-
Notifications
You must be signed in to change notification settings - Fork 1
Research: Thoughts on Bots, Bot Research (Saito)
Tachikoma 4: What I'm trying to get across is that humans want their machines to respond like machines.
Tachikoma 1: Oh, now I get your drift. If we just act a little more robotic...
Tachikoma 2: We might stand a chance of [humans] liking us!
Tachikoma 4: Exactly!
Tachikoma 3: It's the ultimate robot strategy plan!
**All four Tachikomas, in unison: **[monotone] We are robots. We are robots. We are robots.
Tachikoma 3: Aaaaaah I can't do this anymore!
There’s more m2m (machine to machine) communication on the internet than h2h (human to human). Probably more m2h and h2m than we’d like to admit as well. In the series Ghost in the Shell, the Tachikoma (armored robots) are trying to figure out why their superior officer, the major (a human), doesn’t like them. They conclude that they are acting “too human” and they have to act “more robotic,” but also they also feel like they can’t do it... We’re not quite at the point where we have to worry about this kind of conundrum, but the bots operating on social media might be a good precursor to some of these dynamics. Yet, differently from how they’re portrayed in media, the patterns of human-machine collaboration are much less separate and much more intermeshed. It’s not human vs. machine, it’s human and machine. On the path to bot-building, Saito reviewed a lot of the bots that exist. Saito came up with a taxony (see below) and looked at some specialized use cases (novel-writing, protest).
There are two types of twitter bots, ‘sourcing bots’ and ‘output bots.’ Sourcing bots tend to aggregate content from three sources: human action, other bots and the environment. For example, Allison Parrish’s “oneiroposiesis” bot scans for the words “I dreamt,” takes the following language and posts it:
These are Saito’s favorite bots that categorize human action:
https://twitter.com/oneiropoiesis list of human dreams
https://twitter.com/theDesireBot list of human desires
https://twitter.com/censusamericans list of every American, with some info
https://twitter.com/Every_User_ list of every twitter user
https://twitter.com/TheftFromMaster list of historical stolen items
Another of Saito’s favorite bots, Station5100, juxtaposes text (from Moby Dick?) with images from a buoy’s location and data:
These are Saito’s favorite bots that categorize and/or interact with other bots/digital infrastructure:
https://twitter.com/storyofglitch a digital cat
https://twitter.com/one_algorithm list of algorithm actions
https://twitter.com/KatamariItems list of Katamari items
https://twitter.com/NewHorizonsBot list of new horizons probe images
https://twitter.com/dronesweetie list of drone movements
https://twitter.com/adventvrecall crowd sourcing a character’s actions
https://twitter.com/EarthRoverBot crowd sourcing a robot’s movements
Saito’s favorite environmental monitoring bot (or perhaps speculative environmental monitoring bot), 100yearsrising, automates data about rising sea levels in the future.
These are Saito’s favorite environmentally oriented bots:
https://twitter.com/100yearsrising projected sea rise levels
https://twitter.com/a_travel_bot mini travel guides for random locations
https://twitter.com/BPEarth images and locations of random locations
The other category of bots are ‘output bots,’ bots whose authors appear to be more concerned with the resulting output, rather than aggregation of existing material. Saito divides these kinds of bots into mimics and satire bots. Thrice Dotted’s @soft_focuses is particularly poignant example of the former:
Saito’s favorite mimics:
https://twitter.com/the_ephemerides images of outer planets and texts
https://twitter.com/moonmurmur erotic messages from the moon
https://twitter.com/wikisext sexting/gender commentary
https://twitter.com/poem_exe automated japanese poetry
Darius Kazemi’s @HottestStartups is a pretty excellent example of the latter category, where different domains of language collide for ironic effect (in this case, startups and marxist theory):
Saito’s favorite satire bots:
https://twitter.com/hotteststartups marxist start ups
https://twitter.com/CyberEveryword cyberwords
https://twitter.com/ghost_things ghost things
https://twitter.com/nice_tips_bot advice robot
https://twitter.com/action_inaction action groups
Here’s a category that doesn’t exist yet: suggestion bots. A bot as a kind of input method editor (https://en.wikipedia.org/wiki/Input_method). Saito often writes using machine mediation. e.g., Saito will often write a poetic text, translate it into japanese, translate it into chinese, then translate if back into english and pick and choose the mistranslations to assemble the final piece. Could there not be a kind of mistranslation robot that acts as an IDE to help an author create newer/stranger forms of language? It seems to Saito that, at this moment, humans appear to be better at certain things while machines appear to be better at others. Could there not be a class of bots running trainable processes (identifying certain lines in japanese literature, identifying the most recent statements of politicians, identifying tweets posted within a one hundred foot radius in the past seven minutes, identifying data relevant to a topic, etc.) that would turn a writer a real-time data-remixer?
In addition to the short form twitter bots, there are the long form bot-assisted works of literature that come out of NaNoGenMo (national novel generation month). They’re process driven works; if you ‘feel the process’ they are immensely satisfying, and reflective of both human and machine hybrid-intelligence. The works are created with multiple different forms of human-machine collaboration. Teens Wander around a House, by Darius Dazemi, uses a dream bank as well as question/answer snippets from twitter.
The Seeker by Thrice Dotted is a machine traveling through the internet discovering human concepts. According to Thrice Dotted “Every run of The Seeker is completely unique, because it relies on external randomness (in this case, WikiHow). The Seeker is at once an algorithm, an agent, a protagonist, a narrator. At its essence, it is an entity that parses, deconstructs, and reconstructs text. The output of this algorithm are its "logs" of doing this. The Seeker operates in three modes: Work, Scan, and Imagine. When the Seeker Works, it is scraping concepts about human activities from WikiHow. In Scan mode, it searches plain text "memories" for a seed concept it encountered during Work. It uses the concepts it doesn'trecognize from Scan mode (i.e., the ones which are censored out in its logs) to Imagine an "unvision" around the seed concept. And so on. And so forth.”
Allison Parish’s Novel “I Waded in Clear Water” was “generated from the text of Gustavus Hindman Miller's Ten Thousand Dreams, Interpreted. The "What's In A Dream" section of Miller's book functions as a dream dictionary: you look up a word, and find out what it means to dream about that word's referent. Each word has multiple interpretations, and most of these interpretations can be broken down into what I call anaction and a denotation... The text of this novel was made by extracting the actions and changing them to first-person. The denotation for each action is scored using a sentiment analysis algorithm, and the sentences are printed in order by the sentiment of their corresponding denotation, from most negative to most positive.”
At this point, I have to reference Mark Sample’s incrdible work in full, “A Protest Bot is a bot so Specific You Can’t Mistake It for Bullshit.” Mark goes onto describe what exactly makes up the characteristics of a protest bot:
(1.) Topical. Asked where the ideas for his song came from, Ochs once pulled out a Newsweek and smiled, “From out of here.” Though probably apocryphal, the anecdote highlights the topical nature of protest songs, and by extension, protest bots. They are not about lost love or existential anguish. They are about the morning news — and the daily horrors that fail to make it into the news.
(2.) Data-based. Bots of conviction are based in data, which is another way of saying they don’t make this shit up. They draw from research, statistics, spreadsheets, databases. Bots have no subconscious, so any imagery they use should be taken literally. Protest bots give witness to the world we inhabit.
(3.) Cumulative. It is the nature of bots to do the same thing over and over again, with only slight variation. Repetition with a difference. Any single iteration may be interesting, but it is in the aggregate that a protest bot’s tweets attain power. The repetition builds on itself, the bot relentlessly riffing on its theme, unyielding and overwhelming, a pile-up of wreckage on our screens.
(4.) Oppositional. This is where the conviction comes in. Whereas the bot pantheon is populated by l’bot pour l’bot, protest bots take a stand. Society being what it is, this stance will likely be unpopular, perhaps even unnerving. Just as the most affecting protest songs made their audiences feel uncomfortable, bots of conviction challenge us to consider our own complicity in the wrongs of the world.
(5.) Uncanny. I’m using uncanny in the Freudian sense here, but without the psychodrama. The uncanny is the return of the repressed. The appearance of that which we had sought to keep hidden. I have to thank Zach Whalen for highlighting this last characteristic, which he frames in terms of visibility. Protests bots often reveal something that was hidden; or conversely, they might purposefully obscure something that had been in plain sight.
And lastly, an ethical concern from Allison Parrish that Saito believes should be added to this last:
(6.) This is a quote from a great essay by Leonard Richardson called “Bots Should Punch Up:” [W]hat’s “ethical” and what’s “allowed” can be very different things… You can’t say absolutely anything and expect, “That wasn’t me, it was the dummy!” to get you out of trouble.
Saito divides activist bots into two (related) categories, monitor bots and protest bots. Monitor bots keep track of political information. For example, https://twitter.com/congressedits notifys when Wikipedia Articles about congress are edited from IP addresses within congress:
Saito’s other favorite monitoring bot https://twitter.com/FBIbot tweets random pages from FOIA requests.
The other category is protest bots, relentlessly posting about political abuse:
Saito’s favorite protest bots:
https://twitter.com/stopandfrisk stop and frisk, with results
https://twitter.com/nra_tally data about gun deaths
Imagine a world in which there was a bot kit. In fact, imagine a world in which there were multiple bot kits, they were open source, and had a great/intuitive UX/UI. Saito thinks that there are essentially three things these bots could do: (1.) gather information, (2.) suggest information and (3.) persist. For example, let’s say you wanted to monitor a political candidate’s campaign contributions. You could create a bot that monitored a campaign finance database and checked the donors against political watchdog organizations. From there, if you wanted to write a letter to your congressperson and/or to your friends, the bot could autosuggest data from what it had monitored while you were writing. And finally, you could unleash a bot onto social media that would persist, finding ways to broadcast this data indefinitely. Westley Argentum, one of the lab fellows, is making investigations along these lines.
Saito is working on a different kind of bot, a bot that gathers geo-located social media, mentions and/or social media from an organization’s members, then does three things (a.) remixes the material into a poetic-political text (main text), (b.) ships the data to a team of writers who can create more advanced remixes with the help of an suggestion bot, and then (c.) incorporate these remixes back into the main text. All of this material is then projected onto an organization’s office and/or used during political events and/or happenings. Saito is particularly interested in working with organizations like the Electronic Frontier Foundation or the Sudo Room that typically would not be employing these kinds of techniques. This isn’t quite the world where everyone has access to their own automated intelligence on their own terms, but it’s a step towards that world.
Saito’s compiled a by no means exhaustive and/or politically fair list of tool kits:
Monitoring:
- https://ifttt.com/ - workflow protocal, alerts
- https://www.google.com/alerts - google alerts
- http://www.socialharvest.io/#story - social media monitoring
Suggesting:
- https://www.crystalknows.com/ - suggestions about how to write to people based upon their personality as determined by publically available information
Persisting:
- https://www.eter9.com/ - art project? service? that automates your facebook profile
- https://github.com/thricedotted/twitterbot - Thrice Dotted’s template for a twitter bot
- https://www.narrativescience.com/ - a company that does 'automatic data textualization'
- Colorbot is a NodeJS slackbot that assigns color values to common emoji and updates a browser window with colors. Think of it as a mood ring for your slack channel. CS
- Simple Slack Bot is a Python Slack bot that responds to search terms (or calls) with responses. There are other Slack bots that do this better, but this is mine. ABH
- Frisco Bot