Porter Olsen: It’s my great pleasure to introduce Darius Kazemi. He began his career as a game developer, which he did for ten years and worked on massively multiplayer online games including Lord of the Rings Online, which I played a great deal. But of course if it said Tolkien I was interested to at least check it out. Then he decided to write an article titled “Fuck Video Games” and that was the end of that.
But he pursued his career as an author. He’s recently released a book on Jagged Alliance 2, a turn-based squad strategy game. He’s perhaps most well-known for his work in the “weird Internet,” specifically bot-making and a program that he wrote called Random Shopper, which goes to Amazon and systematically spends fifty dollars a month buying him stuff which you actually buy, right? [To Darius] And then it gets shipped to you and you see what you get.
He also writes Twitter bots including a museum bot that goes to a collection and randomly shows you some of the content that is not normally seen in the normal course of visiting the museum. I’m very much looking forward to the time at which Random Shopper buys for you Jagged Alliance the book. Thank you for being here. We’re excited to hear your talk, and I’ll turn it over to you.
Darius Kazemi: Hi, everybody, I’m Darius Kazemi and I hope I can live up to the abstract that I sent in ever so long ago. I do want this to be a dialogue and not just me up here so I do have a talk prepared. It it not going to fill a full hour because I expect to talk to you all while we’re here.
That was a pretty good introduction. As Porter mentioned I spent the first part of my career in video games. I’m actually going to talk about that a little bit because it relates to what I want to discuss today.
I make a lot of stuff. This is the projects that I released in 2013. Actually a few more than that. This is like 18 months of projects, and each one of these is—my book that took me a year to write is one entry on here, and little projects that took me an hour to release are also listed on here. I consider them all of equal value because I firmly believe that there is almost no correlation between how much work into something and whether or not it will make an impact, people will like it, that sort of thing. All I know is that I like all this stuff.
I’m here at MITH today, and I wanted to talk a little bit about digital humanities from my position as an interested outsider. I’ve always kept a finger in academia, at first through game studies and people studying video games, and more recently through electronic literature and those fields. I’m not going to go into a “what is it?” debate because I know everyone who’s in digital humanities is very tired of those, but we know when we see it, right? But I do want to talk about what an outsider thinks digital humanities is, which is probably best encapsulated by the Ngram Viewer. It’s this sort of Big Data approach to the humanities that has been written about a lot, and people have written about why it’s bad from a cultural and institutional critique and those standpoints, and how it’s tied to Silicon Valley trends, tied up with the corporatization of the university, and so on and so forth. I’m not that interested in covering those angles, but what I would like to bring to this for a bit is my own experience using these sorts of analytical tools onto the at least somewhat humanistic enterprise of the video game.
I spent five years working as a data analyst for online worlds. I started working at Turbine in Boston. That was in 2005 and we were just beginning to understand what we had on our hands in terms of data. When I started working there on Dungeons & Dragons Online, we had a server that was collecting data about what our players did on the servers, but we had no reporting on it. No one was using it. They were just these databases that got slowly filled over time, until occasionally some admin would go and write a script that would blow away anything over six months. That data wasn’t even engineered for the purpose of tracking player behavior. It was mostly engineered by engineers who in the course of their bug-fixing decided it would be helpful to write a tracker to do this, or a tracker to do that. By the time the game was launched, we had all these trackers that were collecting data that was meant to be technical, but could also be analyzed to learn things about the way that players acted.
But no one was looking at this information at all. I learned about this server and I was this enterprising 22 year-old with an engineering degree and a minor in rhetoric, so I thought I could do anything. I figured that I could build software that turned that data into dashboards of information that we could then act on. So I did that. I was working as a tester at the time, but I got permission to do this project on the basis that it would save us money and time testing if we could outsource a lot of our testing to the players themselves. We could find out that people are wiping out in a particular dungeon if we get statistics on whether or not people are succeeding in a dungeon.
This was a huge success internally at Turbine. In an 18 month span, we went from no formal reporting apparatus at all to a team of four people dedicated almost exclusively to collection and interpretation of data, with myself as the chief interpreter. I like to say that we created this culture of data at Turbine. We went from a company that didn’t talk about these things to a company where designers would come to me before they even had implemented a new idea that they had, and said, “I have these concerns. What can we do to track data about our players that can then tell me once this feature goes live whether or not it was a success?”
And this was new in the industry at the time, as well. I believe there was pretty much one other company that was doing data analysis at this level, and that was EA on The Sims Online. Most of the other companies were simply scraping log files and occasionally looking at numbers here and there. But at best a company would have a single engineer who would do this part-time because it seemed like a good idea. So this was all very new. This was about four years before Zynga really popularized social online games and collecting metrics and tuning their game based on that, to the point where I had to go out to conferences and try and convince people in video games that it might be a good idea to collect data on their users.
Virtual worlds are really interesting for this kind of work because in a virtual world, you’re presented with what seems like this closed epistemological system, where in theory if it’s happening in the simulation, you technically have perfect data on what’s happening within the parameters of that simulation. It seemed like we were solving epistemological problems inherent to user studies and play-testing. We no longer had to deal with all that messy “Well are they really playing like they would if we’d brought them into the company? Are they saying nicer things about the game because we bought them pizza?” All the way down to hairier problems than that as well. So it seemed like we were doing this very scientific thing.
What I’ve learned from that experience is that using data can get you very far…until it doesn’t anymore. One of the hardest lessons that I learned, and I learned it slowly over the course of three years, is that you’re almost always confirming suspicions that you already had. It’s unclear whether you’re coming in asking biased questions, and in fact even the very nature of the statistics themselves can be called into question as to whether or not they’re accurate or mean anything.
We built all sorts of visualizations. This was from my consulting company that I had after that. We looked at people playing Quake and built these beautiful visualizations of how they did it, and you could look at that and make conjectures about it. Next up is an example that I don’t think I’ve shown before. This is a capture from a tool that I built internally at Turbine. This isn’t even live data, this is from our beta server. I built this tool called the Economy Explorer, and we used it to look at trade networks in the game and sort of ferret out who the gold farmers were, and who’s doing really strange things trading with a bunch of placeholder characters and then figuring out who the mule is and cutting them out of the network, and all that kind of stuff.
[Here Darius plays a clip ~11:12–11:55 in the video showing the application.]
It was very exciting work, but this was the building of a tool to help an authoritarian apparatus in the game, essentially. This was the tool that we used to ban people for doing things that we didn’t like in the game.
When it game down to actual insight and figuring out fundamental things about our game and learning about it, I used to get this question at conferences all the time and I always hated it, which is, “What’s the most surprising insight you uncovered through your data?” I never had a good answer for this question. Usually I’d default to something like “Well the big surprise was that there were no big surprises.”
And yeah, we’d uncover unusual stuff but it usually just mystified us. We were smart people, I had a strong humanities background in addition to a tech background. I knew things, I had cross-disciplinary knowledge, like I thought I was smart enough to slice and dice it and figure out what this stuff meant, but the more I sliced and diced it, the more it was like infinite regression of nesting dolls. You just open up this box and then there was another box, and then you open up that box and another box. It’s probably a human head at the very bottom.
We’d uncover things like a whole bunch of people are buying this particular kind of item. Sometimes we’d look at it and at best we would notice that the item’s pretty over-powered for the price. Well great, that’s useful, but it’s not surprising. It’s not an insight that our players are optimizing for the systems that we told them to optimize for. Over and over and over again, it just told us “things are going the way you probably expect them to go, but maybe they’re going worse or maybe they’re going a little better” and that was pretty much it.
People would want to know about money and purchasing habits, and that’s what we actually got out of the realm of the epistemological shield that we had set up because now you’re dealing with humans with credit cards, and there were privacy laws and all these other things. So I sort of said okay, I’ll do the business analytics on that side because I have to for the business people, but on this other side is the game stuff looking at the game as a work on its own merit, and I can still slice and dice that and learn things. Over and over again we either learned that our suspicions were founded, which either meant that our suspicions were always right or that the way we were collecting this data was just set up to confirm our suspicions, or we found weird things. For example, oftentimes when we’d find something weird it was like this story I’m about to tell.
So we have monsters running around this world, because of course you’re not going to have a beautiful virtual world without monsters getting in your way all the time. We put in a few hooks to the artificial intelligence of the monsters because some of the engineers were worried about the pathfinding for the monsters; that they would get stuck in certain places. So essentially a monster would walk around, and at any point in its program if it didn’t know what to do next and said “I lost my marching orders. I’m completely lost. I have no idea what to do.” it would ping out those coordinates and it would just say “Okay, I got confused here.”
So said I could make a map of all these places the monsters get confused and maybe I could find some weird place where the geometry is bad and needs to be fixed in the game. So we pull up the map, and the map looks almost like an exact grid, like all these dots but the dots are all clustered along these lines making this grid. I had no idea. I thought there was a bug in my code that collected this data or something. I took it to a much more experienced engineer who took a look at it and said, “I know exactly what that is.” Our system is divided into land blocks, and the AI just gets confused when it’s basically crossing a latitude/longitude line into another parcel of the server. It gets momentarily confused and then reorients itself. Invisible to a player, but got sent out to the server.
So that’s weird and sort of interesting, but it reminds me of this. Anyone who’s ever searched for “fuck” on the historical ngrams site, you probably were mildly surprised to note that there’s a lot of f‑bombs being dropped in 1650, until you go to Google Books and actually search for that word in that time period. Then it turns out that it’s just the word “suck” or sometimes “such” depending on the font, and the OCR picks it up the wrong way. That reminds me of finding that weird grid pattern with those artificial intelligence entities in the game. It’s equivalent to those AI seams.
This information often tells us more about the software platform collecting the information and the apparatus that we’re using than the actual content that we are concerned with, which is fascinating to me because I like meta stuff and I like software platforms and all that sort of thing. I think that’s great and interesting and I would read a paper on OCR errors in Google Books and what that means. But yeah, data analysis often tells you more interesting things about the methodology than the subject.
So we ended up with these massive epistemological problems anyway. We weren’t avoiding them at all, we weren’t being tricky or sneaky and finding some secret way to have an objective view of what we were looking at. We were just creating new epistemological problems. We were just moving things around.
I often equate data analysis with critique as a practice. They’re both based on decomposing things, breaking them out into parts, figuring out how they work. I want to talk a little bit about this idea of composition over critique. When I say composition, I mean this in the sense of Bruno Latour’s “Compositionist Manifesto” published back in 2010. When Latour talks about composition, he sets it up not in opposition but sort of as an alternative to critique. He comes from a science and technology studies background, so he’s an STS guy, and his point is that critique only gets you so far. The way he puts it is if you have a sledgehammer—not a regular hammer—or a wrecking ball, there’s pretty much one thing you can do with that, and it’s super useful for clearing out old debris and all that sort of thing. But what happens when you’re done with that? I’ll quote him at length here. He’s talking about the difference between critique and composition. He says:
The difference is not moot, because what performs a critique cannot also compose. It is really a mundane question of having the right tools for the right job. [The sledgehammer can] break down walls, destroy idols, ridicule prejudices, but you cannot repair, take care, assemble, reassemble, stitch together. It is
no more possible to compose with the paraphernalia of critique than it is to cook with a seesaw. Its limitations are greater still, for the hammer of critique can only prevail if, behind the slowly dismantled wall of appearances, is finally revealed the netherworld of reality. But when there is nothing real to be seen behind this destroyed wall, critique suddenly looks like another call to nihilism. What is the use of poking holes in delusions, if nothing more true is revealed beneath?
Bruno Latour, “An Attempt at a ‘Compositionist Manifesto’ ”
This in turn reminds me of something that I read earlier this year in the digital humanities milieu from Mark Sample. He wrote a blog post called “Difficult Thinking about the Digital Humanities” and I’ll quote him as well here. I’m sort of taking him a little bit out of context, but I hope he’ll forgive me. He says
Two hallmarks of difficult thinking are imagining the world from multiple perspectives and wrestling with conflicting evidence about the world. Difficult thinking faces these ambiguities head-on and even preserves them, while facile thinking strives to eliminate complexity—both the complexity of different points of view and the complexity of inconvenient facts.
Mark Sample, “Difficult Thinking about the Digital Humanities”
He’s sort of drawing a line between that facile thinking and critique, and he says, “I’m dissatisfied with that word “critical” and all its variations—that’s why my formulation emphasizes difficult thinking over facile thinking.” I don’t think it’s a coincidence that at the same time that he wrote this essay, he was also in the middle of an incredibly prolific period of writing about and building Twitter bots. He wrote this great essay about bot archetypes where he talks about closed bots versus green bots. @everyword is a great example of a closed bot. It takes a dictionary and it iterates through the dictionary and it tweets every word in that dictionary, and then it’s done. That’s a closed bot.
A green bot is something that might source from the greater world. Sample’s bot @whitmanfml is a great example of that. It takes quotes from What Whitman which are a closed corpus and then mixes them with the #FML hashtag, source from Twitter which is an open corpus, and that’s what he’s talking about when he talks about green bots and how green bots have this capacity to surprise and all these other things.
I don’t think it’s a coincidence that he was thinking about these things at the same time. Another great example of a bot that he built is more of an activist bot, sort of the equivalent of— It’s activist in the sense that a Brechtian play is activist. He calls it an experiment in speculative surveillance, and it comes up with all these tweets about who NSA PRISM has caught doing what. If you follow this, it keeps you on your toes, essentially, as you follow it through your feed.
What I want I want to challenge people in digital humanities to think about is how you can reverse the polarity, so to speak, on the tools of digital humanities. If you’re learning things about the way texts are constructed or how they’re situated, that kind of thing, and you’re learning these things through analysis be it through data like ngrams or anything like that, by all means publish the results of that statistical analysis or visualizations or whatever. But I think it’s a really important exercise to also try turning that research on its head and using your findings to generate things that go out into the world.
A really enlightening example for me is the regular expression. For those who don’t know, a regular expression is a computer science construct, it’s mathematical construct, it’s a linguistic construct. It’s one of these great things that sort of cuts across all sorts of disciplines, and you can write mathematical proofs that a regular expression is also a state machine from this other field, and it’s also isomorphic to this and that. But for people in digital humanities especially you might be familiar with the regular expression because it is one of the most common tools for extracting text from other text.
So if I want to find every set of comma-separated phrases in a book, I could write a regular expression that essentially says find a block of letters and then a comma, followed by an arbitrary number of blocks of letters and a comma, followed by maybe an ‘and’ maybe not if it’s an Oxford Comma or what, and then terminating in a terminator like a period or an exclamation point or anything like that. (I’m a programmer; it took me ten years to really internalize regular expressions.) You feed that in, and then you feed in the book, and then it spits out on the other end the matches for all that. You can also use regular expressions to find and replace things. So if I want to find all of those comma-separated phrases and replace them with the word “meow,” I can do that very easily with a regular expression.
What I hadn’t realized in ten or twelve years of using regular expressions is that they also can be turned on their head. You can use a regular expression as a generative tool. There are tools out there where you can put in a regular expression that might say— If you wrote a regular expression that said “find every instance of two vowels adjacent to one another” you could also feed it into a machine that says “given the rule ‘two vowels adjacent to one another’ generate every string of characters that validates that rule.” and then it would spit out a few hundred vowel pairs for you. You could get weirder than that. You can feed in sets of syllables and you could have it generate whole new languages and that sort of thing based on this tool that is normally used for analysis and decomposition and taking things apart. You can turn it on its head and make it into something generative.
I want to show another example here. This is on my mind right now because this is something that is happening at the moment. I organize this thing every November called NaNoGenMo, and it’s like NaNoWriMo, which is National Novel Writing Month where people sign up and they pledge to write a 50,000 word novel over the course of the month of November. Last year I posited that we could do NaNoGenMo where you’d pledge to write code that generates a 50,000 word novel over the course of November. So we have a bunch of people who sign up and do this. Academics, programmers, people who are neither. We have 61 projects going right now. Some of them are already done. This one’s great; it’s called 50,000 Meows, and you just feed it a book and we can get the new all-meow version of Moby Dick. Much improved, I think. It retains the syntax and capitalization and all that sort of stuff. So it still has the shape of Moby Dick.
I’m working on something where I’m looking at people who have studied plotting for novels and reverse-engineering their analysis of the way plots in novels work to generate something approximating a novel and part of what this does is it points out the holes in their model as well. So it does sort of operate as critique as well. I pulled all this stuff from one of these “how to write a novel for a beginner” things, and this person has the “ten steps” of a novel. You start with status quo, then something happens, then the character makes a decision to act. So I read through this whole thing and took it extraordinarily literally, and I built a short first pass of this, which is
If you’re interested, there’s a resource thread where people post resources they think might be useful, anything from tools for natural language processing to corpuses of data that they think might be interesting. I learned about someone who compiled 10,000 plot elements that could possibly be in a book, so I’m probably going to dig into that pretty soon as well. To me as an outsider, this is all digital humanities work as well. We’re analyzing the form of what’s out there and then internalizing it. In this case we’re writing an algorithm based on that, and then spitting it back out in the world.
This is the same thing as when I did @twoheadlines. This was me looking at a lazy form of joke on Twitter and going, “I could parameterize this and just generate this joke forever.” I don’t know if Mark Sample would consider this an activist bot, but I consider it activist for a very narrow band of cultural activism where I want to kill a particular kind of joke. And I’ve sort of succeeded. On a weekly basis, someone tells a lazy joke like this and then someone else rather innocuously goes, “Oh that’s just like @twoheadlines!” And I’d like to think that they look at it and feel bad.
Much of what I do is just noticing patterns in how people communicate in digital media and figuring out how to replicate that and send it back out in the world. When I see people doing just analysis, I’m always holding my breath waiting for the next step, which is what are you going to do with that? What are you going to make with that? Analysis for its own sake is fine, I like analysis for its own sake, but at that point that I’m waiting for someone else to take that and see what they’re going to do with it.
I think that’s it for the planned part of my talk here. I’m interested in opening it up to questions or comments or anything.
This is just a little bonus. I built this this morning while I was working on my talk because I accidentally built it. This just generates a random ngram. So if you want to do some analysis, there’s a tool for that now. This was sort of inspired by this great web site Spurious Correlations. This algorithm has a zillion different data sets and finds ones that correlate well, and then it says, “Well here you go. Drawn your own conclusions.”
Audience: But you’re not actually looking for sequences that are correlated, they’re just…
Darius: No. They’re just random words from the dictionary, but sometimes you’ll notice that “polarization” and “folklore” are pretty well correlated, and what does that tell us about the history of the written word?
Audience: That’ll be for your next talk.
Audience 1: I’m wondering if your novel generation algorithm could be forked and turned into a scholarly article generator.
Darius: There are already tons of scholarly article generators. That’s a well-trodden path. It’s usually people from the hard sciences, who create generators that generate sort of deconstructionist tomes and then submit them. [cross-talk comment from audience member] Random post-modern stuff, and then submit them to actual journals and see who they can get to actually accept them for publication. And this has happened. But it’s also gone the other way. There are scientific papers as well that have been equally, sadly, accepted by journals, too. You should Google it. There’s a lot of good stuff there.
Audience 2: My day job is over in the English department, so I’m going to try to ventriloquize a question from a kind of cranky senior colleague who, were such a person sitting here, he or she might point out that if you look at Literature with a capital “L,” the kind of literature that we expect to do real cultural work. You look at titles and get things like War and Peace, you get Crime and Punishment. These are the big questions, right. And the projects that you’ve shown, one of them is about killing a certain type of joke on Twitter. How would you address a kind of cranky skeptic who sort of wanted to know something about the potential for this form to address the big questions, the human condition?
Darius: I’d show them [Last Words] to start with. This is occurrences of the word “love” in last words of executed Texas death row inmates. This is a corpus that I got my hands on. Someone just pointed me to it, and this is one of those 45 minute projects for me. I didn’t even set out to build this, I just said “I have this weird set of data.” I built software that scraped out all the last words. I wrote a program that just put the entire text of people’s last words in front of me and just sort of flashed it by like flash cards. I noticed the word love appeared many many times. I said “what if I filter for the word ‘love?’ ” It came up with this, and I said, “I’m done. That’s it. It’s done. Ship it.” That was over lunch at work one day. And this has been used by actual anti-death penalty advocates. They use this in their workshops. A Christian newspaper wrote an article about this and Christ and the death penalty, and can Christians really support the death penalty, that sort of thing.
So maybe not NaNoGenMo, yet. I don’t think we’re going to have War and Peace, but I do think there’s plenty of potential to address larger issues.
Audience 2: And one version of that question is kind of this cranky hypothetical colleague, but another version is the concern that a lot of what we often see is either derivative or parasitic in the sense that it’s a kind of in-joke if you will. So the 50,000 meows is funny, but because we have National Novel Writing Month, so you already have to be in on that as a kind of Internet meme or phenomenon in order to then appreciate this wonderful derivative work, and that’s a kind of barrier to entry.
Darius: Right. NaNoGenMo itself is my loving jab at NaNoWriMo because it is very strange to have an event where the definition of novel is “50,000 sequential words.” So yes, I do agree there’s very often a bit of a barrier there. On the other hand, I think you could make the argument that what I’m seeking to do— It’s that whole public intellectual versus university and publications and that sort of thing type of debate. I’m interested in deploying things to public networks like Twitter and Facebook and the Internet at large, where people are more likely to laugh at a reference to NaNoWriMo than they are to laugh at a reference to Foucault.
Audience 2: But ultimately the same sort of underlying message…
Darius: Yeah. I think it’s a lot of the same work that’s being done. And much like with the debates over the public intellectual and their role, they can go on forever as well.
Audience 3: How do you see the field of software criticism or cultural criticism through software evolving? What’s next?
Darius: I don’t know what’s next for the evolution of cultural criticism or opposition through software. I see more and more work done with— I was talking earlier about this today. In my head for a long time, I’ve had this divide between software that does work like utility software, and software that does art. You have utility bots that you can tweet at a bot and say “Remind me at 12pm to have lunch.” and then it would tweet at you at 12pm and say, “Have lunch.” That’s a Twitter bot. It’s not really an art bot, I don’t think, but maybe you could make a case for it. As opposed to my Metaphor-a-minute, which just cranks out metaphors all the time, which is much more situated as an art thing.
But it’s strange because there are also activist bots. I’m hosting something called Bot Summit on Saturday [November 8, 2014] in Boston, where a bunch of practitioners are getting together both online and in person to talk about this sort of stuff. One of the things we’re going to talk about is this whole question of activist bots. There’s a strange tension between activism through bots and the terms of service of places like Twitter, which don’t let you unsolicited reply to people. So I’ve had people say, “Can you help me make a bot that finds people tweeting bad facts about climate change and gives them the data to help them come around on it.” And I tell them no, I can’t do that because that bot will get banned automatically by Twitter as spam, because technically it is spam. They did not sign up to get harassed by an entity for things that they are tweeting. So it makes it very hard to do activist work on Twitter, but within activism you have Mark Sample’s NSA PRISM bot, which is activism. It is also art. I drew a connection to Brecht on purpose there. It’s possible to do art and activism at the same time. It’s also possible to write something on a sign that says “Down with So-and-so” and go out in the street, and that is activism as well, and that’s not art.
So I’m very interested in the future of activist bots and activist software. There’s something that someone built called Block Together, where you can as a community come up with a blacklist of people who are toxic to your community. And then everyone in your community can authenticate with Block Together and press a button and simultaneously block all 5000 of those people. So members of your community can be protected from having to deal with members of another community they might find toxic. But it stays within the bounds of Twitter’s terms of service. That’s very interesting work as well.
In terms of trends, I’ve noticed a lot more visual things happening on Twitter recently. There’s ⋆✵tiny star fields✵⋆ by Everest Pipkin, which has 20,000 followers, blows away any of my creations, and just tweets these beautiful, serene little visual compositions. And I love @ARealRiver, which tweets individual things that look like a sort of scene by a river. As you scroll down you’ll notice it’s continuous, and you can just scroll forever and follow this river. There’s a boat in there, there’s a horse. If you go all the way down, you’ll hit a mountain range that persists for quite some time, and then there’s metropolitan areas it’s passed through. It’s a persistent landscape on Twitter. It’s very interesting and it’s very exciting to me. So I’m noticing a lot more visual stuff playing with emoji as well, but I can’t predict the future. Hopefully I get a better sense when we have Bot Summit this weekend.
Audience 5: I wanted to ask you about the term “Weird Internet.” Can you give us the background on that term, and also talk more broadly beyond Twitter where you see the weird Internet?
Darius: Sometimes when I say Weird Internet, some people thing I’m talking about Weird Twitter, which is its own subset of surreal Dadaist humor that happens on Twitter. But when I say Weird Internet I mean that very specifically. I’m talking about almost a return to a folk Internet where there’s a place for things that aren’t just sanitized of your birthday party that your grandmother can click Like on. People have recently referred to it as like the return of Web 1.0. There was an article recently about that. I don’t know if you’ve seen tilde.club but you sign up—there’s a huge waitlist now, like eight thousand people or something on the waitlist—but I think Paul Ford set this up and it’s literally just… I don’t know if universities do this anymore, but I know they did when I was in school where it’s like, “Okay here’s your shell account, and here’s the Unix server and your username is blahblahblah.edu/~dkazemi” and then you can upload whatever you want on there and host it on there. This is an attempt to recreate that, and if you go to people’s sites […] it’s got this lo-fi look to it. But beyond that, it’s got a lo-fi apparatus as well. One of my friends was saying “I have friends who are completely not technical asking me for help with ssh because they have a tilde.club account and they want to update their site.” I consider this Weird Internet, I consider bots, especially non-utility bots to be Weird Internet. Anything that makes you stop in your tracks or surprises you at all I would consider part of that milieu. I just want strange and unexpected beautiful things to happen all the time.
Audience 6: I don’t know this for a fact, but it was my impression that that started out as a joke kinda similar to how some of the things you talked about have a comical aspect to them, but Paul was quickly overwhelmed by all these people that wanted to use that site. There were some people that I think were part of the joke, like Matt was talking about being on the inside—
Darius: Right. Some of it’s almost like an ironic, hipster, “I’m gonna use ssh in addition to building a bike out of my woodshed.”
Audience 6: But then there were all these other people that got in on it because they actually liked the idea of it. Like, “Actually, no, I don’t mind ssh’ing to this computer and creating an HTML file and having a web site there.” And people that are rediscovering that that was the way that the web started. There’s a new generation of people now that are [observing?] this is actually how it began.
Darius: Myself, over the last few years I’ve been slowly extricating myself from blogging platforms like WordPress. I archived my WordPress site and it exists in static form now but I can’t log in, there’s no database or anything like that. Instead if I want to create a blog post, I open up a text editor and I type “” and then I start like I used to when I was in 7th grade and making little web pages. So now my essays look more like this. I just write it all by hand in a text editor.
But it also gives me the freedom to do things like embed software and make it way more interactive than it otherwise would’ve been because on WordPress it’s like, “Well can I find a plugin that does the thing that I want? Do I have to write a plugin?” As a technical person, this gives me more freedom. But I think also even if you’re not that technical and you’re just learning ssh, it can be exhilirating to have— I remember when I was a teenager I think, and I first learned that I could log into a computer remotely somewhere else, log into my home computer from someone else’s computer and get files off of it, that was this incredible, exhilirating thing for me not just because of that connection, but also it was this thing that I controlled. I could literally install software on my home computer from someone else’s machine, and that was very cool and it felt empowering at the time, and I think you get a whiff of that by joining something like tilde.club as well, even though it’s more managed. But it’s certainly less managed than Facebook, for example.
Audience 6: I guess what I meant to ask was have you noticed this sort of arc in some of the work you’ve done where it started out with this initial purpose of doing something maybe with a comic value or whatever, but then like with tilde.club it took on this other meaning. Like you started out with a particular idea but it sort of morphed into something else that had more use than you anticipated.
Darius: Definitely. A great example of that is Museum Bot. I came home from work one day and my wife was already home, and she said, “Hey, I don’t know if you saw, the Metropolitan Museum of Art released 400,000 high-resolution images along with metadata about its deep, deep back catalogue.” And I as like, “That’s really cool!” And she was like, “You might want to do something with that.” And I was like, “Yeah, that would be cool to do.” And I sat there for a half an hour, forty-five minutes, looking at their collection and figuring out what I could do with it, and I couldn’t come up with anything good. I was just like, “I can’t think of anything interesting. But you got me all excited so I guess I’ll just build the most boring possible degenerate case.” Which is it just grabs something at random and tweets it.
At the time I was making it, it was a throwaway. But because it’s this autonomous agent that goes out into the world and is sort of consistently surprising. The Met’s collection often has, I mean you know, “Petals of a Composite Cornflower Pendant,” that’s weird; An unlabeled just…vase; a drink tumbler, like for your whisky. I would not have expected to see this stuff in the catalog at the Met. People really liked this, and it also says things about curation. People can look at this and look at what’s on display at the Met and go, “Oh. What’s on display at the Met is nothing like what a true random sampling of items in the Met’s catalog is.” So people start to learn things about curation by engaging with this bot, and then for some people it’s almost a status thing. It’s like, “Oh, I’m signing up for culture.” It’s like listening to NPR so you can tell your friends that you listen to NPR. “Have you heard the latest Serial episode? Come on.”
So there’s a lot going on here, and I had zero zero zero idea when I built it. I built it and I had dinner, and I was like okay, I can forget about this forever. And it keeps coming back up with more and more people who are interested in it and want to talk about it and so fort.
Audience 8: I want to go back to your example where you were saying you sort of drifted away from existing platforms and started building your own stuff including blog posts and tied that to the activism stuff that you were talking about earlier. Of course, activists have really embraced the Internet including some of the weird stuff as well. But they will obviously and probably do see the value in moving off platforms like Facebook and Twitter because those are owned by private companies and they exert their control over algorithms, over stuff they censor over their terms of service, and activists don’t like that because that gets in their way. On the other hand, as soon as they want to move away they face a different barrier, which is they need to learn how to make this stuff, and a lot of them—the majority of them, probably—don’t know how to do this stuff. So what would you say is like…there’s obviously kind of a good thing about more people learning about the Weird Internet, but there’s also this barrier of having skills to build their own stuff. And also then letting other people know that stuff’s out there, because unless its on Facebook very few people will learn that it’s there because it’s just some little web site.
Darius: To the second point first, I think that’s unavoidable. I think you see that analog in just traditional activism as well. As an activist, you both need to be able to go out into a public square and protest and maybe do it by getting a permit and following local laws, or maybe not so much. But you also need a private space in your own home or a building somewhere or a public park or something, where you can just gather and do something a little more private on your own terms as well. So to me, I see that as a loose analog to running your own server and services versus going out into the world. I don’t ever want to stop putting my stuff on Twitter as a platform because whether we like it or not, that’s sort of like the public forum, or one of them.
On the education gap, I agree with you completely. I think I’m ethically obligated, or morally obligated to teach people how to do this kind of stuff, so I publish the source code for almost everything I do. But publishing source code is like whatever, it doesn’t help 99.9% of people. So I write prose tutorials, I run workshops, I run things like Bot Summit, I go speak at places, that sort of thing. That’s a personal choice that I make as a creator, to actually give people these tools that’ve helped me so much in me doing those things, because I like seeing other people’s work and what they do with it.
Audience 9: Are you not afraid that at some point there will be so many people doing Weird Internet that it’ll stop being weird?
Darius: For a long time I was the only person doing game analytics and then everyone was doing it, and I moved onto Weird Internet. And I suppose if I really feel crowded out I’ll just move on to something else.
Audience 10: I noticed that when you talk about your bots, often you refer it’s like, “Oh it took fifteen minutes, it took me forty-five minutes.” I was wondering how you wish your audience get out from that kind of comment.
Darius: I think I mention that because I consider making a Twitter bot on the same level as most other creative endeavors, so I’m a big fan of— There’s a guy who I met recently, Jonathan Mann, he does a song a day on YouTube and has been doing it for many years. He passed the 2000 song mark recently, and much like me not even close to a majority of his stuff is a big hit or anything, but [we] probably have roughly about the same hit rate, about one out of every fifteen things that we make gets some kind of response other than the usual background noise of like, “Oh that’s cool.”
I have this blog post I wrote called “Thoughts on Small Projects” that goes in detail on this. I think it’s very important as a creative person to get to a place where not only have you skills gotten to a certain level where you can create things very quickly, but also your expectations have been brought down as well. I think it’s very important to do. I spent until I was maybe 27 years old or something agonizing all the time over whether what I was doing was good enough, and I just sat on so many projects and had so many half-finished things sitting in folders. I don’t have half-finished things that sit in folders anymore because I decide that they’re finished and I just put them out there. It’s been creatively rewarding for me, and I hope that other people can take stuff away from that, too. That’s the great irony or big joke, that Museum Bot, which I made in an hour or two is more popular and I get more people saying they liked it and all that sort of stuff than the book that I spent a year writing. I’m still proud of that book, but it just didn’t happen to be what was right at the time or whatever. I don’t know.
Mostly when I talk about how quick it is, I just want to encourage people passively to do whatever it is they do quickly, whether it’s making comics or whatever. I don’t mean to showboat “I’m so fast and this is so easy,” but also I try to provide the tools to make it easy for people as well. The reason I can do Twitter bots quickly is I have a template, which is a program that I run and it just asks me a set of questions about what software I’m going to need and then I press enter and it generates a placeholder Twitter bot that just tweets “Hi” every hour. Then I don’t have to write any of the scaffolding. It’s like here’s all the scaffolding, now you can just work on the core of it. That saves me significant time on every project as well. But I made sure to publish this and I know other bot makers who use this, and it saves them time as well. So it goes back to that moral obligation that I feel to enable other people to be able to do this stuff.