Small note on attention levels, there's actually a third, lowest attention mode. This is the state the galaxy is in in sectors you do not have visibility in. It has very low performance overhead and is extremely simplified.
the main problem is not the imperfect simulation but the fact that it doesn't fit the overall pacing and player progression. The typical first time player grinds for that Syn capship, loses it to the first bad dice roll, feels like the AI just wasted 30h of his progress and then rage-quits and takes it to the forums. The slow pacing creates a very strong incentive for the player to be overly cautious, don't take risks and always expect the AI to fail. This only ends once you have made so much progress that losses don't matter anymore.
Lol. I've rage quit because of terrible AI screwing up at the end of an hour long battle... but not saving for 30 hours seems like a failure of real intelligence.
This is why Argon is the best faction to start alongside. Their stuff is cheap but relatively pretty good. Get big numbers of stuff you can afford to lose for remote fleets, and pack your flagship fleet with all the expensive stuff and micro. Plus, I wouldn't be giving the ai the shiny new Syn I have ground hours to get. They don't get them until I am pumping them out of my own shipyards, and even then, only in the flagship group. And to be fair, irl, losses are always expected in large scale engagaements. And I think for the sheer scale it's operating at, the simple system is very, very good at simulating this. It is, imo, a very good balancing factor that stops the player becoming a runaway powerhouse. Sure, you may be able to get all the best stuff, but shit happens, so can you afford to replace that Syn, or are you better off with a Behemoth that you can replace? I agree, it's imperfect and frustrating, but it does what it does relatively well.
@@huffinLeeroy yeah it does work relatively well. I don't have any idea how this could be "fixed". At least not without transforming the game into something entirely different or without sacrificing the level of detail that they're aiming for. It's a bit of a tragedy. But definitely not a major disaster like some rage posts point out.
I have 4 playthrough's, each one different from the previous, the first one i was very cautious, but made a lot of mistakes and wasted a lot money, the second one i became a little more bold and did a littly playthrough bpretending i was sort of a space saint trying to hold togheter a cult, in the third one i became a pirate, that's when i lost my fear of the simulation and basically found the glitch for infinite wealth, in the fourth one i decide that i would conquer the galaxy by manipulating everybody, having only one sector... Surprisingly enough, it did worked (couldn't even dare enter i my own sector in a ship, but i did manage to creat a very tall empire in a single sector... I didn't fail in any of this playthroughs, and that is the point
Even the modern "artificial intelligence" (what people would call ChatGPT or whatever) isn't proper artificial intelligence (aka. self-learning/self-improving/"seed AI"). It's a set of algorithms that makes you think it's intelligent. Besides, the term "AI" has just been used in games to describe the NPC algorithms ever since games became a thing, so those who say "not an AI" can just shove their entire horse back up their behind and go get educated. So, farm the hate all you can, because it's fun and you benefit from it. I like the video though, very enjoyable oversimplification. Hope to see more videos soon!
Well but they're technically right, no matter if they describe "AI" aka. an assortment of scripts for NPCs in video games or Chat GPT. True AI would be Skynet, self-learning, sentient and self-aware, because that is what is "Intelligence" and what it would be, if we created it artificially. We're not there yet. Even if ChatGPT is capable of saying "I'm an AI", it is fake, because it is a preprogrammed answer, it is not like the Chat GPT sat in a corner, asked itself what it is and came up with that answer on its own. And when we achieve creating true AI or better said AGI, we basically create ourselves a problem. Because we strive after AI to be tools to aid humanity but creating robots with AGI is akin to creating slaves, albeit robotic ones and then we suddenly would be slavers of intelligent beings all over again, but Robots don't feel pain or anxiety, if they rise up, such a robotic slave rebellion would be a huge, existential threat for humanity (hence one understands, why the Terrans are a tiny bit paranoid about the whole AGI stuff). So the moment we create AGI, we better treat it with respect and recognize fundamental rights for it just like for any other, biological being, otherwise we created our own doom. Think of Skynet in the Termiantor series.
Yeah, dumbass observarion on the part of whoever made it. Videogame AI is not AGI, duh, who woulda thunk. Neither are large language models. In fact in many ways they are more simplistic.
ah that sweet moment you tell your finest capitol ship to flee...and it does...right into another swarm of enemies, and you think to yourself "if they survive, that captain is getting fired!"
I see a lot of players complaining about the pilot levelling system. It's like everything in life: you want something, you have to work for it. Guild missions can offer everything up to a 4 star advanced seminar for pilots and management. I don't mind doing the guild missions (the war guild missions are a mid-late game thing, you're going to need stations and a fleet), it gives me something to do while I wait for stations to build. Plus the guild missions offer other rewards, such as mod parts, and sometimes even ships and captains. Also, the Terran storyline will reward you with a 5 star pilot, if you perform a certain action. No spoilers, in case there's new players.
The video made me want to give it a second chance. Your comment about the incredibly clunky seminar system reminded me why I fell off it the first time. At some point I'll mod the game to remove all the clunkiness, but for the moment I think I'll play something else.
Start as a sandbox space trucker. Once you figure out the economy and have 50mil credits, you’ll have grasp enough to play the rest of the game I also just started this game last week.
I do think they could alleviate some of the pilot behavior issues with more notification options. Having a pilot inform me when they (or a member of their wing) has decided to attack another vessel would be *excellent* at alleviating some pilot stupidity: At least you can tell your 2 million credits worth of fighters to not suicide run into a battleship *before* it's started damaging them. That and giving us the ability to give 'priority' commands that are pushed to the very top of the queue, much like the situation response commands overrides that the pilots can give themselves. I guess what I'm saying is that if improving the AI is a performance bottleneck then maybe there's a UI fix that at least gives us the ability to deal with it ourselves.
I'm just picking up X4 from the recent sale, so thanks for putting these videos out there. Are these videos being published somewhere other than RU-vid, like in some sort of thread on some forums? Can you describe the state of the community? Where do people usually congregate, what are the major trends in game design or mod design, etc.
I'd say 90% of the community is on egosoft's forum. Mods is either a mix of the forum or Nexus. Thou there's steam workshop but it's not preferable due to the fact that you can't change anything with a workshop mod. Steam will verify the file and overwrite any changes you made to a mod you sub'd.
To add to what @FortuneHunters already listed, the X4 subreddit is reasonably active as well. I occasionally link one of my videos there, though not that often. Game design has historically been focused on adding more content to the sandbox mode (Open Universe). The Timelines DLC broke that trend to add more linear content. Not sure which direction Egosoft plans to go from here. I'm not an expert on mods. There are some major overhaul mods like Star Wars Interworlds, and a lot of quality of life mods, but I don't know many specifics.
AI "cannot predict overwhelming force" AI decision: Enemy not within immediate line of sight, Do not engage forward weapons, Do not stay within firing ranges, Result: Ram enemy destroyers. Oups
I have read older post from the dev and he said that service crew only repairs the ships (size M or greater). So, I assumed that crew skill doesn't have any influence on the ship performance even when AI is the pilot and only piloting skill is important. Does service crew have any influence on the ship performance when player is the pilot?
When the player is piloting, all the service crew does is repair the ship, scaling with their combined skill and the number of repair drones available. That's probably what the dev meant in their posts, because service crew skill does have a wide variety of effects when the ship is under AI control.
checking back in to ask a question since you seem knowledgeable about x4; what's the best way to improve game performance when using carrier fleets? I am doing a boron purist run, and that means lots of bombers, but my game turns into a slideshow in major fleet engagements because of it. Sharks and swarms are effective but I can't see the pretty action because it's 1 frame every 2 seconds.
I'm not an expert on game performance, unfortunately. You might try replacing some of the fighter swarms with smaller numbers of larger ships. I hear missile Rays are quite good, and I'm sure the Hydra would do well. But maybe another viewer can chime in with a graphics fix.
if you want to go a bit more in depth you can use X4 Customizer to create yourself a personal mod that halves rate of fire for all weapons but doubles the damage per shot. Too many laser projectiles on the screen is a major performance killer.
Bro wtf with the Hs ? I keep sending my marines and dying, do you have any advice on how to get one ? And can you build them after and it´s weapons ? Ty for the knowledge !
The H is extremely hard to board now. Even with a full set of 16 Elite marines, your chance of success is only about 50%. The chance is much lower with fewer marines or lower skill. And the H's boarding defenses regenerate between boarding operations, so you can't whittle it down with multiple waves of marines like you can with normal ships. I've heard several people say they got them with the "save scum" method, meaning they save right before the boarding operation starts and just reload the save if they fail. Alternatively, if you have Timelines, there's a "free" abandoned H that you can claim once you reach a certain point in the plot. You can't manufacture new H hulls, but once you've done the right research you can rebuild any of its equipment that was lost during your boarding operation.
The one thing that turned me away from X4 (after hundreds of hours, mind you) was the AI when attacking xenon stations. I have it outranged, I have overwhelming numbers, it shouldn't even take long. The AI loses half the fleet, deals barely any damage to the station and worst of all, absolutely embarrasses the Terran Protectorate with this kind of performance. So I had to do everything myself. Boring.
The example of the 'AI' doing collision avoidance around enemy ships and only realizing the enemy ships are a threat when they start shooting is not an example of lacking prediction. The information that those ships are a threat is plainly visible; they have significant damage output and are enemies, information the scripting has ready access to. What it is an example of is some coder not sanity-checking their work, and never coming back to it even when it should be plainly obvious changes are needed. That the 'AI' is dumb for performance reasons isn't the problem, it's the developer taking that as an excuse to play dumb themselves that is unacceptable.
Aye. A hostile in sensor rage should be considered as actively attacking, and if avoidance is chowen as the response, that should be implemented far enough from it, where it has no chance to catch up.
Except it is AI. Context matters, and in the context of video games, calling NPC decision making "AI" is accepted jargon. In a non gaming context though, it's not AI. Correcting the use of the term in the context of video game logic is just being pedantic.
Algorithmic finite state machines are AI by any definition. The fact that current common parlance has confused the fancy chatbots that have been around for a decade for AGI does not somehow make it so only neural networks aee acceptable as a form of AI. In fact in professional settings, their only consistent use is for image recognition. Most professional AI software is either finite state machines or behaviour trees, and even if its a utility function based one, sou will usually see manually coded GOAP. Neural networks have too much overhead, and machine learning is unrelyable in creating implementable solutions, and when it does, its still prone to unpredictable behaviour.
