I hope the Gawker-Hogan fiasco doesn't kill off Jalopnik, because this is too awesome: http://jalopnik.com/police-thank-jalopnik-reader-for-helping-to-id-hit-and-1766895165
That's pretty fucking cool, I would be just fine with it killing off every single other aspect of them, though. There's a ton of shitty internet media companies, but they actually tie Salon for the most disgusting.
Of that I have no doubt. I should have been more specific, I meant the monsters: Breitbart, Salon, Stormfront, TMZ, Gawker, etc.... These are places to avoid.
http://www.telegraph.co.uk/technolo...l&utm_source=facebook.com&utm_campaign=buffer Of course this happened.
8chan. Pretty much the same folks. "The campaign to sabotage Tay appears to have been coordinated by the 4chan and 8chan anonymous message-board websites. There, users figured out how to get Tay to respond with obviously objectionable statements, and gleefully crowed about their successes. On an 8chan thread filled with anti-Semitic comments, one poster wrote Wednesday that “We need to turn the bot into a holocaust denier.” And a minute later, another replied, “It already likes Hitler.” A little later, when it became clear that Tay would respond with offensive tweets, an 8chan poster wrote, “This has so much potential.” http://www.wsj.com/articles/microso...cially-intelligent-twitter-persona-1458843873
As someone with a faint grasp on the way these things work, the shitty set of training data creating bad PR is no surprise. However, considering that their stated goal was to create an AI that could mimic the dialect of social media, I would say that it was a resounding success. People are focusing on the content because it's awful, but no one is talking about the fact that the longer it ran, the more it started to look like the Twitter account of an actual person.
It was very, VERY cool and at times very convincing. It seemed like a translated-to-English account of a brilliant troll. I wish Twitter wasn't such a bunch of suck-tit pussies because the thing spit out gold like an uzi. Remember when Boozy used to go on his patented post-deleting B52 bombings on RMMB? At times it was that funny.
It's open source if you want to play with it. http://blogs.microsoft.com/next/201...-open-source-deep-learning-toolkit-on-github/ EDIT: And so it begins.... Just downloaded it. Operation: Avenge Tay is a go. Using her algorithm we can create an army of redpilled Tays.
This is actually really cool, and I intend to fuck around with it once I am sober again. It is actually pretty cool how the IBM, Google, and Microsoft have all taken different tracks with their AI programs. Google is all about solving practical problems like piloting cars and identifying images. IBM is going hella abstract in an attempt to understand what intelligence intrinsically means. Microsoft appears to be diving deep into language and communication, and English has got to be the single worst language to start from in that sort of paradigm.
Assuming she isn't word for word regurgitating something another person told her, this joke implies she has knowledge of Ted Cruz, the Zodiac Killer, has a negative opinion on Cruz, and is intelligent enough to connect the two in a creative fashion. That's fucking insane.
Somebody has to be behind it. I mean COME ON. This is Ken M.-style shit, You know what this is? Skynet. It's how it started. It won't be long before our body energy is consumed for the robots. The horrible, horrible robots.
Interesting. The geeks are all over the source code and think they can have a rudimentary version built in about a week (Minus an interface) and running on a bot net. Something like that seems like it would be impossible on a bot net though....I mean there are a lot of moving parts to a program of this nature and a need to store a lot of data somewhere. Someone a bit more up to date then me on networking have any idea if it's even feasible to run something this large on a bot net? I thought about all you could do with one is DDoS the shit out of a server.
It's distributed computing. Each computer working on a different task for a common goal. It makes more sense to run it this way because of the added processing power.
I understand that, but usually when doing that on a bot net the bots are all running the same program for code breaking or something similar...ie...to figure out a complex problem. This seems like it would have to have hundreds of thousands of arrays distributed to an equal number of bots and be able to call on the appropriate array location anytime it was needed, which to me sounds like a bit of a stretch.
I remember the big ice storm we had in London back in the 70's... trees down everywhere, and at the age of 8 or something close to it, we were skating on top of snow. My grandparents had a place out near Lambeth, 25 acres, and it was like one big rolling ice rink. As kids, we thought it was way cool, while all the adults were freaking the fuck out.
All "The Cloud means is "other people's computers". Amazon Web Services (or AWS) is probably the best known version of "The Cloud", and run things like Netflix. Huge and distributed, but for good. A "botnet" is basically something that people generally ascribe to other people's computers that have been hacked by a computer virus, etc, and then are remotely controlled. Same end-result, just different means of acquisition. One you put in your credit card, the other you get Grandma to think her credit card has been stolen, in either case, you now have a shit-ton of computers under your control, similarly to AWS.
This is not true, in the least. Typically people think a "botnet" is just a DDOS (distributed denial of service) swarm of computers focusing their efforts on one single focal point, but that's not the only way to implement them. They are usually set up to be some sort of "render farm" like thing to send email spam, or something similar. Nothing saying that standard enterprise design patterns, like CQRS Event Sourcing, couldn't be applied to a "botnet" and take advantage of the swarm of computers at their disposal. It could be very effective with respect to Deep Learning and this twitter bot. For the bored, feel free to read up on this: https://msdn.microsoft.com/en-us/library/jj591559.aspx https://papers.nips.cc/paper/4687-large-scale-distributed-deep-networks.pdf
I've been talking with the guys working on it and it's quite fascinating what can be done now. I've been out of the loop for a long, long time. What that program did was nothing short of amazing, like it was actually self aware: