Turing and Twitter
Azure AI Bots. They can certainly do basic tasks like playing music, answering Google-able questions, return orders, and schedule appointments. But, if properly configured, can they do more? While they're too primitive to pass for a human on a basic level, could they reasonably pass for a human on the most primitive cesspools of social media? Could they be configured to pass for a human on Twitter?
Moreover, as a measure of success, could they be more successful than your average user? To do this, I'd need to find a truly average Twitter User. Someone who shared the thoughts, opinions and mean intelligence of your average Twitter prole.
Friends, meet Target:
Target is a long-time friend, associate and occasional conspirator of mine, and a Twitter user since 2009. Like much of Twitter, he enjoys platitudes, circular reasoning, sharing things that result in his smug, engaging with his own echo chamber and berating anyone on the outside. If I am my Twitterbot's code-father, this is his code-mentor; a role model for him to emulate and learn from as he grows.
The goal? To eventually overcome Target's lofty hurdle, exceeding him in both average likes/retweets and followers. The Twitterbot must be accepted as a human To that end I created:
Twitterbot's Daily Loop:
- Follow everyone., prioritizing whackjobs with few followers. This is especially necessary as while Target practically lives on Twitter, most of his posts comprised of Dad jokes or shared photo memes that would be difficult for my API to parse. This gave me substantially more material to adapt, including a very amusing cuck in Target's followers.
- Adapt the profile at least once a week, changing up the pronouns.
- If someone referenced my account by something other than my name, accuse them of not using my preferred pronouns.
- If someone argued with me, brainlessly twist the argument to be about them OR simply accuse them of being some sort of -ist.
- Also, always play the victim.
- Copy others...
- But use proper spelling and grammar, and in general try to be less pretentious.
- Improve readability and SEO.
The Dull and Lifeless Original: - VS better use of hashtags, more obscenities, consistently "religious" language, and sexual references for the proles:
- If someone's trying to say something profound, twist in it such a way to point out how hypocritical it is, or simply misattribute the quote
- If someone makes a humble brag, twist it to make the blatant hypocrisy, pettiness, or desire for validation self-evident.
- If someone blasts a particular party/demographic, make it a different demographic or make the blasting even more ridiculous and callous. Amp that strawman up to 11. (I can't stand all these trans women transspreading in the train today. It's disgusting.)
- If someone pushes a particular agenda, take that agenda to the extreme.
The variables I was able to tweak:
- Sarcasm
- Sentimentality
- Wordiness
- Pathetic-ness
- Use of emoticons
- Use of obscenities
- My matrix for what qualifies a whackjob,
- How many of Target's posts it likes per week.
My Resulting Twitterbot:
The Timeline:
Oct 14 - After a few days of unit testing, Twitterbot is in Beta mode.
Oct 15 - 56 Followers
Oct 16 - Upped obscenities
Oct 18 - First major bugpatch after Twitterbot, inspired by one of the cuck's "bulls," almost uses the N-word.
Oct 19 - Twitterbot gets its first retweet.
Oct 21 - 98 Followers, Twitterbot upgraded to V 1.1
Oct 28 - Increased obscenities
Nov 6 - 156 Followers. Tweaking posts to run hot and cold with
Nov 12 - A self-described frat boy believes Twitterbot is a 'hilarious troll' and gives a follow. Has a hilarious convo about sharing Twitterbot's nonexistent wifebot.
Nov 18 - Had Twitterbot follow Target.
Nov 19 - Target discovers Twitterbot and proceeds to argue with it, blocking it without ever realizing it's a bot. Turing test passed.
Nov 21 - I discover argument, Somewhat amused. Reduced wordiness. Upped sarcasm and obscenities.
Nov 22 - TERFS find Twitterbot's pro-trans "date trans people; take that female penis, bigot" post and proceed to go absolutely insane all over his account.
Nov 23 - TERF engagement peaking. It is glorious.
Nov 26 - Twitterbot surpasses 200 Followers
Nov 28 - First TERF threatens to dox and 'out' Twitterbot. Sarcasm increased.
Dec 1 - Twitterbot tweets from hospital after discovering Chris Pratt will be playing Garfield in an animated movie for children. Attempted to jump from Lego tower into puddle of stale La Croix.
Dec 2 - First TERF figures out Twitterbot may actually be a troll after being called a 'big fascist meaniepants doo-doo head.' And possibly the Chris Pratt thing.
Dec 6- Last TERF comment left.
Dec 9 - 236 Followers.
Dec 16 - Twitterbot surpasses Target in followers with better engagement.. Goal met. Experiment terminated.
Target's Stats:
VS Twitterbot's:
- Hashtags have a significant impact on SEO. Target tended towards custom hashtags, while Twitterbot tended towards more popular ones.
- People actually like the pronoun thing. The pronouns themselves can be idiotic.
- Users with fewer followers are more likely to give a follow.
- Most Twitter uses care less about spelling./punctuation, and more about wordiness.
- Obscenities increase engagement. People seem to see them as a mark of sincerity/
- Most Twitter-users either skim or don't think terribly critically. They're willing to like and retweet pretty much anything if it seems to harmonize with their ideology. I had people liking/retweeting some posts about Twitterbot being cool with
- Right-identifying people were more likely to question Twitterbot's sincerity. This is likely because they're more used to being attacked/trolled on Twitter.
- TERFs are possibly the silliest people on Twitter.
- Target investigates anyone who follows him, but doesn't engage long enough to determine if bot or human. To be fair, he may not think it's worth it.
- Azure Bot AI is actually pretty comprehensive and adaptable.
- My faith in humanity has been lost. I may never leave my home.
Comments
Post a Comment