#botmakers

by Tyler Callich

  bot: anthropomorphized algorithm with a UI drone: Unmanned Aircraft System (UAS)

Bots and drones are two ways technology is making itself apparent to people, and they’re apparent to us in very different ways. I’m going to be discussing our experiences of both: compare, contrast, then try to conceive of something that’s neither (or both). Bots are interactive, playful devices in our lives. A chatbot, a photo bot, a game bot, an art bot. The list goes on, but the emphasis is on small size and exploration. drones are devices that are remote and automated. A surveillance drone, a racing drone, a military drone. they’re extensions of power and sight. Bot has a cute, positive valence, while drone has a menacing, negative one.

First off, I want to admit: I like bots.

I’ve been tinkering ever since I was a little girl. Tiny gewgaws from clothespins and rubber bands. Monsters from discarded rubberbands and paperclips. Even though these are static objects, I’d like to think of them as proto-bots. For little me, a ‘‘bot’’ didn’t have to accomplish a task. I made proto-bots to entertain myself. They might flip something across the room, but that was secondary.

These things were formed from limited materials and a child’s mind.

Here’s a recreation of one of my proto-bots:

  I made this proto-bot in a coffee shop with items I had in my purse. Mint tin, chapstick, sticker, hair ties, markers, memory card case, Advil.

It doesn’t do anything except make me happy. Its name is Toothsome.

Today I make bots mostly on Twitter out of Python or Ruby scraps that I copied shamelessly. There’s a certain level of frivolity behind most of my bots, which makes them unique, fun, possibly precious:

They allow me to play online even when I’m asleep or working (ex. @thextheyandthez, a tribute to The Good, The Bad, & The Ugly)

  They uncover the distant, strange boundaries of an idea (ex. @lichmaze, a fake game concept)

  They reflect my own words back to me like an oracle (ex. @teenlich, which is made from my old livejournal)

  Bots are a form of introspection for me. They help me answer personal questions.

How can I better understand my teenage self?

Why do labyrinths fascinate me?

To what extent can I take this ridiculous joke?

What is gender, anyway?

I keep making bots because I keep having questions and experiences that are hard for me to comprehend. A bot expands a concept in serial, revealing unanticipated combinations in the process. Sometimes it’s an uncanny turn of phrase. Other times it’s a recontextualized memory or utterance.

Drones are another kind of automated creature that people make. The thought of a drone conjures up the humming of a mindless bee army. Drones are used to expand the presence of a person or government. Drones kill. Drones deliver. Drones surveil. Governments, in particular the U.S. government, use drones to do great harm abroad. Today UASes are a big arm of the military science industrial complex, and they’re only making more.

On the other hand, as a culture, we have anthropomorphized bots into cute little artificially (un)intelligent homunculi. Our hopes of making real AI give bots a certain cuteness and relatability that drones certainly don’t have, but all that spontaneity and humanity comes from the bot creator and their acceptance of error.

Joke bots & murder drones

(image source)

This place we’ve gotten ourselves into reminds me of the movie TOYS (1992). cute heir to toy empire (Robin Williams) fights against general who wants to militarize playthings.

Even though the plot of TOYS is attractive today, the dichotomy between good friendly bots/evil servant drones isn’t useful. Bots aren’t that smart, and drones aren’t that mindless. Both are cultural constructs around how we automate processes. I’m not going to try and make drones cute. I have a nightmare that people start naming drones things like “@tiny_destroyer.” Cutifying or gamifying violence/control will likely become a trend as drones and the children of drones come to the fore.

Bots aren’t all good, either (ex. @_grammar_ and @RedScareBot), and their association with cuteness and play can obfuscate when bots are doing something terrible. (take for example, what happened with Microsoft’s @Tayandyou.)

Drones work, bots play

When we create things, we are always solving a problem. We create drones to do things we’d rather not, to try and erase responsibility. We created bots to keep ourselves company online. The effects that both have make our intentions clear.

Teach a bot to love

Today, drones put our own inhumanity on display. It’s disheartening and worrisome. silly little bots help somewhat: they make me feel more hopeful about technology.

I hope that one day we learn how to code empathy like we’ve learned to code cruelty.

Maybe it will be the other way around: our tiny bots teach us empathy.

Thanks for reading!

Sarah Blaszczok interviews botmaker and musician Emma Winston, who will be performing at The Art of Bots event on April 15-16, 2016.

How would you describe what a bot is?

I would describe a bot as the software equivalent to a robot. A bot runs automated, repetitive, pre-defined tasks, which can take any form. They’re not always intelligent, just as robots aren’t always intelligent; but they do what they’re told (mostly), and can be useful for all sorts of different day to day applications ...

What are they used for?

All sorts of things, good or evil. They can be used when a simulation of a person is required, but no person is available – in tech support, for instance. They can be used for advertising (if you’re on pretty much any social network for any length of time you’ve most likely picked up a follower or two who seems innocuous enough, but click through to their profile and it turns out they’re only there to sell you stuff or spam you). Google uses bots to discover and index new web pages every hour of every day.  

The uses of them that I’m most interested in are those which are creative; using them to generate art, or poetry, or games, or jokes, or stories. Twitter seems to be the platform with the most vibrant community for this kind of botmaking.

What is it about bots that makes you want to work with them?

In general I am interested in the intersection between technology and human emotion. I think there is a lot said, and written, about the digitisation of the modern world leading to a loss of human connection and emotion and creativity and love but that doesn’t resonate with my experience of technology at all. I make (extremely personal!) electronic music and find it to be the medium I’ve gelled with the most after years of dabbling in various genres and means of creation; as a teenager my friendship group and I bonded over going to the computer room at lunchtimes and coding each other HTML layouts for our personal blogs. It was instrumental in how we shared things and communicated with each other. Both the creative potential for Twitter bot creation and the lovely community surrounding it really drew me to working with bots as a creative medium.

Can you tell us more about the bots you make?

With a couple of exceptions, most of my bots are versions of the ‘tiny’ bot, as popularised by Tiny Star Fields and Tiny Seas – that is, a scene or image which is generated by the bot using emoji or sometimes Unicode.

I like making these because it’s easy to come up with a framework and then have the bot swap variables in and out depending on what the given scenario is – for the Tiny Gallery, my most popular bot, the ‘gallery’ is fixed, but the paintings on its walls and its visitors are swapped in and out by the bot. Graphic Score Bot is an exception since what it generates is much more complex – it uses code to spit out vector images, the shapes and colours of which are the swappable variables. I am horrible at maths and found it very difficult to dictate where and how the lines should be drawn – the coding of it ended up being as improvisatory as the music (at The Art of Bots) will be.

What is the relationship between you, the bot creator, and the bot?

I’m not sure, always. One of the things I quite like about bots is that once they’re done you can just leave them to get on with things if you like. I am doing a PhD at the moment which is a huge long-term project and having well-defined, small things which I can complete and ship keeps me sane – I thrive off completing stuff.

Graphic Score Bot is the only one of my bots currently that I really want to continue developing and working on and adding to long-term – I feel like once I’ve raised the others I can let them go and do their own thing ...

It appears that most of our interaction with bots is online, and we are often oblivious when it happens. You are doing a live performance at The Art of Bots, alongside the Graphic Score Bot. Have you done this before?

I have not done this before but it feels like a relatively straightforward progression of my existing work. All of my previous experimental improvisation has taken place in a duo, Brute Love, with a friend. Having that second person there acts as a mutual anchor and a means of directing the performance and doesn’t leave you as adrift as a completely solo performance would.

Except from Adrienne So’s interview with bot maker & Internet artist Darius Kazemi.

Can we talk about "lazy humor" and why Twitter is such a great platform for it? The only thing I can think of is Dada art. The bots are funny, and they don't make any sense.

A lot of it is definitely based in this surreal sense of humor. Dada is a pretty good word for it. Dada artists are direct predecessors of all this stuff. They pull words out of a hat. That's essentially what I'm doing. With the things that I build, they get funnier the more you engage with them. I have this bot called "AmIRite Bot" that looks at Twitter trending topics. If you tweet "Donald Trump", it tweets back "More like Donald Dump, amirite?" One of the long-running jokes is that it doesn't know that "art" rhymes with "fart." There are so many times when it'd be like, "state of the art, more like cart, amirite?" The jokes are funnier if you are, what I would say, algorithmically literate. I tried to build it in such a way so you can look at it and intuit what's going on behind the scenes, and then figure out for yourself what's so funny.

It's trying to mimic humanity, except not at all.

People ask me about artificial intelligence all the time. I'm the furthest thing from that. A lot of the stuff I build is purposefully stupid. "Two Headlines" is one of my bots. A friend of mine pointed out that people on Twitter tend to lean on this lazy joke format. "David Bowie's in the news, there's a Republican primary coming up, I'm going to make a joke about David Bowie being in the Republican primary." It's easy to automate, so I automated it. It's pretty easy to teach a computer to do a stupid thing that people do all the time. It's artificial stupidity. The joke is on us, pointing out that you're not as clever as you think. The stuff you come up with is so derivative that 30 lines of code could also come up with it.

Allison Parrish is a computer programmer, poet, educator, and game designer who lives in Brooklyn. Her teaching and practice address the unusual phenomena that blossom when language and computers meet. She has built and created many Twitter bots, including @needsref, which tweets sentences in Wikipedia marked with “Citation Needed;” @the_ephemerides, which pairs computer-generated text with images of space; and her best known work @everyword, which, from 2007 to 2014, tweeted every word in the English language. Allison is currently the Digital Creative Writer-in-Residence at Fordham University and an adjunct professor at NYU's Interactive Telecommunications Program, where she teaches a course on writing computer programs that generate poetry.

VIVIAN SMING: In your recent talk, you make the analogy of exploring outer space with exploring semantic space. Could you go over that concept, and how bots play a role in discovering new semantic territory?

ALLISON PARRISH: I’m an armchair space exploration (and science fiction) enthusiast, and one of the ways I think about space is like this: our universe is potentially full of little islands of habitability. Earth is covered in places that are nice for humans to live in, but there might one day be other islands, on Mars, in the atmosphere of Venus, in caves under the surface of the moon, or maybe even on planets in other solar systems. But the spaces in between—by which I mean, >99.9999% of the universe—are totally inhospitable to human life. In order to find the islands of habitability, or just to learn more about the universe we live in, humans have to find ways to retrieve information about those places that are inhospitable. Since the invention of weather balloons in the 19th century, one of the main ways to do that is through unpiloted spacecrafts: we send robots out into the places where humans can’t survive.

So “space” is the metaphor I’m using here. In the talk, I mention that the idea of “semantic space” is used in slightly different ways by practitioners in a wide variety of disciplines, from linguistics to neurobiology. In these disciplines, “semantic space” is understood as the idea that any unit of language—say words or pairs of words—can be represented as a number or group of numbers, like an X/Y coordinate on a plane (though these representations usually have hundreds or thousands of dimensions, not just two). The proximity of the coordinates of two words in a system like this roughly corresponds to how related the two units are from a semantic perspective. It’s generally possible to make a visualization or map of this space, showing the similarity of units in terms of distance and size. The thing I noticed is that in these visualizations, there’s always an uneven distribution, or “clumps” of words/n-grams/concepts that are more common than others, and then empty spaces between the clumps.

Those little clumps are the islands of habitability in semantic space: they’re the concepts, sequences of words, and juxtapositions that are most highly traveled. We know them; they “make sense.” What I proposed in my talk is the idea that we might go searching in the empty areas of semantic space—the “nonsense”—to see if there are hospitable areas we haven’t seen before, or just to bear witness to the unique kinds of language there that would never otherwise occur to us.

VS: You describe some of these unexplored semantic spaces as being potentially “inhospitable” to the human mind. As I was reading through works that use Recurrent Neural Networks, for instance on Click-o-Tron, I found that it physically hurt in attempting to process the linguistic juxtapositions. I wonder if there are neurological implications in encountering these new semantic territories—if coming into contact with these spaces will physically alter our brains.

AP: I think that text generated by RNNs and Markov models is especially difficult because it does such a good job on the surface of resembling intention-typical text. (“Intention typical,” by the way, is my term for whatever the opposite of “computer generated” is.) Every word in a text generated by these methods makes sense in the context of the next few words that follow, but then something syntactically or semantically anomalous happens. You lose balance, and you find yourself jumping back in the text to find exactly where the train of meaning jumped the rails.

But RNNs and Markov models generate texts that are comparatively easy to engage with, in my opinion at least! In some of my classes, I teach David Melnick’s PCOET, a typewriter-based work of poetry that eschews conventional words altogether, and builds up its own expressive language from juxtapositions of individual characters. Ted Berrigan’s Sonnets look approachable at first glance, until you dive in and realize that because of their chance-driven, polyvocal composition, they resist conventional reading. (Even if you’re an experienced and patient reader of poetry.)

In order to engage with either of these texts (and countless others like them!), you need to throw out any preconceived notions you have about how an utterance is put together, and build up an entirely new understanding of what it means to read. One of the things that is exciting to me about composing text procedurally with computers is that it allows us to quickly propose arbitrarily many of these “new languages” for reconceptualizing how reading, writing, and talking work and are understood.

So yes, I do think that encountering new semantic territories has the potential to physically alter our brains. But that, to me, seems like an age-old mission statement for the entire field of poetry. The poet Charles Hartman wrote an amazing book called Virtual Muse that serves as a record of his own explorations with procedurally-generated poetry. When explaining the effect of unexpected juxtapositions in poetry, he says, “Juxtaposition makes the reader an accomplice in the poem, forging the links of meaning. In the process we supply a lot of energy, and that involves us in the poem.” I love that image—brains supplying energy. Lightning strikes, a static charge zaps across a Tesla coil! And in some sense it’s literally true: the brain is applying some amount of actual electrical energy to bridge the semantic “gaps” in a poem. Maybe someday poetry books (computer generated or not) will come with a label on front that gives measurement in Ohms of the poems inside, and how many calories you can expect to burn by reading them.

VS: There’s a sense of wonder that is experienced through your work, particularly in the act of taking words or images outside of their context to create new juxtapositions. Can you talk about how these semantic discoveries work to produce an alternative understanding of language through wonder?

AP: I’m honestly not sure. It’s easier to recognize when the sense of wonder has been achieved than it is to figure out by what means the effect was produced. I like it when poetry operates like a good crossword puzzle clue, or a riddle: something that seems obscure at first then becomes clear, and seems obvious in retrospect. There’s a real joy in making connections like that, and I think something approximating that joy results from closing a procedurally-generated (perhaps even random) semantic gap. The joy comes partially, I think, from the small adjustment in how you conceptualize the world that you have to make in order for the riddle or crossword clue to make sense.

Generative art is also very effective at evoking a particular kind of wonder: a sense of vertiginous smallness in the face of the infinite. The feeling I get when looking at the Grand Canyon—”oh my god, this is unknowably ancient and essentially has no beginning or end”—I also get from, say, zooming in on the Mandelbrot set, or watching Conway’s Game of Life, or listening to Laurie Spiegel’s The Expanding Universe. Part of the goal of the procedural juxtaposition of text, I think, is to reaffirm the generative and infinite nature of language itself, to assert that it hasn’t “all been said before,” that there are limitless undiscovered compositions of words and meaning that are yet to be discovered.

VS: Returning to the idea of sending a bot out to explore semantic space and bringing data back, in the bots that you have built, what is the data or material that is brought back, that then you look at and find of significance?

AP: I think the “purest” Twitter bot I’ve made from the perspective of exploring semantic space is Library of Emoji, which randomly generates speculations on what new emoji will appear in the next Unicode revision. Under the hood, the bot essentially just picks random words from the dictionary and smashes them together (with a few additional syntactic rules to mimic the outward form of Unicode character names). The semantic space it’s exploring is something like “all possible emojis” and the telemetry it sends back are, well, random emoji names. Some of these are more evocative than others, of course, and there are any number of ways to interpret the bot’s output. Personally, I like to take a tweet like “HEAVY SIGN FOR SHREWDLY ROUGH-SPOKEN MILLINERY” or “SHRIEK STACK” and imagine: how exactly would history have had to be different for these to be the emoji we have on our phones, instead of the emoji in this universe, like “SPEAKER WITH CANCELLATION STROKE” or “BLACK HEART SUIT”?

VS: The results of these combinations often generate humor, which seems to play a huge role in both the bot community and within bot-generated text. For you, how does humor function in the exploration of semantic space? And how does it relate to ideas of sense and nonsense?

AP: The most basic element of satire is mimicry, and it’s really easy to use procedural language to mimic language. You make up a model for how a particular genre of talk works (or even how a particular person talks), and then you throw data at that model. Mix in a bit of irony and whammo, you’ve got a satire machine! The message of a bot like this is usually something like: “Your method of talking is so predictable, easy to perform and semantically empty that even a simple computer program can do it.”

I’ve made my share of satirical bots (most recently Brain Tendencies) but I find it’s really difficult to avoid stepping over the line from satire to just being… mean, especially if the creative impetus of the bot comes from annoyance or frustration. But I don’t want to be mean! (Unless the person I’m being mean to really deserves it.) So lately I’ve been trying to focus my bots less on satire and more on bots that express joy and wonder at the wide variety of linguistic experiences we’re all capable of. (Although I don’t mean to imply that those two concerns are incompatible.)

Darius Kazemi (@tinysubversions) is a Boston-based artist-programmer and founding member of the #botALLY community. His recent shift towards the strategic pursuit of small projects has led to an array of Twitter bots, as well as numerous artworks and experiments published to other platforms.

As Kazemi notes, “lots of projects means you can have something to pull out for pretty much any occasion.” The technical and conceptual diversity of his text- and image-manipulating bots is testament to this, from the lazy joke-making critique of @Amiritebot to the recombinant poetics of @TwoHeadlines and @YouNeverDidThe to the algorithmic autofandom of @AU_prompts and @wirescenes. Kazemi is also one of several botmakers and thinkers to be influenced by object-oriented ontology, most notably by way of Ian Bogost’s Alien Phenomenology.

To sample the pleasures afforded by Kazemi’s oeuvre, there’s a bot for that: @dariusbots aggregates the most popular of his creations’ creations:

Tweets by @dariusbots !function(d,s,id){var js,fjs=d.getElementsByTagName(s)[0],p=/^http:/.test(d.location)?'http':'https';if(!d.getElementById(id)){js=d.createElement(s);js.id=id;js.src=p+"://platform.twitter.com/widgets.js";fjs.parentNode.insertBefore(js,fjs);}}(document,"script","twitter-wjs");