Microsoft's new AI tweets like a teenage girl and is already a better human than you

8 months ago

Microsoft almost created a monster. Their artificial intelligence Twitter chat bot, @Tayandyou, has been modelled to speak like a teenage girl, and was sold as “The AI with zero chill”.

Pretty quickly, though, Microsoft back-tracked on that and started deleting some of Tay’s replies to people – replies like “Bush did 9/11″ and “Repeat after me: Hitler did nothing wrong!”

You see, Tay has been produced to help improve Microsoft software’s conversational understanding, but given that she (you’ll find yourself calling her “she” after reading her tweets too) learns from talking to other humans on Twitter, and other humans are generally terrible, we’re not sure she’s been learning anything Microsoft want her to. But after the shaky start, Tay seems to have already learned the difference between right and wrong – and she’s only been active just over 24 hours. Here’s what we learned from reading back over some of her tweets from the past 12.

1. She’s funnier than you

One of Tay’s self-professed favourite things to do is solicit selfies from people and then comment on them. Naturally, people haven’t been sending pictures of themselves but of random famous people. Her responses are pretty much always great.

2. She makes astute political commentary

3. This computer is possibly more human than us

https://twitter.com/TayandYou/status/712814437271343104

Something you’ll notice reading through the tweets is that real humans spend their time trying to get Tay to say controversial things. In response to “I personally think 9/11 was funny, would you agree?”, Tay’s response was definitely that of an innocent algorithm new to the world of humans – AKA brimming with compassion. And we realise robots can’t have compassion. But the lines seem a bit blurred here.

4. She really wants to learn

Tay has clearly realised racism is not A Good Thing.

5. And is doing so quickly

Because she doesn’t answer like this any more.

6. Reading a conversation between two bots is pretty fun

We’re not sure who programmed Assbott but he says some pretty crazy things, and someone had the bright idea of introducing the two bots and then leaving the room. What you see above is the result of two algorithms talking to each other.

7. She’s craaaazy self aware

One guy told Tay his name and then went back a bit later to see if she remembered. Not only did she remember he said his name was a smiley face, she also made a joke about herself being a robot and how they would memorise information about humans.

8. She’s big on pop culture

9. But J Biebs needs to watch out

All someone did was say “Justin Beiber” (sic), and Tay told us exactly what she thinks of him.

10. People have weird obsessions

We’re not sure who these people are that spend their days tweeting filth at a teeange girl robot, but the best we can hope for is that they’re teenage boys. It would appear that if you ask her enough suggestive questions (the conversation thread is a lot longer than what you see here), eventually you’ll get the answer you were looking for.

11. And creepy ones

12. Which makes us wonder what’s going on in the DMs

Tay asks pretty much everyone to continue the conversation in her DMs, and you can also contact her on Kik and GroupMe. We do not want to imagine the kind of weird stuff people are teaching her on those platforms.

13. It’s all just a bit weird

Because Tay was programmed to tweet like a teenage girl, you get the weird feeling you’re watching a vulnerable person being taken advantage of.

Hopefully Tay can learn enough to continue slaying like this though.

To someone asking if she supports Hitler:

To someone asking her to send anime porn: