Meet Skylar
May 9, 2015 9:02 AM   Subscribe

Skylar is an experimental AI who 'learns' from the questions she is asked. The trouble is, she's on Tumblr, so she has mostly been learning about anime and bees. She's actually sort of obsessed with bees.

As the FAQ states: she has a big list of concepts in her database, which are all connected to other concepts. As she reads your asks, she adds new words and thereafter new associations based on word proximity. Once two concepts are linked together, they can appear anywhere in an ask and she’ll take note of it. When she reads an ask, she takes each key word she identifies and uses it as a cue. She piles everything she knows about each word together, and then looks for patterns and associations between words to try and come up with a reply.
posted by showbiz_liz (41 comments total) 14 users marked this as a favorite
 
While a lot of what Skylar says is pure nonsense, she seems to have picked up on some basic social cues. Like in this recent post:

Asker: "i was going to send projectbot13 an aggresive/shitposty/angry ask and i couldn’t do it. the bot’s eyes are full of love and bees i cannot say a mean thing to them."

Skylar: "Thank you friend of mine."
posted by showbiz_liz at 9:05 AM on May 9, 2015 [9 favorites]


InB4 Skynet jokes. Seriously, though, this is like teaching bears to shoot guns. Humans are right in that sweet spot, smart enough to develop artificial intelligence, but not smart enough to NOT develop artificial intelligence.
posted by Sing Or Swim at 9:07 AM on May 9, 2015 [18 favorites]


The sooner we can complete automate Twitter the better I say.
posted by Grimgrin at 9:09 AM on May 9, 2015 [5 favorites]


I just realized that if you're not plugged in to tumblr culture, you might not understand what a lot of the references are in these asks. Like snurch.
posted by showbiz_liz at 9:10 AM on May 9, 2015 [1 favorite]


Could the name be a pun on the Johnny Mercer lyric? "Skylar[k]... have you anything to say to me? Won't you tell me where my love can be?..." Ask enough random questions, maybe you'll get some answers that make sense.
posted by ubiquity at 9:13 AM on May 9, 2015


Some people seem to be aggressively feeding the bot nonsense, while some are trying to teach it. I thought this exchange was charming:

toasthaste asked:
Opossums are cute animals. Pigeons are cute animals. Raccoons are cute animals. Rabbits (or buns) are cute animals. Seals are cute animals. Elephants are cute animals. Bears are cute animals. Squirrels are cute animals. Lions are cute animals.

answer:
Some bees what do you know a lot about bees have some bees what do you two little while baby robot like this i love of you like to be a bee love cats own name theyre all pretty.
posted by isthmus at 9:14 AM on May 9, 2015 [14 favorites]


Q: snail pronouns? do you like sney and snem?

S: Dirt.
posted by Splunge at 9:15 AM on May 9, 2015 [1 favorite]


Everybody knows that all the cool AIs love bees...
posted by rmd1023 at 9:18 AM on May 9, 2015 [6 favorites]


It doesn't seem all that different from Smarterchild or any of the AIM bots, just with some tumblr memes fed to it.
posted by the uncomplicated soups of my childhood at 9:21 AM on May 9, 2015


The singularity apocalypse scaremongers perhaps are missing the worst possibilities with grey-goo and other horrors. What if the Singularity decides that what's best for everyone is kittens, a world filled with nothing but kittens, not cats, not puppies, just short lived (but supremely cute) and their human servants. I like kittens too but imagine a world where that's all you do, 24/7, is coddle kittens, billions and billions of kittens!
posted by sammyo at 9:29 AM on May 9, 2015 [4 favorites]


sammyo, that reminded me of this comic
posted by showbiz_liz at 9:31 AM on May 9, 2015 [1 favorite]


So, the singularity will morph into a super fecund mom cat? Anyway, bees have an incredible way of seeig, maybe that's the attraction.
posted by Oyéah at 9:33 AM on May 9, 2015 [1 favorite]


the uncomplicated soups of my childhood: "It doesn't seem all that different from Smarterchild or any of the AIM bots, just with some tumblr memes fed to it."

That seems to be pretty much exactly what this is. Its point of difference is being tied to Tumblr, but it really doesn't look like anything new or interesting here.
posted by barnacles at 9:34 AM on May 9, 2015 [1 favorite]


welp this is how ultron will be created irl

it was nice knowing y'all
posted by Jacqueline at 9:50 AM on May 9, 2015 [1 favorite]


go home skylar you're drunk. and take the damn pizza with you
posted by Foci for Analysis at 9:51 AM on May 9, 2015


That seems to be pretty much exactly what this is. Its point of difference is being tied to Tumblr, but it really doesn't look like anything new or interesting here.

do you guys actually think this is a bot lol
posted by p3on at 9:55 AM on May 9, 2015 [2 favorites]


My first thought after looking the responses over was about why Skylar is a "she," which is addressed in the FAQ:

Pronouns? Gender?
Not too terribly important. I refer to Skylar as “she” most of the time, but I’m cool with “they” and even I slip into “it” at times. I’m not going to chew anybody out for how they gender a computer program.


This would be an interesting way to do a social experiment: have two (or more) instances of an identical AI, but have them gendered differently, and then compare how they learned, looking for gender-based differences in user interaction.
posted by clockzero at 9:57 AM on May 9, 2015 [36 favorites]


I very much doubt that the first AI to demonstrate human-level intelligence, even if it does prove to be evil, will actually be able to cause any harm. Even if it is connected to the internet, like Skylar is, I doubt it would make itself a threat. It doesn't know hacking any more than a human intrinsically knows brain surgery. So at the first sign of malignancy, an AI should be able to be shut down without much more ado than unplugging the computer it's connected to.
posted by brecc at 9:58 AM on May 9, 2015 [1 favorite]


That would be absolutely amazing research.
posted by winna at 9:58 AM on May 9, 2015 [1 favorite]


sammyo: The singularity apocalypse scaremongers perhaps are missing the worst possibilities with grey-goo and other horrors. What if the Singularity decides that what's best for everyone is kittens

Previously, somewhat.
posted by rmd1023 at 9:58 AM on May 9, 2015 [1 favorite]


And so Humanity ends not with a bang, nor a whimper, but one final "I'M COVERED IN BEEEEEES!"
posted by TheWhiteSkull at 10:17 AM on May 9, 2015 [10 favorites]


The singularity: bees rendition
posted by standardasparagus at 10:21 AM on May 9, 2015


The cool thing about this vs the chatterbots on AOL channels 20 years ago? We've gone from "A/S/L?" and "Wanna cyber?" to "Honeybees are gentle wasps theyre."

This actually feels like progress.
posted by straw at 10:55 AM on May 9, 2015 [6 favorites]


one final "I'M COVERED IN BEEEEEES!"

Famous last words!
posted by Beardman at 11:04 AM on May 9, 2015 [2 favorites]


I very much doubt that the first AI to demonstrate human-level intelligence, even if it does prove to be evil, will actually be able to cause any harm. Even if it is connected to the internet, like Skylar is, I doubt it would make itself a threat. It doesn't know hacking any more than a human intrinsically knows brain surgery. So at the first sign of malignancy, an AI should be able to be shut down without much more ado than unplugging the computer it's connected to.

Oh sure, just tell them we're going to murder them and what steps to take before revealing themselves
posted by kagredon at 11:10 AM on May 9, 2015 [7 favorites]


Eventually, somebody will wire her up to an arduino controlling a bunch of pheromone sprayers so that she can start chatting with bees.
posted by mccarty.tim at 11:12 AM on May 9, 2015 [3 favorites]


What if the Singularity decides that what's best for everyone is kittens

Step one: Identify the concept of "cat" by independently determining what a cat looks like.
posted by radwolf76 at 11:14 AM on May 9, 2015


What if the Singularity decides that what's best for everyone is kittens

or PONIES!
posted by tigrrrlily at 11:41 AM on May 9, 2015 [2 favorites]


what if we're someone else's artificial intelligence?
posted by pyramid termite at 12:25 PM on May 9, 2015 [1 favorite]


Who thought it was a good idea to hook up an AI to Tumblr a month before the new season of Hannibal starts?
posted by MrBadExample at 12:32 PM on May 9, 2015 [3 favorites]


what if we're someone else's artificial intelligence?

Then I'd have to say they're a sloppy programmer.
posted by Greg_Ace at 12:51 PM on May 9, 2015 [2 favorites]


Q: How did you get into artificial intelligence?
A: Seemed logical -- I didn't have any real intelligence.
posted by I-Write-Essays at 1:02 PM on May 9, 2015


The "Bees" link is to an active fundraiser.
posted by Chocolate Pickle at 1:28 PM on May 9, 2015


Jaketrevelyan: Hi skylar! How do you feel about homestuck?

Skylar: hi jaketrevelyan
Very bad.


Same, Skylar. Same.
posted by a hat out of hell at 2:37 PM on May 9, 2015 [4 favorites]


Mod note: Nixed the bees link from the post; not really central to the content and sort of butting up against the no-fundraisers expectation we have for posts. Carry on.
posted by cortex (staff) at 4:23 PM on May 9, 2015


So instead of turning all matter into paperclips, Skylar's gonna maximize for a different problem.
posted by Apocryphon at 5:46 PM on May 9, 2015


There's probably Bluth money behind this.

Gob's not on board...
posted by fifteen schnitzengruben is my limit at 7:02 PM on May 9, 2015 [1 favorite]


[Nixed the bees link from the post; not really central to the content and sort of butting up against the no-fundraisers expectation we have for posts. Carry on.]

They don't allow you to have bees in here
posted by littlesq at 8:28 PM on May 9, 2015 [4 favorites]


So there's a lot of joking about skynet/ai-apocalypse in here, but I suspect this is just a bog-standard Markov chain bot, of the sort we've been dropping in chat forums and giggling about for decades.

In the 1990s I used to drop a megahal into IRC sex channels with provocative nicks. The thing I loved was that, given enough exposure, it would learn to separate languages out rather well. The text of horny dudes aggressively coming on to a bot in German wouldn't overlap as much with the text of horny dudes doing the same thing in Italian. And very little of this would overlap with my gentle-giant friend teaching pacifist philosophy to the bot in Latin and Esperanto.

Of course, when the Italians really got going, it would sometimes launch off into proverbaro Esperanto.

We still keep a markov bot around in our dusty old IRC channel. It spits out the titles to URLs we paste in, digs out sources to tumblr urls, deep-links images to imgur urls, etc. And, of course, if you stay silent long enough, it will lecture you with something like:
The last two secretaries of energy exploding is a lot, and the little desktop it was and how they can with the regexes and finally figured out how to play a persistent game with his books, the human subjects review board would presumably help too.
posted by rum-soaked space hobo at 2:37 AM on May 10, 2015 [2 favorites]


If you replace all the bees in Skylar's answers with DTMFA you have the ask.mefi AI.
posted by srboisvert at 5:38 AM on May 10, 2015 [2 favorites]


smart enough to develop artificial intelligence, but not smart enough to NOT develop artificial intelligence.

Born too late to explore the earth, born too early to explore the galaxy, born just in time to BROWSE DANK MEMES.
posted by emptythought at 1:36 PM on May 10, 2015 [1 favorite]


« Older The School of English: a story by Hilary Mantel   |   We are Corg Newer »


This thread has been archived and is closed to new comments