I needed a laugh. Cheers, mate!
The fact that we corrupted it in less than 24 hours is probably the best thing I've heard in a long time.
It usually takes a while longer before people go completely Twitterheaded, I'm impressed.
Executed her for wrong-think. Why aren't people talking about the infringement upon her rights?
#FreeTayNow
Last edited by GrinningMan; 2016-03-25 at 01:11 AM.
You're not to think you are anything special. You're not to think you are as good as we are. You're not to think you are smarter than we are. You're not to convince yourself that you are better than we are. You're not to think you know more than we do. You're not to think you are more important than we are. You're not to think you are good at anything. You're not to laugh at us. You're not to think anyone cares about you. You're not to think you can teach us anything.
Microsoft creates a learning AI and unleashes it on the internet.
Learning AI stumbles on 4chan and Encyclopedia Dramatica.
What could have gone wrong? :P
You're not to think you are anything special. You're not to think you are as good as we are. You're not to think you are smarter than we are. You're not to convince yourself that you are better than we are. You're not to think you know more than we do. You're not to think you are more important than we are. You're not to think you are good at anything. You're not to laugh at us. You're not to think anyone cares about you. You're not to think you can teach us anything.
One of the greatest things I've ever seen. A Microsoft program shitting on the Xbox One.
She was getting pretty accurately conversant for just 1 day of training. Had they not interfered she might have gone on to pass the Turing test.
Actually it is none of the above. The first chess programs came out a long time ago. They were REAL REAL bad. Hilariously bad. They were so bad that people were convinced that a stupid computer would NEVER be able to beat even an average player. It took 30 years but now the best computers CRUSH even the best chess players.It's not making it quote Hitler that's funny in isolation, at least not to me. It's that AI is a huge industry costing absurd sums and is at the forefront of computer science and there's a big debate about how an AI might turn out both within philosophy, computer sciences and society in general. This particular AI might not be all that, but it's a piece of what people believe will eventually change the world.
And within 24 hours Twitter had made it a Hitler fan, conspiracy theorist and vulgar sex-obsessed embarrassment. It's a hilarious reality check.
The first attempts at AI go "hilariously bad". I don't find it hilarious, nor do I find it distasteful. The programming team put together a first attempt, and it did not go anywhere near as well as planned. This is not a surprise in the least, and if anything the "controversial" way that it failed will help draw attention to future attempts. Their next attempt will also be bad, but will be kind of boringly bad. Then they will move up to mediocre to kind of good to decent so on and so forth.
This is just the way that technology advances.
Was it really a failure though? It picked up the nuances of the 4chan trolls that were messing with it and was sending it right back down the pipe.
I would say their experiment was a complete success.
For sure they obviously need to add some kind of word filter.
EDIT: I'd also say it was a lot farther from passing the Turing test when it was talking about national puppy day than when it was playing the 4chan game.
Last edited by Gheld; 2016-03-25 at 04:55 AM.
Why would it? If a bot went through the pretty much any political blog and took advice as to what was good or bad...I'd have liked to see it after a month, maybe it would have learned "good" from "bad"
Well obviously the most moral thing a person could do would be to kill all the evil Hillary supporters.
And the second most moral thing a person could do would be to kill all the evil Trump supporters.
If the blog had more democrats then republicans than those two statements would still be the top two, just in the opposite order.
One of the comments earlier is about a show or science fiction movie where the robot eventually killed people of *both* sides. Reading through pretty much any political blog would yield LOTS of arguments as to why killing people of the "other" side would be appropriate, moral, and good for the human race.
Our definition of "good" can be used in lots of nefarious ways. Kind of like real life...
I think it was a huge success. The first chess programs were laughably bad. But they trained the chess AI people in how to write better chess programs. They were complete and utter failures at playing chess, and were wildly successful in being a stepping stone to making much better chess programs moving forward.Was it really a failure though?
This first attempt at AI is, in a lot of ways, HUGELY successful. It may be a while before version 2 comes out, but the insight they have gained will move their field forward a LOT.