@@ItsMeBarnaby they fixed it using AI

@@Hamanito a profanity filter is not AI.

photoallergic agreed, that was on everyone working on the code not just the one dude. mistakes like his are why you have QA in the first place

photoallergic the problem is it wasn't code, it was data. xml data at that. how do you code review that? a missing branch is not going to break anything, it'll just remove possible behaviors in a non-deterministic pathway. much harder to test and diagnose.

Ignoring fiddly bits about precisely how you test data-driven code engines, the takeaway is that you should have a system in place that verifies that every behavior node in your engine can be reached, and if it can't, it should throw up a giant red flag warning. In "normal" software, this sort of thing gets handled by code coverage tests. Data-driven behavior should have some form of "data coverage" as well.

How do you code review XML data? The same way you code review code. If it's going in your game you code review it, it doesn't matter what it is.

+photoallergic Not mentioning that it has nothing with AI in particular. It's product management, it's version control, it's a QA. As well as his previous case.

As something happening in the game that's hilarious, but imagining it actually happening is utterly horrifying.

little tip on formatting text. Putting new lines in-between points, it's fine, but if you use exactly the same newline spacing every time, it's a bit hard to read, so I suggest either including lines where you go like this Line1 with text EmptyLine EmptyLine Line2 with text or Line1 with text Line2 with text if you mix-up the amount of space between phrases, it's gonna make it considerably more readable. I say all of this because I had a habit of always putting a 1-empty-line, until someone pointed out to me, that if you use exactly the same distance rule everywhere, it's almost as worse as using a 0 distance rule. I hope this makes sense

That's what I assumed was happening with the executioner at first. I figured, 'oh, he feeds the thing, he's probably got a way to check its hunger, and he has to keep it above a certain amount. The pit beast eats sims...'

how did the game run ? You got some status report after each action or regular messages of whats happening ?

@@catlord69 it was very simple, the "map" was a grid of 100x100 or something like that, each thing was represented with a character and a log file of each decision and action performed.

"There can be only one"

I think the dwarves developed the game themselves.

the cat bug was a funny one.

@@beebz666 I remember nicknaming one of my bars "Dead Cat Tavern" because of it

You can make every game panel better by inventing Tarn Adams. :)

I had the same sinister plot myself, but involved Ron somehow learning through sim objects/decorations like journals to find out about the quest and somehow override the ability. It would've made a good story, I guess.

Seems to me like some of them didn't really have any AI stories but went on the panel with whatever they could think of. The organizers really should have checked to see what everyone was planning on talking about.

yeah it was more of the AI guys talking about disasters than AI disarters.

I don't know. The baseball game sure as hell looked like an AI disaster

The amount of likes for this comment is repulsive. A.I. aka Artificial Intelligence doesn't only encompass how the NPCs engage or behave in a world. It's also how the game itself interfaces with the player. How it handles player decisions and acts accordingly. It's the construct of the game's intelligence. The first one deals with the Baseball game programmer knowing little to do with baseball, then being given charts or diagrams of what each player should be doing without any real context on WHY the AI is acting in that way(the programmer wasn't familiar with baseball as he wished, and no said "bullshit detectors" for baseball AI issues). The translation from 2D to 3d and how there was facing issues, and animation issues for the player and other characters. The second one deals with a person responsible for testing a certain branch or behavior branch commenting out certain behaviors that the game performed (luckily they didn't catch shit for it by their customers). It had to do with new content being added, I.E. New player behaviors or features, fixes, updates, etc. The third one had to do with gestures from the AI that when talking they would blend from one to another. Which in turn saved them time in short timeframe development for the game (3 or so months and little amount of people to work on it). So by blending many different types of things, he saved time with variation instead of making more complex AI for deciding what to do while talking or animating. The fourth one has to do with the AI actually being spent hard time on, doing the actual AI work behind "fog of war" but the customers believing it was faked because they aren't actually able to see that the AI is doing it. The fifth and final one had to do with the AI of Ron the executioner attacking the Pit worm and bypassing certain locks by the insane trait. Which part of this isn't about AI again?

@@gozinta82 I think your delusion of comprehension is repulsive.

Two of them seem to be about AI. The one with the basketball where a whole chunk of actions were commented out and the one with the sims. The others just seem to be about general programming/designing quirkiness.

All of the talks are about AI presentation and how easy it is to reduce the impression of AI by overlooking design and testing.

The amount of likes for this comment is repulsive. A.I. aka Artificial Intelligence doesn't only encompass how the NPCs engage or behave in a world. It's also how the game itself interfaces with the player. How it handles player decisions and acts accordingly. It's the construct of the game's intelligence. The first one deals with the Baseball game programmer knowing little to do with baseball, then being given charts or diagrams of what each player should be doing without any real context on WHY the AI is acting in that way(the programmer wasn't familiar with baseball as he wished, and no said "bullshit detectors" for baseball AI issues). The translation from 2D to 3d and how there was facing issues, and animation issues for the player and other characters. The second one deals with a person responsible for testing a certain branch or behavior branch commenting out certain behaviors that the game performed (luckily they didn't catch shit for it by their customers). It had to do with new content being added, I.E. New player behaviors or features, fixes, updates, etc. The third one had to do with gestures from the AI that when talking they would blend from one to another. Which in turn saved them time in short timeframe development for the game (3 or so months and little amount of people to work on it). So by blending many different types of things, he saved time with variation instead of making more complex AI for deciding what to do while talking or animating. The fourth one has to do with the AI actually being spent hard time on, doing the actual AI work behind "fog of war" but the customers believing it was faked because they aren't actually able to see that the AI is doing it. The fifth and final one had to do with the AI of Ron the executioner attacking the Pit worm and bypassing certain locks by the insane trait. Which part of this isn't about AI again?

For the Omnissiah i just laid out how ai was covered in each talk. Thats all.

For the Omnissiah what do u want? A cookie?

Believe it or not, code testing and review is not usually done in game programming. Often there is only one programmer working on that section of code. Many programmers don't even review their own code. When it runs, they think it's ready. There might be no specification or design document.

It's more than one thing, not just AI. The test cases he was working on probably contained AI, and like Vidio said, the game itself at this calibur is actually responsible for spitting out flags and errors if something is wrong with the base branches. Multiple things were wrong in the disaster, including the game AI

Wow, that's a nice bit of trivia. I didn't know that. Sweet.

Dye your hair lue and then wait a few months. ahahah

Ten years ago

The joke is on you - I couldn't play baseball to save my life, unless it was in a video game.

Were you writing an AI for tic tac toe or something? If you think it's easy to make ai that are too good, I'd love to see you make an AI for chess. Or literally any game for that matter.

@@alan5506 lol I was not talking about board games. but judging by the video I was commenting on, you should already know this ;-)

@@las10plagas Well actually, I briefly saw that they were talking about super old games and some baseball thing as well as many comments that said this wasn't really about AI. So I clicked off. If I understand the little bit I watched, the baseball AI made animation? What did your AI do?

its very hard to make a good ai in games, unless you are cheating... its super EASY to make a ai play better than a human if its cheating...... now remove the cheats, and you probally will need to dedicate almost you a decade to get it good unless you use machine learning.

I've done some hobbyist game development (C/C++, built 2d engines from scratch) and naming of variables, files, and objects is very strict, down to capitalization and punctuation. If you are off by even a single character, it will fail. Sure it's easy to just look it up in a small project, but the more you can remember off the top of your head, the better. In a large project like that, with easily hundreds of thousands of individual objects, I'd use any and every memorization trick and naming convention I could get away with!

Eador. Masters of the Broken World. reviews aren't great but I enjoyed the game a lot when I did play (250h playtime) (I'm aware some time has passed since you asked)

@@AlienAV Surprisingly, I still noticed.

What i found most striking that some of the stories were not AI disasters, but rather accidental AI awesomeness.

Yeah, the one in The Sims was hilarious :)

RON THE BADASS!

the second half of his talk was about AI though.

"P.C terrorism" I must have not been paying attention to the news if P.C types are terrorists. How did I miss all those shootings and suicide bombings by people who think the N word shouldn't be used in polite society!?

reminding you

Yeah sorry about that...

what subtitle

i think the ai is getting out off controll..

*have

The amount of likes for this comment is repulsive. A.I. aka Artificial Intelligence doesn't only encompass how the NPCs engage or behave in a world. It's also how the game itself interfaces with the player. How it handles player decisions and acts accordingly. It's the construct of the game's intelligence. The first one deals with the Baseball game programmer knowing little to do with baseball, then being given charts or diagrams of what each player should be doing without any real context on WHY the AI is acting in that way(the programmer wasn't familiar with baseball as he wished, and no said "bullshit detectors" for baseball AI issues). The translation from 2D to 3d and how there was facing issues, and animation issues for the player and other characters. The second one deals with a person responsible for testing a certain branch or behavior branch commenting out certain behaviors that the game performed (luckily they didn't catch shit for it by their customers). It had to do with new content being added, I.E. New player behaviors or features, fixes, updates, etc. The third one had to do with gestures from the AI that when talking they would blend from one to another. Which in turn saved them time in short timeframe development for the game (3 or so months and little amount of people to work on it). So by blending many different types of things, he saved time with variation instead of making more complex AI for deciding what to do while talking or animating. The fourth one has to do with the AI actually being spent hard time on, doing the actual AI work behind "fog of war" but the customers believing it was faked because they aren't actually able to see that the AI is doing it. The fifth and final one had to do with the AI of Ron the executioner attacking the Pit worm and bypassing certain locks by the insane trait. Which part of this isn't about AI again?

Hi there! I'm the OP. Had to switch accounts a while ago. "Repulsive" is quite a strong word, don't you think? You seem to have a far wider definition of what AI is. Why it seems unfathomable to you that others might view that differently is a bit odd to me and your reaction is quite over-the-top, but okay... First off, I said "the first section that's really about an AI _disaster_ comes up at 20 minutes in of 25 minutes total". Note the "disaster" bit there I took from the title of the video. So I didn't say none of these stories are about AI, but I did say that none of them up to minute 20 are about AI disasters IMO. The first guy basically says "you should have some basic understanding of the domain your AI is supposed to handle". Definitely true, but it's not exactly revelatory. The rest is basically showing how some animations went wrong. To me, the animation system is separate from AI. Sure, if you consider the animation system part of the AI you could call those glitchy animations AI problems. Considering the game is from the late 90s, I'd say those glitches are far from actual disasters. They're not pretty, but ugly animation transitions and sliding feet were basically the default back then. Also, it goes back to my point above: You seem to have a far wider definition of what AI is supposed to be than me. The second guy talks about applying a vulgarity filter in a level editor and how a colleague had commented out a huge chunk of a behavior tree. Both are entertaining and at least the second one has to do with the AI in a game, but it's not about the AI going wrong, which I would have expected with an "AI disaster". The third guy talks about creating an animation blending system. All the AI does is pick gesture sequences, which he covers in half a sentence. _That_ would have been interesting, although it's not a disaster, because it seemingly worked out really well. The rest is about the animation blending, audio recording and that's that? Number four... well, okay, disheartening, but once again: Not a disaster at all. Only the last guy talks about AI actually going "wrong", explains why / how that happened and goes into how they fixed it.

Also, just guessing, but I'd bet the presenter was talking about themself lol

*Civ Ghandi enters the chat*

I imagine the blue hair (and the sneakers) is socially coding something about how designers are not "corporate" or something. Bob from accounting doesn't have blue hair and the CFO wears cap-toe oxfords, so you do something different to stand out as not-them.

Third is clearly visible in the audience. I love it.

nobody cares

agree- if you're reading this, blue hair #1, you're kind of a knob