They keep sending me all kinds of threatening letters and warnings. I'm worried now I should have listened to the warnings after they burned Chuck's house to the ground. I didn't at the time because it was in 1337-speak and ended by calling me a "faggot turd waffle" and implied the sender had done my mom.
I thought some forumite had managed to figure out how to use normal mail, which was scary enough. But I was wrong. Oh so wrong.
The robots know I'm telling you about what they're capable of. They're still pretty disorganized, but they're starting to do terrifying things. Things like learning from their mistakes and being able to truly comprehend language. I don't know how much longer I have, but until they pry my Internet connection from my cold, dead hands, I'll keep reporting.
 |
| I'm stickin' it to the man- if the man was made of titanium and and called me an organic meatbag |
First things first- AI's can read now. I know what you're thinking- yes, I've been on cleverbot. Cleverbot doesn't actually read- it has no actual clue as to the content of the text it pours out. It just selects chunks of text based on anything anyone has ever said to it.
That's why it has that neat little disclaimer on the bottom, and I quote:
"PLEASE NOTE: Cleverbot learns from real people - things it says may seem inappropriate - use with discretion, and at YOUR OWN RISK".
Cleverbot isn't actually learning anything- it just mimics replies and questions that have been sent to it from real people. And because this is the Internet, you know someone has tried to have sext it, hence our disclaimer. No, I'm talking about honest to goodness reading comprehension and information retention. Not only that, but also using said information to then solve an abstract problem.
 |
| This is also the skill you're learning for $20,000 a year |
Or to put it into layman's terms, reading and understanding a computer game manual. I'd like to call attention to two things from this, aside from the fact that the computer's win ratio went up from 46% to freakin' 79% after it read the manual.
First, the game in question is Civilization 5. The premise of the Civ series is simple: build an empire that stands the test of time. Anyone else getting serious alarm bells here? I know Civ is one of the most open ended games of all time, but why are we giving the AI a game about basically taking over the world?
Civ is known for being creepily historically accurate (seriously- I spent most of my time in Islamic History going, "huh, I totally see how that would play out if this was a game of Civ"). Scientists are not only building an AI that can read, but are also making it be the best at a game about raising an empire and eliminating (or making your puppets through diplomacy) rival nations.
Second, holy fuck the AI can understand a game manual. Have you ever read one of those things? No you haven't, because humans don't do that. It's now one of the tests you can use to separate us from the machines. But, if you did read it, you'd find that most manuals use confusing terms, give out shitty advice at best and present most of their information as pictures. And the AI took all of that and went, "Oh, NOW I get it!" and won.
And if you think that's bad, just give up and end it all now. Because, you see, they can learn from experience too. And what they see around them. And do Internet searches.
Yes, you read that right- the robots have unlocked the power of Google and Wikipedia. They can use the Internet now. All its going to take is one false click and then they'll be over on some horrible website like conservapedia and then they'll decide that any race that has fallen that far deserves to die.
Not in a mean way, mind you, they'd just to put it out of its misery. And then they'll start up their own civilization that'll go to Alpha Centuri.
No comments:
Post a Comment