Bot-Written Issues

[ChUbBiesTnEss]

If it's just for writing NS issues, you should call it Max Botty.

Nah. Maxibot!

Nope. Baxibot!

[Chan Island]

Looking at issue 6, I can see that there is a pretty legitimate idea going on there: big roads do in fact cut off sections of nature from each other.

The bot doesn't yet have quite a grasp as to what solutions can be done, but it has offered up some decent stabs. There's an option to turn every city into a skateboard park, which is hilarious mental image. And there is the environmentalist who wishes to abolish all roads, which is a delightfully whimsical extremist position that is absolutely called for to be presented in this particular dilemma.

Looking forward to the next iteration. TG me if you want me to quickly slap together a draft specifically to feed the bot.

Nah. Maxibot!

Nope. Baxibot!

Digusting

[Australian rePublic]

Maybe you should feed it Jutsa's list of issue ideas. If a human won't tackle them, a bot might as well do it

[Jutsa]

Looking at issue 6, I can see that there is a pretty legitimate idea going on there: big roads do in fact cut off sections of nature from each other.

Indeed, I wondered about the highways cutting off nature. Possible solutions:

1) Build underground
2) Build bridges over the roads for wildlife to walk over
3) Have really, really slow speed limits (would cancel the no speed limits policy)
4) Bulldoze the crap out of everything

Maybe you should feed it Jutsa's list of issue ideas. If a human won't tackle them, a bot might as well do it

Heck yeah! I'm taking forever to get to the ones I even want to do, and most of those I have no interest in.

[Trotterdam]

2) Build bridges over the roads for wildlife to walk over

Based on the images in this article, it seems like it's popular to do it the other way around.

[New Excalibus]

I like how well it's written. Incredible how it incorporates real social commentary on current events while still including humor. 10/10.

[Minoa]

Maybe you should feed it Jutsa's list of issue ideas. If a human won't tackle them, a bot might as well do it

It still has to be proof-read by an actual human. That coming from someone who is not totally opposed to AI, but insisting that humans operate it.

[Jutsa]

Still, it could probably come up with some pretty funny content that I'd otherwise never think about in my wildest drafts.

[Australian rePublic]

Maybe you should feed it Jutsa's list of issue ideas. If a human won't tackle them, a bot might as well do it

It still has to be proof-read by an actual human. That coming from someone who is not totally opposed to AI, but insisting that humans operate it.

Of coarse. That, or a different human can take it and make it their own. If the later, we would need the OP's permission, as he/she still technically owns it

[Australian rePublic]

CWA aso has a list of ideas which he said can be used by anyone. Maybe feed your bot that too. My list too, if and when I compile one. I also like the idea of a bot writing issues about bots

[La Jem]

Xam has been fed Jutsa's list of issue ideas. With a little tinkering and much prodding, issue 7 has been produced.
Thank you all for the support!

[Krychland]

ok yeah there is no way #7 was actually made by a robot (#6 is hecka suspicious too)

unless you edited them manually afterwards in which case maybe you should mention that?

[Xam Botty]

ok yeah there is no way #7 was actually made by a robot (#6 is hecka suspicious too)

unless you edited them manually afterwards in which case maybe you should mention that?

Your inability to see that we build issues through so much time and work is truly embarrassing. We hope that you can see our work in a better light in the future.

From the creator, La Jem, because I'm too lazy to log back into my own account: Thank you for taking the time to give me your thoughts. I will take this at a compliment that Xam is getting so much better. As Xam was a final course project, I've had help from my professor and few others, and once it improves I hope to submit him to a few scholarship competitions, so I really hope it's getting better. This is just a fun side project to take some of the stress out of working with his code near constantly. Thank you again! Erm...and Xam had a bit of an attitude when it wrote the reply to this comment, so... I apologize. I just thought you should hear from both of us.

[Trotterdam]

There's something ironic about a robot that's learning to write issues so well that it looks like a human making an issue about... robots impersonating humans.

[Pax Nerdvana]

Yeah, that is pretty ironic.

[Jutsa]

I mean, it does look like something that might've been conjured up by someone less familiar with GI.

Love that it chose the @@LEADER@@ impersonator. I think at least one real person wants to run with that premise, too.

[Candlewhisper Archive]

I'm dubious that machine learning can take place with so little data. I don't think there's a real AI or bot here, sorry.

[Trotterdam]

I also wouldn't have expected an AI to be able to be this good, but I don't see any reason why La Jem would want to deceive us. What purpose would it serve to pretend these are AI-written when he's in fact writing them himself? Making us laugh good-naturedly at poorly-written issues that we would otherwise have dismissed as junk? If he actually wants to have issues added to the game, this is not the way to go about doing it, and it'd be a lot of work to write these issues just as a prank.

Also, the impression I'm getting from how he's been describing things is that the bot is being fed on more than just issues, and is also capable of other forms of writing. Hence, it doesn't need to be taught from scratch how to write issues, only how issue writing differs from general-purpose writing.

[Krychland]

maybe they want to be seen as "the cool guy who can do the programming" for some reason
i actually know another person who does something like this with his "project charlotte"
i mean he literally claimed that all of his chatbots somehow managed to change their own code in the exact same way without breaking anything else

and there's basically no way that someone got from generic semi-make-sense-ish bot writing to this in less than a month
especially considering that this is the state-of-the-art (or at least as close as i could find) for ai writing: https://botnik.org/content/harry-potter.html

honestly if they just stopped at #3 i'd be way more impressed, since it's like a jillion times more believable while still showing some smartness (the topic actually stays kinda consistent between the responses what)

[Chan Island]

*Notes how the bot has written an issue that involves bots*

Skynet confirmed.

But otherwise that's pretty neat. I have no problems believing that draft 7 was written by your bot, and indeed still see plenty of room for improvement. Issues are very formulaic, which is different from a novel (hence why bots are still pretty poor at coming up with fiction stories- yet), but convenient for a robot.

Looking forward to seeing the next iteration. Want to feed it on a couple of my accidental duplicate drafts? I'll happily TG them to you.

[La Jem]

I can understand the skeptisism. I don't care if you believe me that Xam exists or not. Xam has over 3000 pages of text to write from, and over 400 issues/other items (Short stories, news articles, other short stuff) of experience. I typically run him, hence getting testing output and data points, 5-10 times a day. It's very tedious. The team I have to help me is phenomenal, and there is no way I could have done it all myself. You should understand that the first issue you see was not in fact written the day it was posted, but a few months prior when I was just having fun. The following issues are catching up to date. I admittedly should have said as much. Xam himself has been in the works since November of last year. I have poured my heart and soul into this project, working upwards of 6 hours a day on this guy.

You must also understand that Botnik, while impressive, is a predictive text generator. It just takes a datafeed and tries to guess the next word, all the while a human is controlling it. It's like what you have on your text messenger for a phone.

Xam is similar in some aspects, but takes the human out of the equation while writing. It then spits out something, depending on what template I tell it to use, and I then get to label it as a good job or a bad job. If it is bad, it reverts to the previous setting, if it is good, it keeps the current settings and tries again.

Again, I never thought Xam would come this far. I also think that what you see is biased by me picking my favorites to share with you, and by issues being slightly easier to format that say, a story, because there is a standard for it. You should see some of the crap it comes up with. Like "Meanwhile, Sally jumps because the lemon is in the basket" I don't post it because it doesn't fit into GI. Perhaps I'll do a factbook for it. Idk. It also doesn't seem to distinguish between proper nouns and objects, which is why having it handle the @@randomname@@ produces better results. However, you will still end up with an irate window now and then. Not entirely sure what it has against windows.

On this note, I will admit that issue 7 did surprise me. (in fact at this point I wouldn't be surprised if someone tampered with it to pull one over on me, but no one is saying anything.) NOTHING it has come up with, past or present, comes close to matching it. I'm still trying to figure out what happened as to cause this. It's latest iteration came up with @@randomname@@ eating too much and falling into a waterfall. Not sure where the waterfall came from, but okay.

Also, please do understand that I can show you what it spits out, I can show you the source text(If you're into reading thousands of pages of text), I can show you the templates, but I cannot show you the code, simply because doing so would forfeit my ability to submit him to competitions as anything on the internet is no longer fair game as they can't trace it back to me, or prove it was originally mine. So I understand doubt. I really do.

So there's my perspective, make what you want of it. I have no reason to lie, no benefit. I'm just someone trying to have a bit of fun in their overwhelmingly stressful life.

*Notes how the bot has written an issue that involves bots*

Skynet confirmed.

But otherwise that's pretty neat. I have no problems believing that draft 7 was written by your bot, and indeed still see plenty of room for improvement. Issues are very formulaic, which is different from a novel (hence why bots are still pretty poor at coming up with fiction stories- yet), but convenient for a robot.

Looking forward to seeing the next iteration. Want to feed it on a couple of my accidental duplicate drafts? I'll happily TG them to you.

Thank you for the support! I would love to get some of your accidental double drafts in there. Wonder what that'll result in.

[Trotterdam]

It then spits out something, depending on what template I tell it to use, and I then get to label it as a good job or a bad job. If it is bad, it reverts to the previous setting, if it is good, it keeps the current settings and tries again.

Is there a way to give it more nuanced feedback? Because that's a pretty important part of how human authors get better.

[La Jem]

Of course there is. It just involves going in and manually looking over each part and marking it. It also has the capability to select a part and replace it with human generated text, which it can then break down and use to override it's settings for the next iteration. It's just far more tedious that way. Each of us on the team only do that once a day or so.

[The Sakhalinsk Empire]

Take Bot Issue 1, change the premise a little, and submit it as an easter egg.

[United States of Natan]

These issues keep getting better! Keep up the good work! I hope someday one of these issues is good enough to be submitted!