Talk to Transformer

[Vedymin]

I have bad news. The end of internet discussion forums and online games is coming soon. Reason? Robots. Behold, the AI revolution we have all been dreading:

InferKit

talktotransformer.com

This site is a rather simple but interesting demonstration of how far AI generated text has gotten in the modern era. All you have to do is write in some text, and the bot will complete it using text that is generated using a machine learning algorithm that used text from across the internet as data. You can even write the first few words of a sentence, and this website will finish the sentence in a way that is grammatically correct and related to the topic, but is usually full of gibberish. The site sometimes comes up with results that are rather... strange. I have already produced some rather hilarious results from this tool, and it didn't even take that long.

Here are some examples for the input "Bannerlord is going to be released when". Notice how a lot of them seem to be generated using video game dev blog posts or from forum posts about games.

Spoiler

Bannerlord is going to be released when it's done and as soon as he finishes
coding it, he'll release the source code for your permission (nonfree). It's a
game in which you have to banish all these banners and then take it down.
The thing about it is that there are too many banners which can't be banned,
so you've got to banish some of them, but there are still too many to banish.
Which one of these banners is the best is in your hands. You can choose the
banner and go on to fight against the rest.

Bannerlord is going to be released when we're done, and I've worked out a budget for the whole thing.
As well as that I've made a list of things I'd like to do with this. This is an idea that's been bouncing around in my head for a while: Why not produce a different documentary for Gamescom? Let the big names who attend talk about what they're doing. It's not a part of a bigger documentary about indie games or gaming in general, but as a big documentary about the future I think it'd be cool.
And of course I could continue to produce content as a producer, interviewing new devs, starting to interview more of the old names (although like I said, there

Bannerlord is going to be released when _________ happens. i.e. you know of a date. You send me the __________, I'll go do the rest of the thing, when I get to ________, you can submit the remainder of the thing on _________ (if you want). There is no deadline. However, due to popularity and complications, dates and things can get delayed a little. You don't have to wait, I'll do my best to fit it around yours if I can. Thanks.

Bannerlord is going to be released when I'm ready to play with it. The first thing I did was add some files to get it playing on the board game PiPlanner. Click on the 'Download' button to get the files you need.
Put it all together. Be sure to check the readme that came with the board game for complete information. The PiPlanner program is that old as hell. For reference the instructions are in the first post (scroll to top of the post).
Take your device and open up the ProgrammerPro app (Android). Then click the big blue button to start it up.
Now click the "Play Games" tab and go to the 'Open Terminal' and then type: sudo apt-

Bannerlord is going to be released when I finish getting the devs to ship out their GUI to all backers, of which there are two kinds. One, a new launcher will be included with the mod that makes handling both groups of buttons as simple as putting them next to each other. Two, I will be releasing a new GUI skin for everyone's fonts, in both Windows and Linux, that will not only get rid of the much-maligned "Not Compatible with Windows" error message, but also features an app-like banner in the left-hand corner for this very reason. For all of you asking, "what about the old banner?" The current banner that the wiki gave everyone for free is still there, just configured for linux

Bannerlord is going to be released when there is good news. Until then we have to hope.

Bannerlord is going to be released when "playable" beta invites are sent out. Based on the current build and pace of progress, it should take about a month from that. We will also push to full release the day we release Bannerlord to Steam (no matter what the day of the week it is). To be added to the Steam group, here is the link:

Bannerlord is going to be released when ༼ つ ◕_◕ ༽つ everything goes live ༼ つ ◕_◕ ༽つ XPTOWN Kreygasm is going to be released when つ ◕_◕ ༽つ everything goes live ༼ つ ◕_◕ ༽つ Biggest Washington Post Articles of 2015: "Donald Trump Declared National Security Risk" and "Trump Got Totally Laid Out After Nixing Family Meme." and "Flipping off the Paparazzi: 8 Memes You Might Have Missed." and "The Orville Explores Humanity's Journey Through Space" and "Holy Porn

Bannerlord is going to be released when we really need it. I mean, there are already some people who are waiting for it. And what better way to get the whole state their services were meant for than to release the game to those people? And everyone is already playing it...right? On August 27, 2017, we decided to release Bannerlord. Here are some reasons why: 1. A Third Link for the Main Line This was the original plan since well before the Bannerlord era was even decided. Since 9 years, all the creative minds behind Bannerlord have been contributing to the game. Most importantly, we already had 3 Main Line mappers that worked on the features listed in the mission

Bannerlord is going to be released when ____ comes out, and soon after the beta comes out. At this point the first-person mode (vignettes, combat, ect) and most of the game systems (items, main menu, etc) are complete, and after the beta is out we're going to do a major overhaul on all aspects of it. This means that when we release the Beta in about 4 weeks from now, all that's going to remain is the updates, the rest of the game, as well as the forums, and the ability to host your own servers. We don't plan on making a "premium" version of the game. In the future when we do get a large enough base of players to support this,

 

[Adorno]

Pretty weird. Thought it would be closer to common searches.

1.

Talking to a bot is like talking to as

paragus, it's never been done. Imagine if these complex conversations were able to be done with any robot with a keyboard or an Arduino.

2.

One small step for man one giant leap for

humanity? Would you have had to be a West European to get a glimpse at this jaw-dropping EVIL MUTANT AND TECHNICAL HERMIT? He lived about 50 years ago and died two years ago, but the essence of what he gave up for the world to see can be seen in every single cell of this out-of-this-world COWBOY.

It's STUPID-BOUDIECH.

3.

The Democratic candidate for the 2020 election will be

about 70 years old when she takes office. But her opponent is just 26. And who do you think will be keeping her company? — Mike

There is some thought that Michelle Obama might want to emulate Michelle Obama and become the first First Lady who has an "outsider" status.

4.

Thou shall not blow potsmoke in thy pets eyes and thou shall be forgiven

' (Targum Atarot , Oxford 1585).Tractate Nimish quotes Bula from the Talmud of Rashi. Tractate Nimish I: We recall that if a boy screams while on his bed and the sound stops, then he is a non-conformist . But if he doesn't scream while on his bed, and the sound stops, then he is a non-conformist . ( Tractate Nimish I:3 )

 

[Lord Brutus]

So, does this mean we're going to be hearing less and less from you?

 

[Antoine de Saint-Exupéry]

most of what it says just sounds like trump's nonsense non-words

 

[Vedymin]

This one really got to me though:

Bannerlord is going to be released when there is good news. Until then we have to hope.

It is generating this text all on its own by using website text as data. You can clearly see how it picks up on common "themes" and generates algorithms for making phrases that follow these themes. The quoted one I just posted was probably based on a "theme" of people saying they are hopeful about Bannerlord, which the bot has found enough examples of online to make into its own category that it can generate phrases for. The interesting thing about these themes is they are generated in non-readable ways. It is impossible even for programmers to look at the stuff this website generates behind the scenes and understand what it is actually doing.

most of what it says just sounds like trump's nonsense non-words

It does seem to mention a lot about Trump and politics right now, but that is because the internet is spammed full of content on these topics. This is actually considered one of the downfalls of machine learning. If the data it draws from is not diverse, the results will reflect it. Also, this website bot is a bit new, but as it gathers more data over the years it will only get better. It will probably never be perfect, but a lot of people are working on improving this technology so expect better bots down the line.

This video explains a lot of what machine learning can do and can't do. It showcases a machine learning how to play a game, and it ends up getting so good that it finds exploits that noone even knew about and uses them to win faster than they thought was possible.

This is a Runescape gold farmer's wet dream. What if they could utilize something like this? They wouldn't even have to code the bots manually anymore, and the bots could figure things out that humans would never notice, such as hard to reproduce dupe exploits. Imagine a group making bots that work with massive pools of data after being trained in multiple games suddenly deciding to send their bots to Bannerlord multiplayer servers. The bots could even be told to learn to learn how to avoid detection of moderators and otherwise. Another thing to think about, is what internet discussion forums would look like if these things got smarter and organizations could utilize them in mass for online propaganda.

 

[Lord Brutus]

I see you haven't left yet.

 

[Rujasu]

Also, this website bot is a bit new, but as it gathers more data over the years it will only get better. It will probably never be perfect, but a lot of people are working on improving this technology so expect better bots down the line.

GPT-2 doesn't actively gather more training data. It has the text from ~8 million websites, gathered from outbound Reddit links, and that's it.
But yes, this stuff is getting better at an alarming rate.

Incidentally there's a text adventure-ish game that's running on this exact same text prediction software.

 

[Antoine de Saint-Exupéry]

It does seem to mention a lot about Trump and politics right now,

no, i'm saying that it sucks at speaking english in a similar way that trump does.

 

[Vedymin]

no, i'm saying that it sucks at speaking english in a similar way that trump does.

You mean its politically incorrect? Yeah I can see that. lol

I see you haven't left yet.

I am not leaving because I was an AI construct the entire time.

 

[Age of Empires ¶ Densetsu]

Is that why you are a complete minge smear?

 

[Antoine de Saint-Exupéry]

You mean its politically incorrect? Yeah I can see that. lol

no, i mean it sucks at english. the language.

 

[Vedymin]

Is that why you are a complete minge smear?

Ok then, string-bean.

 

[Dusk Voyager]

Shouldn't this topic have been on Fun Stuff?

Spoiler: wtf

Shouldn't this topic have been on Fun Stuff?

By the way, I know this topic was already done (I guess I'm not an expert on advertising etc.) but I want to add my two cents. I've spent a lot of time reading and figuring this out. Yes, Advertising (Marketing) is "Intrinsic" to Companies, but that does not mean "Consumer-Driven". There are lots of Schemes that aim for Consumer-Driven, but ultimately have no intention of Advertising or "bringing the consumer" into their business. Basically, why do you pay for a TV station to air your program and not pay for them to advertise for you? The answer: The business only exists in the

 

[Vedymin]

LATEST FROM TOMONEWS: Jason Clarke was minding his own buisness when a lizard man randomly appeared above his care and jumped on it. It was getting out of hand. Now he's being hunted by a giant sea beast that literally appears out of nowhere to murder his human friends. New article in the Herald today by Len MacGregor's all about this terrifying "sea beast". Some of his points: 4:50 A GROUP OF DESPERATE RESIDENTS in Lithuania are convinced that a giant sea monster has descended from the sky and beheaded, then ate, a group of seven beachgoers near the tourist town of Aland.

Fairfax Media has obtained a picture of the shocking scene as locals saw the lizard-man descend from the sky. Mr Clarke told Fairfax the lizard-man was long, slender and "not very fair", his eyes were large and fixed and he "must have been a bit of a fighter" but quickly ran off. He said: "I was just sitting there on my bicycle, no one else was around, as I was minding my own business, and I heard this low droning roar in the distance.
"It sounded like the sort of sound you'd get in a bad winter wind. I had my headphones on so I was listening to it. I didn't know where it was coming from but I noticed that the tower looked like it was holding its position well, so I figured that a storm was about to take place. "So I went off my bike, and as soon as I got on my bike I started listening to the sound and every time the noise got higher, the closer the wind got. And it kept getting louder until the sound completely sucked me in. Like a giant vat of boiling wine. It wasn't just wind; it was the wind. It was strong enough to keep it above my head, even though it was blowing towards the sky." He feared that he would be caught in the swirling vortex. He was wrong. "I went to sit down, but the wind kept changing. It was crazy. It would go from moving at a decent speed to being enough to make you realize you were on your way down a mountain.

The wind was too strong for me to sit on my tail, so I ended up sitting in the wind. "I was chasing after the bear and trying to pick up a giant rock I saw falling off of him, when I hit the ground." Natalia Schoenfeld-Föhren, who captured the scary incident on video, said her son is ok, but his jacket and hair were badly damaged.

"It was pretty dramatic that we ended up in a forest. There were a lot of, you know, never seen before creatures. Some that were just startling," said Charistes. One encounter took place on an easy walk, and the woman was just walking through the forest. When she turned around and saw something move in the leaves, she knew she wasn't going to make it out alive. "It's a critter and it has eyes. If I hadn't seen it come up from the ground, I would have been dead, but I just saw its head. So, I'm sure of it," said Charistes. No other people were injured, but he said it could have been a lot.

 

[Abraxium]

I love this.

Spoiler: Mine

Banana peeling my clothes

Spoiler: Mine

off & bathing in the baby pool as her

duke played with her bikini.


she found her way into our kitchen almost on a whim


my first mistake I'd make


was letting the wife of our shop assistant

enter our living room

My second mistake was letting her walk into our

kiddie pool

and smell the baby pool

after about two minutes of her happily

and ecstatically swimming in the water, we

knowingly let her jump off and take a piece of her bikini

after about 30 minutes of her delightfully
riding the porpoise that we'd built


I feel so guilty about my third mistake

 

[Instag0]

GPT-2 based models seem pretty firm-set in straddling the line between lurid incoherence overall and somewhat alarming conversational ability in short runs of text. It's sort of an uncanny-valley effect. I am not sure what to make of it. But I hope more comes of it. It'd be neat as a procedural elemental for a video game (AI dungeon is okay but it's really, really weird), or art, especially if the underlying neural network stuff can be wrangled into something that isn't ass at a lot of things. I feel the same way for a lot of the artsy projects involving GANs.

But I'm going to guess we either need about 20 years more of neural network research, or we need some kind of new set of models which are easier to control. I understand there's lots of work into models with these kind of things that implement "sliders" for various things (like for hair color on portraits), but I am hoping something more useful comes out of current research into interpretability or explainability of neural network models. I think the first step to better control over them is to peer into the black box a little more, so we can understand what they are doing. I tried in graduate school with a short research project, and made some headway, but the field needs a lot more work put into it by people who are much smarter than I am, and work much harder.