LLM knows.
Jess: Yeah. Do you use any kind of like AI detection kind of software? Would that even be helpful for you? or is it like I you can tell. You can tell when AI is being used.
Blake Rayfield: okay, I we of course most universities have this ability, right? They have something, but it it's my personal kind of thing that I don't think it works very well, right? It's it's kind of statistically motivated and so detecting whether or not an LM wrote something is difficult, right? And so I can tell. Most of the times you can tell, right? Bullet points are a big giveaway. of course emojis. Hopefully must in some put emojis in there. But if there's emojis, I know that that's probably not your writing. and of course no one has spelling or grammatical mistakes anymore, which is actually better. In some senses. but yeah, so you can tell. I don't like to use the other the software because I I don't think one, I don't think it works very well, and two, I think it's hard to accuse a student of cheating without actually see seeing them in the act, right? I mean, just like any crime. We can't accuse them of cheating if they have a right answer if I didn't actually see them using Elam. So yeah.
Jess: Cool, it makes me happy I'm not an educator these days. That's hard. how do you think about teaching like Python now? is that different than you might you know than you would have pr you know, pre LLM days? you know, in it sounds like, you know, if if students can get from kind of like square one to more advanced very quickly, then like using LLMs in this process is helping them helping your students get
Blake Rayfield: Yeah.
Jess: Further along faster, helping them build stuff with Python faster.
Blake Rayfield: Yes. so what a loaded question again. I mean, these are topics we could do whole hours on, but learning Python. So when I learned Python, it was like the old way, right? I had my syntax guide, I learned like the for loops and while, and then I had to go learn what packages were relevant, mostly through Anaconda, right? Seeing which ones were like important for scientific learning and things like that. and of course LMs have changed that, right? The students that even students that don't learn programming necessarily, they can
Jess: Totally.
Blake Rayfield: just go in and plug something in and get a python script and and run it and see if it works. So Python learning Python has changed. for me personally it was really difficult because like I said I had learned python the older way learning the syntax first and so I almost had to relearn how to teach and learn Python if that makes sense right because yeah because everybody who's learning python today like
Jess: Yeah, it does.
Blake Rayfield: Yeah, maybe they go learn the syntax, maybe they go learn the ecosystem, but most of it's just gonna kind of be serendipitous, right? They just Type something in and look at this, it says import, you know, numpy. Well, I don't know what that is, but it's it does this matrix algebra. So yeah, it it it was quite a challenge, I think, at first. And so the my the way my students learn Python, I still go through syntax, but the time frame is significantly condensed, right? Because they don't have to memorize it anymore. but we're able to build things a lot faster and then it it excites them more because they see the the cooler things they can build. Build a website, they can build an app, whatever it might be. And so I think it again, just like everything, it's been positive.
Jess: That's awesome. Yeah, and then correct me if my assumptions are wrong here, but I imagine like LLMs were not like really built to do like good math. So like have has like the emergence of LLMs like changed the way you teach math at all? Or you know, or maybe less less of a change than the way you might teach like a code you know, a a programming language.
Blake Rayfield: that's a good question. I think the programming language has definitely dramatically changed. I think math is one of those foundational things that yeah, it's changed a little bit. but yeah, like you said, LMs aren't the best at math. Of course they keep getting better, and who knows by the time this podcast comes out they may be really great, right? but Yeah, you still you still need to know math, still gotta make sure it's accurate. When we talk about math, of course there's many different branches, right? And the way LLMs approach the different branches, like just say algebra versus statistics, that has dramatically changed, right? Algebra and LLMs, it doesn't really mix as well, right? But of course, LLMs and statistics, that has actually changed, right? So my students ability to do statistics in a meaningful way with LLMs is gone way above and beyond, right? so yeah it depends on the subcategory of math as well.
Jess: Yeah. Gli like tell me a little bit more about that. Is that because L LMs are predictive that it like aligns better with statistics and less with algebra? Or how does that work?
Blake Rayfield: yeah, I'd say predictive, but I also say there's a little bit of a logical element to statistics, right? So statistics, yeah, it does have some formulas, but there's a lot of like When you see this data, here's some prescriptive type methodology you can use on this type of data. And a lot of things with statistics, which drives you know the regular math people crazy, is that there are multiple ways to get multiple different answers, right? There's not just one specific number. And so with take a data set, right? so if you just have a rudimentary data set and you're trying to find housing prices based on characteristics of homes, for example, it sounds very simple, but there are many different ways you can approach that problem. Right. So whether or not you're using just linear aggression or or some other kind of model or whether or not you're just doing t-tests or and then how you interpret all of those variables, whether or not there's multicollinearity or you know, all these different things, those are the things that LLMs are actually really good at flagging for my students. So like when they go to let's just say they're trying to predict these housing prices with this data set, the LM is able to tell them, hey, you should look out for multicollinearity or hey, warning. You know, bedroom and bathrooms are related, don't put them in the same predictive model as the housing price. So those are the kind of things I think are getting a lot better actually with math and LMs.
Jess: That's great. That's really great. as you said, we could go on forever about a lot of these topics, but we are running out of time. what what's next for you? You've you've wrapped up your first book, sort of what are you planning to work on next or sort of are you kind of really just focused on on your teaching?
Blake Rayfield: I took some time after writing the first book and and I think this is common for authors, like you write the first book and you say I'll never do that again. but as I look back, I think I'd like to start a new book. maybe more in the LLM education space or something like we've been discussing here today. and so Well, I don't know. I guess the next topic. I I think I have a lot to say and I'm ready to say it. So yeah, let's figure that out, Jess. Yeah.
Jess: Let's get another book going. I love love to hear it. I know I it is it is an addictive process. And it is true. Everyone's always like, no, I'm done. I'm done for a while. And then you're like, well, I do have other ideas. Like I could do it again. So I love it. Well, let's make sure folks you know, they can go get your book now, applied math with Python.
Blake Rayfield: No. Yeah.
Jess: by Wiley and available wherever books are sold. Is there anything else? you know, anywhere people should find you or anything else that you want to plug?
Blake Rayfield: People can find me on LinkedIn, they can find me on my website, blakerayfield.com, which also has links to the book. I love to connect with any of the people who read the book. I've already gotten some feedback from individuals. And so hopefully get out there and get your coffee today.
Jess: Brilliant. Love it. Blake, thanks so much for joining the podcast today. It's been a pleasure. Looking forward to working with you on your next book. And if you enjoyed conversations like this, stay tuned for more. we've got some other really exciting guests coming on the podcast in the weeks to come. So thanks and Blake, have a good one. We'll see you again soon.
Blake Rayfield: Thank you, Jess. Bye.
