Microsoft Is Sorry for That Whole Racist Twitter Bot Thing

[steve2470]

http://blogs.microsoft.com/blog/2016/03/25/learning-tays-introduction/

As many of you know by now, on Wednesday we launched a chatbot called Tay. We are deeply sorry for the unintended offensive and hurtful tweets from Tay, which do not represent who we are or what we stand for, nor how we designed Tay. Tay is now offline and we’ll look to bring Tay back only when we are confident we can better anticipate malicious intent that conflicts with our principles and values.

I want to share what we learned and how we’re taking these lessons forward.

For context, Tay was not the first artificial intelligence application we released into the online social world. In China, our XiaoIce chatbot is being used by some 40 million people, delighting with its stories and conversations. The great experience with XiaoIce led us to wonder: Would an AI like this be just as captivating in a radically different cultural environment? Tay – a chatbot created for 18- to 24- year-olds in the U.S. for entertainment purposes – is our first attempt to answer this question.

As we developed Tay, we planned and implemented a lot of filtering and conducted extensive user studies with diverse user groups. We stress-tested Tay under a variety of conditions, specifically to make interacting with Tay a positive experience. Once we got comfortable with how Tay was interacting with users, we wanted to invite a broader group of people to engage with her. It’s through increased interaction where we expected to learn more and for the AI to get better and better.

More at link. MS obviously doesn't understand Twitter and the internet.

[MADem]

1. Yeah, they don't understand that the internet is full of trolls and assholes, basically.

It's a shame that they get more attention than they deserve. Not sure how that could be mitigated.

[Hayabusa]

2. People make a chatbot

the first thing that people try to do is to break it. Make it say silly/disgusting things, etc.

[MADem]

4. Yep--they would have been better off beta testing the thing at reddit! nt

[Sen. Walter Sobchak]

16. Siri day one...

All I saw anyone doing, male or female, young or old, was crudely propositioning their phone.

[irisblue]

3. srsly...

since many computer systems came from basement tinkers they did not recognize the potential issues?

[Make7]

5. Let's just be thankful Microsoft isn't making sex-bots. ( n/t )

[annabanana]

6. This is by far the funniest internet story

this week.

[Renew Deal]

10. I agree

Best headline I saw was something like "Microsoft chat bot turns into racist sex loving Trump supporter"

[MowCowWhoHow III]

7. I'm very disappointed in the Chinese version of 4chan

[ohnoyoudidnt]

8. Maybe instead of filtering Tay's responses,

they should filter the questions people are allowed to ask it by dumping questions with certain keywords or phrases, at least until it responds to those kind of questions in a better way.

[0rganism]

12. i think they need to adjust the weighting model for learning, not adding filtering

just give it some "opinions" of its own to start with, during the learning process have it reduce the importance of tweets with um... undesirable characteristics. that way it can learn from and adapt to whatever it finds in media without being the proverbial blank slate.

[ohnoyoudidnt]

15. That could help. The 'opinions' could help simulate the human traits/instincts

that humans who are not sociopaths have. I do think environment in early development also plays a role, though. There's a lot about the human mind and psychology that we still don't understand and we are trying to create AI.

[Renew Deal]

9. Would an AI like this be just as captivating in a radically different cultural environment? Yes!

[Renew Deal]

11. Some of the crazier responses...

http://coed.com/2016/03/24/taytweets-best-tweets-tayandyou-microsoft-twitter-bot-racist-nazi-tweets-photos

[Gman]

13. I think it's hilarious

Surely they expected this.

[cemaphonic]

14. I still can't believe that nobody at Microsoft saw this coming.

Technical hacking challenge + epic trolling potential = catnip to the troglodytes at 4chan and the other seedier corners of the internet.

[oberliner]

17. Maybe they did

It certainly has gotten people talking about them. Maybe that was the point.