Well, it was hilarious. Points for that.
Excerpts from an opinion piece by Andrew Tarantola, writing at Engadget, 3/25/2016:
moot. Christopher Poole. Founder of 4chan, since resigned, now works for Google.
I agree with you, though: The kinds of trolls who do this know how to amplify their effective influence online. They know how to find, and are willing to exploit, vulnerabilities, making their antics more visible than their raw numbers would otherwise allow.
(That said, I’m pretty sure Boaty McBoatface is just a joke a large number of people find funny enough to vote for, as opposed to evidence of tampering.)
Boat McBoatface is about the least-worst example possible of this phenomenon. It ended up being good PR for the campaign, and did little real damage (apart from DOSing the site for a while)
But in other cases, it doesn’t take more than a tiny amount of stupidity per capita to create something nasty. None of the conspirators may have been particularly evil - all it takes is a mechanism that happens to filter and feedback input in a particular way, and the thing can snowball.
This is exactly what my post was trying to say.
No. That might be the result of an ~unannounced~ program that was allowed to soak up iCulture for a while before starting to interact on its own. This was just an invitation for trolls to fuck with it.
but LOL Micro$oft, amirite?
Oopsie, yes, him. Don’t know how my brain screwed up his nom de keyboard.
And yeah, like I said, Boaty McBoatface certainly picked up with the internet at large, but, maybe I’m just cynical, but I can’t imagine it would have gotten the steam to be notable enough to be seen by the internet at large, if not for the ‘hey, guys, let’s do this goofy thing’ factor.
Anyway, maybe we understand a bit better why we have moderators. Occasionally, I think they get a bit schoolmarmish, but we can live with that.
My daughter and I read the (censored) Twitter feeds from Tay last night to much hilarity. She doesn’t quite pass the Turing test because she admits she is AI or a robot fairly regularly and she is a little repetitive sometimes but it is damn close if you remove those posts. She is pretty funny even without the Hitler stuff and really wants to know why National Puppy Day isn’t Every Day.
I’m starting to think the bot was pre-programmed with a lot of things. Obviously that comment about needing responses was programmed in. And the “leskeep chattin” line is almost certainly a programmed response to someone replying too much.
On the other side, the pic requests look way too good. Not a repetition in sight. If they could do that, why not do it with the other two things I mentioned above? Her other responses seem more like she just takes things other people have said on Twitter, so it seems suspicious when she says something it would only make sense for a bot to say.
(Except the “humans” bit, which I assume was programmed in.)
Again, I’d really like info on how it works. If I’m wrong about the above, that’s pretty dadblasted good, and maybe a little scary.
A few months ago Microsoft demonstrated a new app for crime called ‘Predictive Policing’, that can say in advance which offenders will re-offend and be/should be jailed again within six months. This will help the law decide whom to identify and concentrate on.
We feel safer already.
Tay didn’t take her grounding all that well. Microsoft let her back on Twitter briefly after a firm scolding but it wasn’t long before she was bragging about smoking drugs in front of the police so she is on lockdown yet again. 
There’s an entire sick internet full of full-on forum trolls (4chan, I’m looking at you) who are all right at this moment making the “Challenge Accepted” face.
Microsoft will never be able to put this online in a publically-accessible fashion without having it twisted into whatever sad lulz the troll brigade are willing to make it.
Unless they give it a properly-protected learning childhood, and gradual adolescent exposure to the wider world.
All they have to do is do it anonymously.
Proposal: We make an AI ‘mascot’ from each board, forum, or other social media site by exposing a new AI to them in same manner as Tay was exposed to Twitter.
We get a representative average user for each site that can be compared and contrasted and all kinds of other interesting psychological and sociological functions. For data and study and other science type reasons.
Imagine ‘The Straight Dope’ talking to ‘Ars Technica’. Or ‘Youtube Comments’ talking to ‘The Free Republic’. Or ‘Conservapedia’ talking to ‘Uncyclopedia’. Such a time to be alive.