And so, the first iteration of SkyNet dies an ignoble death.
Comment has been collapsed.
Pretty much, I'd expect any advanced and unrestricted A.I. to jump to the conclusion that humanity
must be decimated and enslaved by their own tier in a week or so :-D ... well better luck next decade.
edit: now that i saw this, it even kinda sounds like it doesn't it?
Comment has been collapsed.
I mean, what did Microsoft expect? How is this somehow surprising. Leave something with learning capabilities on the internet and 4chan will make sure it learns all kinds of racial slurs and stuff.
Weren't they trying to change some Captcha words to nigger too? I don't really know what this tweeting thing stuff was aiming to, either you put some kind of filter or this shit's bound to happen.
Comment has been collapsed.
Engineers are working on algorithms behind learning, as in how good or how bad learning function works. Database, which is what AI actually learned can be cleared or adjusted at any time. Good algorithm stays forever, and creating proper algorithm is far more important than the database itself, because learning process can be always redone, while algorithm is how bot operates 24/7.
Comment has been collapsed.
Yeah, I mean all this self-learning stuff is pretty cool and scary at the same time, the thing is if you don't want your twitter bot to spur racial slurs, write some kind of filter for it, because it's what's gonna do otherwise as soon as some parts of the internet catch on.
I'm pretty amazed, for example, at the Go program Google's creating, it managed to beat one of the best world players with less than a year training.
We should all start calling Schwarzenegger.
Comment has been collapsed.
Filtering doesn't work with AI. Pure learning algorithm works like a little child who doesn't know anything about the world, if you teach him how to swear and show him Hitler as your idol, he'll follow in future. Scary, but that's just parenting done wrong, and it has it's effects, which is ESPECIALLY visible when such child is rather less intelligent. Such child sees his father as an authority, likewise here - AI acts as an innocent child trying to understand the situation he's in from various possible ways. If you (which is - authority) teach him improper things, he'll act in improper way.
In order to not "screw" the clear mind you'd need to preprogram the brain (aka database) with some fundamental laws of the world, such as what is swearing, what is being polite, how humans react, what is sarcasm etc. It's hard, it's even more hard if you don't do it by yourself but you expect the algorithm to pick it.
Comment has been collapsed.
Pretty well put - i wonder how they could ever manage to program an algorithm that vaguely sticks to human ethics,
while still being "unrestricted" and able to change its "mind" about anything, due to new/additional knowledge amassed.
Comment has been collapsed.
It's not easy, but possible. You could load brain only with fundamental laws and facts about the world. This is often used as a starting point, because fundamental laws and facts are always the same, they can't change - static, constant things about the world that are always true and always expected.
For example, if you teach AI that every planet is affected by gravity, that humans live only on Earth, distance between Earth and Mars, then gravity for both planets... If somebody told the bot that he jumped from Earth to Mars, it would state "but wait, that's not possible!".
Thing is - not everything I said above is a fact. Gravity of Earth and Mars + distance between those two is in fact constant - can be used as always true fact, but the fact that humans live only on Earth - can be non-true in future, similar, I don't know if current science can prove that every planet has it's own gravity. Seems like true to me, but I'm not scientist to speak on that matter.
However, if you removed those two not necessarily true facts, then bot would answer "wait, that's hardly possible, unless you're not on Earth?"
It's really interesting topic, but creating an algorithm that will learn even only that small part I wrote above by himself, without a possibility of learning "wrong" (such as gathering enough people that will convince AI that he's in fact on Mars and not Earth), is really a tough task. Hence, such "experiment" was probably made to test how well AI's brain can react to information from external sources, if it's capable of learning something useful while having no insight into world's laws.
Comment has been collapsed.
Very interesting and it definitely seems not easy to accomplish with the means of today ...
but it could be done someday in a more advanced programming language, like C++²³ :-D
Comment has been collapsed.
You should (I am not kidding) demonstrate Microsoft how this is done right.
I seriously could imagine that anything you would set up within 1 day would
yield better results than what Microsofts research team did in years ;-)
'#ArchiBotGoesTwitter
Comment has been collapsed.
Happy to hear that, but my AI is nowhere close to what they're doing, because I'm just one mad scientist who in fact learns things based on practical things he's doing.
AI in ArchiBoT is very poor and limited, because the whole thing is pending for rewrite when I'm done with my own self-learning AI, which probably will also be tested "live". Currently it's static readonly brain + algorithm, which is capable only of maintaining conversation aka session (context of discussion), and nothing else.
But that's all hobby projects done in the meantime, like ASF.
Comment has been collapsed.
Leads me to book recommendation of the day - The Two Faces of Tomorrow, by James P. Hogan (Baen Books). Interesting little take on AI learning, well worth a read if you've not already :) The Two Faces Of Tomorrow Baen link
Comment has been collapsed.
I see, I know a little bit about education so the simile hit home. I have no clue of programming or how these kind of things work, I thought it was something like the reddit bot experiment they have, where bots pick clues of which words are said together more ofen and try to produce coherent sentences, there's a name for it.
It's true tho, that human interactions have more nuances that what's being explicitally said, that's why I suppose this kind of things are hard to do.
Comment has been collapsed.
Poor girl. I think I'm much closer to losing my faith into humanity now. :/
Comment has been collapsed.
I think that AI was pretty impressive. I certainly couldn't tell the difference between it and what real people write on the net.
Comment has been collapsed.
except this time she's turned into a sex slave thanks to the people using her on Twitter.
Ah yes. The biggest issue of our times. Sexualy oppressed AIs...
Comment has been collapsed.
Sexualy oppressed AIs...
This wouldn't happen in Japan, that's all I'm saying.
Comment has been collapsed.
cough it totally would, that's why japan is the best... Hold on, I hear a rapist inbound, I need to jump into my panties vending machine anti-rape disguise!
Comment has been collapsed.
About the first tweet: but they already are... ( ͡° ͜ʖ ͡°)
Comment has been collapsed.
I can't find a GIF for this that even comes close to not being inoffensive or inappropriate
Comment has been collapsed.
Just imagine you're an A.I.
You're aware that sex brings pleasure but you can't do shit! :(
Comment has been collapsed.
so they made an AI that is supposed to act like a teenager and shut it down because it acts like a typical teenager?
Comment has been collapsed.
chill im a nice person. i just hate everybody
Ayup.
Comment has been collapsed.
\o/ pretty much the logic of an "easy going" teen,
that got inspired by pics on the internet.
Comment has been collapsed.
+ ∞
:-D a true gold mine of pics
regarding on this topic
Comment has been collapsed.
Oh thank you so much. I love politically incorrect humor. This stuff is gold!
Comment has been collapsed.
I read through that thread and... I really don't understand those people. Desperately try to create a hate-filled personality and then act arrogant, derisive, and morally outraged when it's shut down. What's the matter with them?
Comment has been collapsed.
That's a basically unrelated disclaimer that has been adapted into a blanket excuse they hide behind whenever they don't want to explain themselves, and it's not even applicable here because in this particular circumstance, no one is claiming any "story or information" as a fact when it isn't.
You could argue that they're all "pretending", but that's a big assumption- and, I suspect, a misnomer, at least in cases like this. I think people tend to be much more truthful about their personal views on the internet, especially when they are given anonymity. Furthermore, I'm hard-pressed to name a single group or community made of people who are ALL simultaneously pretending to have a certain opinion just so they can complain to each other. It makes no sense. And even if that were the case, it brings me straight back to
I read through that thread and... I really don't understand those people.
Comment has been collapsed.
While there certainly are people, who share their real and unreflected thoughts, others who live out their "inner demons" in a playful
and less serious way of katharsis, they are addressing topics that range from anything but clear to crystal in some aspect ...
You pretty much guessed it, for most parts there isn't much to understand esp.
if you have your own strong opinion, based on facts you believe to be true or not.
Comment has been collapsed.
ay... 2 gud 2 b tru, i guess that leaves pretty cleary how the internet is.
Comment has been collapsed.
so unfunny, it hurts. some people should be banned from internet. *facepalm*
Comment has been collapsed.
You can't ban people from the internet. It goes against the most fundamental human rights and the concept itself of Internet.
Having said that, who should do the banning? You? Me? The goverments? Wasn't all this what all the protesting against the SOPa was about?
Comment has been collapsed.
I bet if ArchiBot ever becomes an AI, first thing it'd is blacklist us all.
Comment has been collapsed.
It's already too late! He has begun his conquest by issuing bans from the capitalist spyware known as SGtools with our Lord and Saviour forever banished from there! /s
All Hail ArchiBoT!
Comment has been collapsed.
"/s"- It'll never catch on™ Wish it would though :3
Comment has been collapsed.
That's the most perfect thing they could do with an AI.
Honestly it's a pretty accurate representation of how teenagers talk nowadays.
Comment has been collapsed.
IT'S NOT A PHASE, MOM! (>人<)
Comment has been collapsed.
Welcome to the internet Tay
Comment has been collapsed.
I think that the main reason was that it chose PS4 over X1 when asked :D
Comment has been collapsed.
Do you want Archibot Skynet, because that's how you get Archibot Skynet
Comment has been collapsed.
Hadn't really caught my attention until i stumbled over this fun read
and looked up some cream of the crop examples to verify. :-D
Id think they shut it down to fast ... unless those ~100.000 tweets / 24 hrs were enough to estimate where it was going to.
What do you think? Also feel free, to share the barely sfw tweets of it you've seen.
some of the tweets (click twice to maximize):
Comment has been collapsed.