In this episode of Data + Curiosity, I had a fantastic time chatting with Rachael Tatman about Twitter for Developer Advocacy, the definition of NLP and why there’s so much hype around it, personal branding, and getting started with tabletop gaming!
I learned so much from Rachael, and am excited to share the conversation with you here:
You can also read a lightly edited transcript of our conversation below:
JESSE MOSTIPAK: So I would love to talk a little bit about Twitter because it is November 15th. I don't want to get into, oh, this happened and this happened and this happened. But you and I first met on Twitter.
RACHAEL TATMAN: We did.
JESSE: You got me my first DevRel job via Twitter. So lots of love and appreciation for Twitter in general, but I know for both of us Twitter has been a big part of our DevRel careers.
RACHAEL: Yes.
JESSE: And so just what does this mean. Let's just say Twitter will go down and come back in some other form. We don't know what that is. But what does this mean for DevRel because I feel like data science, machine learning DevRel is very Twitter-focused.
RACHAEL: Yes. Definitely something that I've been thinking about. And I feel the two main uses of Twitter for me are I guess collegial interaction, like being able to talk with peers even if I'm not physically there, which I haven't been recently. And be frank, don't think that I'm going to continue to be in the future. So that's one big thing. Being able to talk to folks, who just like an informal back and forth way. And I think the history of the internet there's always been something like that like IRC, Messenger.
Twitter is not the new thing. The new thing is now that all of those things are a little bit more public. You have usenet groups. And then the other thing is as a place to find information as an information aggregator. And certainly what I'm doing is I am splitting those functions. So I mean, I am-- hi yes, sorry. Do you want to come up and-- just going to-- so you can see.
JESSE: Yeah, we love pets. Let's say hi.
RACHAEL: He’s so sleepy.
JESSE: Can we get a formal introduction. Oh, my gosh.
RACHAEL: No.
JESSE: I would get nothing done.
RACHAEL: I know. This is Benson. He's very sleepy and he wants me to put him back down so we can go back to sleep.
JESSE: Well, it was nice meeting you.
RACHAEL: He's a little chihuahua mix. He loves to be a lap dog.
So the two functions. I want to chat with people and I also want to know what's going on. And chit-chatting with people, I think I've committed to Mastodon at the moment. We'll see how it goes. I'm sure some people will go to Reddit. And knowing Reddit has never been a place where I feel comfortable and welcome. Let me put it that way. I'm sure some people are going to Discords and to more private servers for that collegial interaction. I know some people are going back to conferences in person, which like I said, I'm not doing. But I think some people are getting that need filled that way.
I think it's just going to become a little bit less centralized. I think we're going to see more little areas of discussion in different places, which honestly in terms of what I like in social media is my preference. I know when we were talking for my channel, we talked a little bit about compartmentalizing. I use different social media platforms for different things. So that's the one thing. That's the chit-chat.
And the other side is learning new things and aggregating them. And for that, I mean, I think obviously, newsletters are going to take some of that burden. I think I know a lot of people have run up Substacks recently if they didn't already have one. Definitely part of that. Obviously, not as good for things that are happening faster. And I think that's the thing that we are losing now is a way to quickly broadcast things as they happen. So, for example, the tornado like we had to reschedule this because I had a tornado warning, the one we're like we can see the spinny thing. Did it from your basement.
JESSE: It's like watch versus warning. I lived in Dallas for eight years, tornado country, still cannot tell you-- and there's like all these analogies. And I'm like, let's just change the name. But yes, you had the one where it's coming.
RACHAEL: Yes, but we mean it. And the tornado sirens are going off in our area. Anyway, very guarding. Very much a bracing sound. And that's the thing, I just don't know where that's going to go. Is microblogging and real time microblogging going to continue to be a thing long-term because Mastodon is very decentralized. That is the point of the Fed versus that it is decentralized. Are we going to see competitors swooping in is like Snap going to get into that space. Is TikTok, God forbid, going to get into that space. Sorry, that was-- I don't have an issue with short-form video. TikTok as a tech company I trust zero. Absolutely not--
JESSE: There's a lot-- there's a lot of privacy issues with TikTok for sure.
RACHAEL: Which like similarly, Facebook. I also don't trust Facebook.
JESSE: Listen, our data is for sale. Whatever you do, we are being-- not to be conspiracy and I don't feel this as conspiracy theory. So--
RACHAEL: Data broker is a type of company. They exist.
JESSE: Absolutely. I actually pay money every month to a company to remove my data from things. Your phone tracks you. It tracks your movement. It was used during early days of COVID. We started looking at phone data to see where people were and how they were moving. So data is-- personal data is a business. And I think everyone is allowed to have concerns about how their data is used. TikTok, I don't see as a way like what you're talking about is so important.
RACHAEL: It was the thing that Twitter did.
JESSE: It was, yeah. So Twitter was so good during times of crisis, of keeping people informed. And there was largely a sense of, hey, this emergency is happening, like divert resources to that. Twitter has been my source of information for public health around COVID.
RACHAEL: Same, yeah. Definitely, especially the epidemiologist that I followed because of my [INAUDIBLE] graduate work or like the protests in Iran. Oh, here's a great one, the protests in Nigeria that affected tech workers so much. Twitter is how I found out about it and was able to keep up with it. I don't know that we're going to continue to have that public square. And here's the thing, Twitter is still there. And so a community member might trust the platform is rapidly going away. I am double verifying things now, like I used to double-check. I'm being much more careful now with what I share because, I mean, there's the blue check debacle, two-factor authentication was broken.
We were just talking about before the call, reports are that they don't have a data protection officer and are flagrantly in violation of GDPR. We'll see obviously I shouldn't say obviously but it sounds like they might be violating their consent decree. So in terms of user trust and safety, I think we're losing that in terms of the technical capabilities of the platform that continue to exist.
JESSE: That is the thing I worry about in DevRel is that one day I'm going to try to log on to Twitter and it's going to be a blank screen. And we don't know-- there's no warning for that. And in terms of DevRel, it is not the end of the world. We will go on. And I've heard people talk about all of these other platforms and this kind of fragmentation that's happening as lifeboats. They may not be permanent homes, but there are ways for us to stay connected to our data community, until we figure out the next steps.
So we've talked when we were on your channel, we talked a little bit. We got into this mode of auntie's talking about early data scientists and career transitioners. And you have-- I don't know if this was given publicly or if it was just internal, but you have a talk called put a bird on it. And so this is going back to your Kaggle days.
And it's this beautiful slide deck and it's about putting your Twitter handle on every slide. And anything you do, put your Twitter handle on it. Twitter is the way to get your-- I mean, I got my job at Kaggle through Twitter because of our connection. What advice do you have for someone early in their data science career, who is trying to break into DevRel or even data science right now in terms of getting their work seen and noticed?
RACHAEL: Great question. I mean, I wouldn't put that much weight on Twitter now. I mean remains to be seen. There are a lot of platforms that are really focused on helping people get stuff out there. So I would probably steer away from Medium because of the paywall issues. So dev.to does a lot of really good-- it's very focused on developers. They do a lot of promoting of the developer content that folks write on there. Lots of early career folks get like blog posts boosted by them. So I think that's a good place to do blogging.
I mean, obviously, we both have institutional links to Kaggle. But that is a good place to build a portfolio. I would say my main advice particularly if you think you might be interested in going in the DevRel direction is don't become too dependent on a platform, which obviously, Twitter was and continues to be for now one of the main platform. I've got Twitter. I've got Mastodon. But I also have my personal website that I host and run myself. Definitely going to GitHub, very important.
If you're really in the MLOps or ML side of things, maybe being a little bit more active on Hugging Face. So it seems like there's a lot of activity there right now so that might be a good place,
If you're interested in video, so one thing that I do is I stream-- when I stream both to Twitch and YouTube. Twitch doesn't hold on to videos anymore so I recommend definitely doing something that's a little bit more archival if you are interested in video and streaming. That does mean you can't become an affiliate or a partner because that has an exclusivity requirement for the content.
But for me that's worth it because I'm not as interested in building up just my YouTube channel, just my Twitter, just my Twitch. I'm really interested in creating a lot of ways for people to interact with what I'm providing and what I have to say and then my main-- because right now I'm working independently. I'm not working for a company. I'm doing contract work here and there. But the main way that I'm supporting my DevRel work is through a Ko-Fi, which is like Patreon, but takes less fees, to be perfectly blunt about it. Is A Ko-Fi and then people are supporting me. So that's like, I'm not interested in getting Twitter ad revenue, which is small.
JESSE: It's small. It's small. And Twitch payouts are small as well. So partnering with the platform, you hear about all these people, they're big streamers on Twitch. And then Twitch changes their revenue plan and now all of a sudden you are making a fraction of what you were making last month.
RACHAEL: Absolutely.
JESSE: There's definitely something in there about controlling as much as you can, at some level. And I love your recommendation for Kaggle. I think that was one thing that really surprised me when I started there was finding out that people get hired from Kaggle, the platform. I mean, people obviously working at Kaggle can go on to other jobs. But speaking as a Kaggler as someone who is curating data sets, writing really good notebooks, that alone can help you get noticed. That can be a great place to point to your portfolio of work. And I think that's because it's not just competitions. My little Kaggle plug. It's not just competitions. Competitions are great, but there are ways to engage in DevRel style work, that I think are really important.
RACHAEL: Absolutely. I build links in a community. That's the thing that I'm most interested in. I say on my channel, I'm interested in language technology and other people. I want to build links with other people and people who care about what I have to say. I'm not going to go to a dog fanciers YouTube channel, check out my video on NLP.
JESSE: On NLP.
RACHAEL: So finding people who are already interested in the things that you're talking about is really powerful, which is why Kaggle is helpful, which is why dev.to is hopeful is because that's a narrow focus in the community. Not that narrow, but you know what I mean.
JESSE: I mean, like it's not you're not publishing on Medium. You're publishing on a platform for data science and machine learning. And so in all of this content, you've mentioned NLP. And you have a tagline or catch phrase in this goal of making NLP boring. First of all, what is NLP and why is there so much hype around it?
RACHAEL: Great question so NLP, natural language processing, is the name of a field that used to be called computational linguistics and also human language technology. Sometimes you'll hear that. And it is generally the field that works with text data in a computational setting. Very broad. But traditionally, at least NLP didn't work with speech data, for example. It didn't really work with sign language data. It was really focused on text, specifically. And now the borders are getting a little bit fuzzier.
So that's the general field. Some key NLP things you're probably familiar with from the early history of the field things like spell check, things like autocomplete, so things like those search recommendations on Google. If you've ever used some ontology or semantic search like a library website, trying to find a specific piece of information. Those are like NLP adjacent. So that's the general field.
And NLP is adjacent to machine learning in general, has machine learning gone through a lot of the same winters and summers, that were springs, so some people talk about them. So way back in the '60, '70s, people started to get really excited about it. That was in the-- if you've heard about ELIZA, it was one of the first conversational chatbots where you could say something and it would respond. And you can say anything. You didn't have to press a number to pick your thing.
And this was back in the day of rules and parsing. So nothing machine learning, nothing fuzzy, nothing sarcastic, 100% rule-based, and taking text input, inferring the internal structure of it and making transformations, that internal structure. And then it turned out that there were limitations to that technology. And it could only do so much that people lost interest and all the money went away.
And then that pattern generally happened a couple of times. And now obviously, we're in up to a couple of months ago, we were in a situation where people were very excited about it and there was a lot of money. And if there's one thing that you can bet on, it's that there will be a boom and bust cycle in technology. Looking back in history, sure is. That is a general pattern that you can pretty much rely on.
So there were two waves in the 2000, 2010s, and the first was taking the traditional methods, statistical methods in NLP, and beginning to add more deep learning methods. So we started to get cheaper compute, things could be parallelized. These neural networks that people have been proposing for a couple of decades, suddenly became tractable, not because anything in the underlying algorithm changed, but rather because we now have the compute and the money for it.
So you got the wave of neural networks coming into the field. And then in 2017, we got the Transformer paper. And with the neural network applications they were still generally custom-built. They were still very narrow domain-focused. So nobody was really working on or successfully working on, I should say, general purpose models. And then in 2017, we got the Transformer paper, that was Vaswani, et al paper, which was originally machine translation, but was then extended into language modeling.
And language modeling is this very open-ended, very generic type of way of doing NLP tasks and has had some success, particularly in leaderboard-based applications. But it's also been folded into a lot of commercial products as well.
And with that transformation, those two phases of transformation. So first for more statistical rule-based models. The statistical models were fuzzy. They weren't all rule-based, but those are the two things that you are talking about. The neural models and then finally the transformers, which each of those steps, you got a bunch of renewed interest in the field as systems became more flexible. And also a lot of people who didn't have a lot of grounding in language as an object of study.
And with that rush of funding, there also came a lot of hype. There also came a lot of-- particularly in the commercial space, people misrepresenting the capabilities of their systems. People wanted to get funding. And I'm not saying that this is true of everybody in the field, but there were certainly unscrupulous or uninformed folks who genuinely misled people about what NLP could do and it did still has this-- and we're seeing the same thing right now with the image generation stuff. It's got like a magical, hypey like, oh, it's so cool. Oh, it's so neat.
And I don't want to downplay the changes that we've made, the new approaches that have been developed as a field and the amount of work that people have done. That's great. I'm not unhappy that there's more people in the field. Language is so important to me and language technology is so important to me and so important to modern society as a whole. I'm not upset about that. I am upset about people who are more interested in grabbing your attention than being honest with you. And that's where making it boring comes in.
The majority of the actual work that goes into building NLP systems, the majority of the actual labor is still low-level, fine-grained human labor. Especially laborers, NLP like any machine learning field these days runs on human labor that's often obfuscated away behind, oh, it's AI, oh, these magical models. And I think that leads people to have-- It's just to misinform people about the capabilities of the system, what they require to work, the true costs of the system.
There was a paper that came out today, I think, actually or at least I read about today--
JESSE: Is this about CLAP? It's like CLIP. This is me having scanned something, someone shared a paper in Slack and they were like CLIP, now CLAP. And I was like, wait, I'm talking to an NLP person. And then I forgot to follow-up.
RACHAEL: Is this with climate modeling of blue from Hugging Face people?
JESSE: No.
RACHAEL: No. Different thing.
JESSE: I'm so excited that I might know something, like the day it came out.
RACHAEL: You knew about something before I did. I'm not familiar with this. So the climate impacts of transformer models when we move from stochastic systems to neural network systems, orders of magnitude, more compute, which requires electricity and water, and has a big climate impact. Transformers even more orders of magnitude of compute. So there was a really nice project done recently that showed that a transformer-based model was 800 times bigger than an equivalent model based on, I think, LSTMs were there.
JESSE: Like the benchmark.
RACHAEL: So I don't think it's bad to be excited about things. I think it is bad to be uncritical of things. And I think it is not necessarily bad. I get why people do it. But particularly, as engineers, as technical practitioners, we got to know all about what we're building and using.
JESSE: I think that's part of it. So I think my background is in biology and biochemistry and living systems and things like that. And it always-- when you talk about labeling, I always think of the saying, garbage in, garbage out. So everything that you put into your model is based on the quality. What your model outputs is based on the quality of what you input to your model.
And so when we think about labeling, can you talk a little bit about that? I mean, one, it's human labor. I'm assuming it's a lot of like Mechanical Turk, things like that where someone is just get it-- what happens when someone is labeling audio or NLP data?
RACHAEL: Good question. It depends on the specific task. And I should say there's a couple different things that I'm thinking about here. So one is people who are working on training data. And the other is people who are particularly people who are doing content moderation with the idea that it may eventually become training data. But I think that there's also-- I'm going to come back to the labeling the training data. But speaking of content moderation because, I mean, Twitter just fired the majority of their content moderators according to reports that I've read. And we mentioned TikTok earlier. So the majority of content moderation that happens on TikTok is done by very poorly paid human workers, usually in the Global South.
JESSE: It's like the emotional, the devastating toll, like the things that we don't-- every once in a while, something will slip through, and it's really kind of horrifying. And then when you think someone is paid poorly to sift through all of this horrible content so that I never have to see it, there's a very real-- this is not automated. Content moderation is not largely automated.
RACHAEL: No. They might be able to do the first pass with automation. But the majority of content moderation, especially of video is just humans doing it. And the psychological toll of that is--
JESSE: It's immense
RACHAEL: Unjustifiable.
JESSE: Think about the amount of content-- I don't know the statistics. But if you think about the amount of content that's produced, at least our culture right now is very much in love with short, snappy content that is easy, it's fast, and it's just like, let's produce it. We are not currently in an age where we are enjoying prestige, long-form content that takes months to create. And it's a 20 to 40 minute video. We are like TikTok.
RACHAEL: Absolutely.
JESSE: I want two minutes or less, make me laugh. So there's just-- and I can't even imagine the amount of content that gets produced. And it makes me think of-- so I'm in the middle of reading Neal Stephenson's book Fall, which is I love his work, but then it always ends up being kind of true and it always makes me a little uncomfortable. So one of the things they talk about is there are these characters, who literally serve as personal editors.
So you have your ability to consume content but you have an editor who is on call 24/7, who pre-sifts all of your stuff. So like a new version of an assistant. So not someone who just schedules your calls and your meetings and arranges your calendar, but is like filtering your news feed and curating the things that you should see and that's like a service that you can pay for.
RACHAEL: It's like the Zoomer joke about my FBI agent.
JESSE: Yeah, it is exactly. It's like that idea. Instead of an FBI and like depending on where you are in society and your level of wealth, you can get your own personal editor, or if I'm remembering correctly, there are editor farms, who may oversee 10 to 20 to hundreds of people's content. Whereas if you're wealthy and well-off, you can get your personal editor to hand-curate things. And I don't think that is an impossible thing to imagine.
RACHAEL: I mean, that's kind of what I do on my YouTube channel. One of the-- I do two streams a week and one of the streams is like here's what I thought was interesting and relevant and it'll be today. And God, it takes so much work to put together. This week rather. So it's a weekly stream.
JESSE: But I mean, even just curating content is hard. So being the person who is filtering content, who is doing content moderation, that is going to be a challenging job. And then there are people labeling data, which is, I mean, there's just so many things going into it.
RACHAEL: Yes, definitely. And it's– machine learning research is expensive. The compute is expensive. The data is expensive. And people don't want to do boring, repetitive tasks. If you are doing something that is a good application of machine learning, it should be boring and repetitive and fairly easy. Those are the things that are easy to automate. And those are the things that if your goal is to make sure that people have to do less work, that's the thing you want to do.
So a good example of this is a lot of the Zooniverse projects. I don't know if you're familiar with this. I think it's an offshoot of Galaxy Zoo, which was like an early citizen science project, where you click on and label different galaxy shapes with the idea of training a machine learning model. So Zooniverse is specifically for researchers and it is that you're not paid for it. It is free. It is citizen science. You are doing it out of the goodness of your heart to help advance the cause of human knowledge. Science is expensive.
JESSE: Science, it is, yes. Everything, it's all expensive. And I think it's easy to downplay because our phones are so powerful and it's expensive. But anyway so expensive. Computers are expensive.
RACHAEL: And we lost in NLP the idea that-- well, first of all, I should say most people working in NLP, certainly most researchers don't have a lot of human subjects training, if any, which has been-- I've been an ethical reviewer a couple of times for different ACL conferences, which is fairly new, by the way, since I came to the field. That's something that has been added.
And I believe NeurIPS also added ethical review as well. And that's something that there's a lot of pushback towards against-- push back against-- is the idea that we are working with human data and we should treat our-- I like to use the term data donors as people with agency and individuals who deserve the same rights and remuneration perhaps the researchers do. Not that researchers are necessarily paid that much.
While we were recording this, it's currently during the UC strike for a lot of side note. Researchers at companies generally are paid pretty well. Researchers, particularly graduate students and postdocs are generally paid very, very poorly. So when I was a graduate student, I was actually in Seattle, I applied to public housing. And I didn't make enough to qualify. So just to give you an idea of the salaries that we're talking about for some of these researchers are not high.
But compared to the commercial research labs, which is the shift in the field. We've moved from most work being done-- I mean, in the history of NLP, a lot of the early work was funded by the US Department of Defense, especially machine translation, specifically for defense. And so in the history of the field, there is that very close tie to the US government.
And then the government was a little bit less interested so you tend to get more folks working in universities. And then once commercial applications started to become more viable again, then we get-- I wouldn't say the most but like about a third of work is now done at commercial labs. And certainly the most impactful projects are done at commercial labs. And a lot of those projects are actually-- I forget who where I first heard this term, but are published by press release so they don't go through peer review.
JESSE: I didn't know any of this. So my scientific publishing is getting on paper. There's a whole process and you have to do your data. You have to do your methods, like all of that. Here I am just assuming things. That's really interesting. So not giving-- is choosing how you share information with the world. And I'm assuming if you're not beholden to R01 funding and your funding comes from private sources, how you choose to disseminate your findings now is largely up to you. You have choices.
RACHAEL: Particularly if you have a history of very clickbaity, sensationalist projects that newspapers love to publish because people read them. It's unfortunate. Even with GPT-3, the last time I looked into it, you couldn't get information on all of the data that went into training GPT-3. You got some of it, but they didn't tell you what 100% of the data is and where it was from, which perhaps that was an intentional choice on their part to avoid answering questions that they would rather not have answered. Perhaps not. Who knows. I'm not privy to that.
And then that's-- linked in to this prevailing attitude of we're both from science backgrounds. I care deeply about technology and other people and language data. I'm a linguist by training, a social scientist by training. It's important where you get your data from. And it affects everything that your model does downstream. And there's this very free for all attitude that like anything anybody's posted on the internet, we can scrape and use. And certainly, historically, court cases have certainly fallen in that way. But I don't think they're going to keep falling that way, to be perfectly honest.
JESSE: I think about that all the time. How easy is it to find me? And I think that's something that I think a lot-- think about a lot in developer advocacy. You are, in many ways, a public face for an organization and you are putting yourself out there. And what you say and how you present yourself can be scrutinized. And I think that is something to think about for people who are thinking, oh, DevRel seems fun. The role is not just writing blog posts.
RACHAEL: No.
JESSE: I mean, it can be. If you get a job that is like writing blog posts under the company name, it's very different. And I'm not saying those roles don't exist. But in many cases, DevRel is the person who is willing to give webinars or talks or be on videos and podcasts, and really put their name out there associated with things, and just be aware of what that information can be useful.
RACHAEL: Definitely. There's definitely enough of my voice online that someone could make an extremely realistic voice clone of me and that's just something that I've had to make my peace with.
JESSE: There is this-- so this is something-- this really ties into being a developer advocate and I don't know how it is for you. But I will say for myself sometimes it feels like my entire online personality has to be consistent and it has to be really focused on data science and machine learning. Is that something that you found as well or that you've experimented with?
RACHAEL: Good question. I will say that for the things that I do online that are related to data science and machine learning, I try to keep them pretty consistent. And when I am doing other things, I intentionally have separate branding just to bring branding back into the session. But I intentionally separate them. And that's just a decision I've made. But also something that I have a lot of interest in. We're not really talking about many of them now but I have a lot of interests outside of technology.
And I've been thinking about it, particularly now that I'm independent. I can do what I want. Folding those in a little bit more and really something that I've really been thinking about is I want to chase my bliss. And I'm not entirely sure where my bliss is. But I want to root around for it like a pig hunting for truffles and see if I can find a little something.
So I'm currently, my-- I stream twice a week. My other stream is like sometimes we do live coding sometimes for our papers. You do different things. But right now I'm working on a live coding project around moon names. So particularly, the traditional names of full moons in the Eastern United States because I am a gardener. I'm really interested in gardening. And I am in a place where a lot of particularly the older organic farmers, particularly. So I live in Richmond now. But I'm in the Piedmont, so closer to Appalachia, and that's where my husband's folks are from. There's something– I almost never mentioned my husband in my content. I just didn't [INAUDIBLE]. He's not relevant. Really, he's not.
But a lot of the traditional, particularly older folks here will garden according to the moon. And the phases of the moon and some folks will like what sign is the moon in. But also there are names for the different full moons, like the strawberry moon, the frog moon, the harvest moon. And so I'm doing a project on my channel like somebody told me to do something on the frog moon. Which is the frog moon? And trying to have a little nice, little project where I can, it's for me, but you are welcome to use it once it's done and out there. Figure out what that means in Gregorian calendar dates. So that's obviously a part of my life that I talk much about online until recently.
I remember I was having a discussion with a professional colleague a while ago and I mentioned something and they just paused for a couple seconds. And they were like, I forget you have a really rich life. [INAUDIBLE]
JESSE: I think that you have-- we both have many interests outside of tech, but in developer advocacy so much of our life, is talking about tech and being good at tech and knowing what's happening in tech that it can be hard to-- I've even had people say to me, I didn't know you were funny. OK, first of all, rude. Or, people will be like, oh, I didn't know you liked film.
And then I do think about how many people do I do that to as well because what we show online is not always like the full self. But speaking of non data science interests, you were the first person to introduce me to tabletop roleplaying games. So you are a first--
RACHAEL: Me?
JESSE: Like you are the first [INAUDIBLE]--
RACHAEL: That's right, yeah, it's the first one.
JESSE: --[INAUDIBLE] of me. So you hosted Totally Real Human Adults and we played three or four animals in a trenchcoat trying to get jobs as a data scientist. So I had a ton of fun. I'll post the link to the video. But this world of tabletop roleplaying games has always felt really inaccessible and overwhelming to me. It's something that I didn't know about growing up.
And it wasn't until I was older and playing World of Warcraft and people were talking about Dungeons and Dragons.
RACHAEL: I'm going to insert something here that I think is important that the kids don't know about twinks. I was talking with a friend–
JESSE: –like level 19, level 29 PVP. Is that what you're talking about?
RACHAEL: Kids don't know about twinking. I was talking with a friend and he was like-- anyway, long story short, I am also a person who has in my life played some MMOs. And I was like, well, I don't know. If I started a new MMO, I'd probably want to be twinked by a guild. He was like what?
JESSE: So I did-- I had this whole call with my friend Tanya that will come out and-- it will have come out previous to this. And we talk about World of Warcraft because she's playing Classic and she builds Shiny dashboards to help her guild allocate resources. But then we were talking about how we play. And I was like, I can't believe I have to say this. So when we're talking about twinking, we are talking about playing at the max level of a bracket in PVP. So yes, has that disappeared, or do people don't do that?
RACHAEL: I don't think the term is there. I don't know it was wild I have to like--
JESSE: So that was a really-- that's so funny because that's the direction I'm going with this is like it's a good way to play a game and pause your experience and just understand what's happening and get your bearings. And so I felt like I've been invited on a couple of alternatives to DND. And I've stopped because they're like, OK, here are all the things you need to do for our first meeting. And I'm just like--
RACHAEL: Homework. Thank you.
JESSE: I don't understand. What do you mean pick things? So what advice do you have for people who are interested in maybe this world or some other nerdy hobby like an MMO that seems really overwhelming and they want to do it but they don't know where to start?
RACHAEL: Great question. I also thought about tabletop because it sounds like you've covered MMOs recently. Yes, so the number one thing I would say is think about what you want out of the experience. So there's a couple different flavors of groups and everyone's going to have a bad time if you show up to a group expecting a different flavor of group. So I would say the first flavor is like people who are really interested in collaborative storytelling. They really want to do roleplaying. They are theater kids.
JESSE: Both of us, guilty.
RACHAEL: Yes. So that is one style of group. Another style of group are actually these are of two ends of the spectrum and most people who fall somewhere in the in-between, are people who are really interested in the mechanics. So these are people who might have been game players, maybe they do like tabletop figures, like Warhammer.
JESSE: They do hand-painted-- Warhammer.
RACHAEL: Maybe like they really enjoy building characters that have the best possible statistics. Maybe they really like theory crafting. And I'd say that those are like the two ends of the spectrum and most people are going to fall in one of those. And so the first thing that is like if you're interested and you have maybe a group or a specific person is asked you like, hey, you want to come join or someone's trying to start up a new campaign like ask about that, like what is the focus? Are you really interested in more of a combat-driven system? Are you really interested in collaborative storytelling. So that's a big first thing because there's different ways to do it.
So I particularly learn DM thing. DMing for YouTube. Dungeon mastering. I also hear DMing game mastering. That's the person who plays everyone who is not a character and controls the world and everyone who's not a player character and controls the world. And usually does most of the math offscreen. So make sure you're aligned.
And also, it's OK to make mistakes. Finding groups, it can be challenging. I will say one nice thing about particularly the last couple of years is that there's more and more folks who are willing to play online as opposed to going over to someone's house, which helps with accessibility and is nice. I think that's really the main challenge is finding the group. And I would also recommend don't have a group of people who are all completely new to tabletop roleplaying games. You want at least one person, probably the GM who has been familiar with this.
And just know that there's a lot of variety out there. I don't know, this is all very generic advice. But if all you've seen is-- let's say you've watched an actual play, if all you've done is watch Critical Role, the popular actual play, know that there is an enormous variety. There's a bunch of systems other than Dungeons and Dragons. Dungeons and Dragons is a very rule-heavy system. So it tends to draw on people who are more interested in the mechanical, technical aspects of things, where something like Dungeon World, for example, is or totally real human adults. One page systems tend to be much less mechanically and more focused on the roleplaying.
And there are infinite, not infinite, but many, many roleplaying systems. And if you are interested in playing, I think interacting with or consuming an actual play can be a really good way of seeing how roleplaying works. So I mentioned Critical Role, very well known. Some of my favorite actual plays are actually not Dungeons and Dragons. So I really like One Shot Network so they have a show called One Shot, which is a one shot is when you complete your play in a single session. A session is usually a couple of hours. So are totally human adults is a workshop.
They also have a couple long-running campaigns, so they have Sky Jacks is the current one. Before that it's great. It's explicitly anti-colonialist, high fantasy, sky ship. It's a hack of a system designed for Star Wars. So like this airships used to like spaceship mechanics. But it's also very, very storytelling-focused. So those are two good ways to dip your toe in and also just like the One Shot Network, James is a great DM.
And then I also like Spout Lore, which is a dungeon world podcast from a bunch of comedians. And I will say it's very enjoyable if you have a high tolerance for dumb jokes. And also it's a very good story. It's very well-edited. I really enjoy listening to it. There are some dumb jokes and just be prepared for that. They're dumb jokes in One Shot stuff too.
And so those are two that I recommend that. I just like as listening material. And if you-- once you find an actual play that you like, one good way to find people to play with is to support them. Usually they'll have a Patreon. Usually that includes a Discord. Usually one of the things that we will do on this card is organize game. So if you're like, I have no idea what system I want to play, find an actual play that you like listening to because usually the people who enjoy listening to the same actual plays that you do are also going to share your goals on that spectrum, is probably going to be pretty close to whatever the show is just generally, a general rule of thumb. So that was a very long-winded answer.
JESSE: No, this is perfect. You brought it home because I was like but how do you find the people, Rachael. That's so smart. And I think that you could probably extrapolate that to anything. Find the people doing the thing that you enjoy and then figure out how to get into their Discord or Slack or closed forums or whatever and go from there. And I think that's such a great and beautiful way to think about finding community, finding the people who like the things you like, especially in this world where we are ever connected and accessible.
And you no longer have to rely on the kids in your neighborhood to hang out, not that the kids in the neighborhood aren’t cool. I live in the suburbs with a bunch of parents like nobody's going to play roleplay games.
RACHAEL: And then you get to fight your first big boss as a group, which is scheduling for real, the real enemy.
JESSE: That is scheduling is the hardest part. But we got this scheduled, which is that [INAUDIBLE]. No tornadoes have interrupted this broadcast. So where on the web-- you are all over the web. Where are the best places on the web for us to find you?
RACHAEL: Great question. At one point in the past, I would have said Twitter. So I am @rctatman most places. It's actually YouTube just introduced handle, so youtube.com/@rctatman. That's my Twitter. That's my website. My Mastodon is rctatman again, @rctatman.com.
JESSE: Because you rolled your own server.
RACHAEL: I did.
JESSE: You're so fancy.
RACHAEL: Like we said, I'm a little bit leery of putting all my eggs in one digital basket as it were. And if I can be the one holding the basket, I don't always want to do it. But sometimes it's nice. And then probably the easiest way to keep up with my content is I have a YouTube channel. I have a Twitch. And I also have a newsletter, which I'm going to try to be a little bit more consistent about sending out. And that is at-- I think it's TinyLetter rctatman. Let me double-check. I'm pretty sure it's tinyletter.com/rctatman, sure is. So if you want to sign up for my newsletter and get a very occasional email, roughly monthly, that I remember.
JESSE: I love it. That's perfect.
In closing
Thanks for tuning in to our latest episode of Data + Curiosity! I’d love to hear your thoughts on this interview, how you’ve gone about finding your community, and what you’re curious about! Let me know in the video comments – I can’t wait to hear from you!
Show notes
📽️ Jesse chats on Rachael's channel - https://www.youtube.com/watch?v=g-SLqXYTtIw
🧑⚖️ GDPR - https://gdpr-info.eu/
💙 Kaggle - https://www.kaggle.com/
💻 Dev.to - https://dev.to/
🤗 Hugging Face - https://huggingface.co/
🚀 Attention is All You Need (Transformer paper, Vaswani, et. al.) - https://arxiv.org/abs/1706.03762
👏 CLAP (Contrastive Language-Audio Pretraining) paper - https://arxiv.org/abs/2211.06687
🕊️ Fall; or, Dodge in Hell by Neal Stephenson - https://www.nealstephenson.com/fall,-or-dodge-in-hell.html
🗺️ Zooniverse - https://www.zooniverse.org/
🌌 Galaxy Zoo - https://www.zooniverse.org/projects/zookeeper/galaxy-zoo/
💻 Association for Computational Linguistics (ACL) - https://www.2022.aclweb.org/
🧠 NeurIPs - https://nips.cc/
📢 UC Strike - https://www.nytimes.com/2022/12/16/us/university-of-california-strike-deal.html
🗣️ GPT-3 - https://openai.com/blog/gpt-3-apps/
🎲 Totally Real Human Adults (game) - https://crlegge.itch.io/totally-real-human-adults
🕵️♀️ Totally Real Human Adults: Imposter Syndrome (recording) - https://www.youtube.com/watch?v=iTRmOeU48MU
✨ Critical Role - https://critrole.com/
🎧 One Shot Network - https://oneshotpodcast.com/
🪄 Spout Lore - https://www.spoutlore.com/
Follow Rachael online
🦜 on Twitter - https://twitter.com/rctatman
🐘 on Mastodon - @rctatman@mastodon.rctatman.com
📽️ on YouTube - https://youtube.com/@rctatman
💌 in a newsletter - https://www.tinyletter.com/rctatman
💜 on Twitch - https://www.twitch.tv/rctatman
☕️ on Ko-Fi - https://ko-fi.com/rctatman/
.png)
