The last time I posted about AI, Crashed said I’d hit the nail on the head. Interestingly none of you people have any critique on any content, just a reflexive derision to anything anti Kabam… Very interesting.
No critique because there's no point in providing any since the content of your post shows you dont know how programming works therefore anything anyone would say you would not understand and subsequently deny out of plain ignorance
If you believe that every change that has happened to the AI is unintended then you’re really bad at this game. Have you ever place celestial or play at the top of war? If you knew my work history you wouldn’t make assumptions that make you seem disingenuous.
I’ll simplify it for the people who aren’t being disingenuous. Programming is hard and we don’t expect perfection but to tell us to wait for AI 2.0 while there are major problems right now is absurd. And no not every single AI change has been inintended. That’s not how programming works unless you’re completely inept (which seems to be your defense of Kabam).
It doesn't actually sound like you understand how the AI works. The Deep Dive is a very simplified version, to give the average player a basic overview of what's going on. But you seem to be inferring things no technically competent person would assume in this situation. For example: "I've said on many occasions that the AI appears to be frame perfect (just frame) and this means that 60 times per second (used to be 30 times a second), the AI can read your input and react accordingly." You seem to be inferring that every single frame the AI picks a random action to perform in that specific frame. But that's not how the AI works. In fact, no game AI could possibly work that way.
What's missing from the Deep Dive, because it would mostly cloud the discussion with technical details almost no player would have any use for, is that all game AI systems are state engines. The AI chooses to transition from state to state, and within each state there are valid actions it can perform. Because of that, the choices it can make at any given moment in time change over time - you can even see a hint of that in the animated display in the Deep Dive. The AI is not deciding sixty times per second whether or not to intercept you. If that was the case, it would basically never fail to do so. It would simply have too many opportunities to select that action.
You're also getting hung up on "frame accurate." Most AI systems in video games are "frame accurate." They kind of have to be, and the more simple they are, the more likely this is the case, because as the Deep Dive states, the AI has no memory. As a result, it can only make decisions now based on the things it sees now. In other words, all of its actions involve what it sees in a single physics and animation frame, and by definition it must take action now because it will not remember what it was supposed to do one frame later. Because the AI has no memory, it has always been frame accurate going back to the beginning of the game. That just means it is not programmed to deliberately whiff or make other similar errors. Why is a more complex discussion, but the bottom line is that an AI that did that would be too easy to cheese without sophisticated mechanisms the AI simply doesn't have. It can't remember, it can't predict, and retrofitting such awareness into the state engine would be non-trivial.
The Deep Dive is meant to be a way to communicate the issues to a non-technical audience. But if you're going to use the Deep Dive to make critiques about how the AI is being reworked, you would need far more extensive experience dealing with such systems, enough to infer what MCOC looks like given the limited information the Deep Dive contains.
If you think you understand how the AI works, if you think it just sits there randomly rolling dice until it lands on an action that is possible and then does it, can you explain why the AI sometimes just stops acting altogether? If you think that's the AI just completely breaking, it isn't. The AI is still running, and somehow electing to do nothing. I actually asked the devs a clarifying question about this, because a) the answer should be obvious but b) the deep dive said something else that confused me about this, thus the need to clarify.
The last time I posted about AI, Crashed said I’d hit the nail on the head. Interestingly none of you people have any critique on any content, just a reflexive derision to anything anti Kabam… Very interesting.
No critique because there's no point in providing any since the content of your post shows you dont know how programming works therefore anything anyone would say you would not understand and subsequently deny out of plain ignorance
If you believe that every change that has happened to the AI is unintended then you’re really bad at this game. Have you ever place celestial or play at the top of war? If you knew my work history you wouldn’t make assumptions that make you seem disingenuous.
I’ll simplify it for the people who aren’t being disingenuous. Programming is hard and we don’t expect perfection but to tell us to wait for AI 2.0 while there are major problems right now is absurd. And no not every single AI change has been inintended. That’s not how programming works unless you’re completely inept (which seems to be your defense of Kabam).
It doesn't actually sound like you understand how the AI works. The Deep Dive is a very simplified version, to give the average player a basic overview of what's going on. But you seem to be inferring things no technically competent person would assume in this situation. For example: "I've said on many occasions that the AI appears to be frame perfect (just frame) and this means that 60 times per second (used to be 30 times a second), the AI can read your input and react accordingly." You seem to be inferring that every single frame the AI picks a random action to perform in that specific frame. But that's not how the AI works. In fact, no game AI could possibly work that way.
What's missing from the Deep Dive, because it would mostly cloud the discussion with technical details almost no player would have any use for, is that all game AI systems are state engines. The AI chooses to transition from state to state, and within each state there are valid actions it can perform. Because of that, the choices it can make at any given moment in time change over time - you can even see a hint of that in the animated display in the Deep Dive. The AI is not deciding sixty times per second whether or not to intercept you. If that was the case, it would basically never fail to do so. It would simply have too many opportunities to select that action.
You're also getting hung up on "frame accurate." Most AI systems in video games are "frame accurate." They kind of have to be, and the more simple they are, the more likely this is the case, because as the Deep Dive states, the AI has no memory. As a result, it can only make decisions now based on the things it sees now. In other words, all of its actions involve what it sees in a single physics and animation frame, and by definition it must take action now because it will not remember what it was supposed to do one frame later. Because the AI has no memory, it has always been frame accurate going back to the beginning of the game. That just means it is not programmed to deliberately whiff or make other similar errors. Why is a more complex discussion, but the bottom line is that an AI that did that would be too easy to cheese without sophisticated mechanisms the AI simply doesn't have. It can't remember, it can't predict, and retrofitting such awareness into the state engine would be non-trivial.
The Deep Dive is meant to be a way to communicate the issues to a non-technical audience. But if you're going to use the Deep Dive to make critiques about how the AI is being reworked, you would need far more extensive experience dealing with such systems, enough to infer what MCOC looks like given the limited information the Deep Dive contains.
If you think you understand how the AI works, if you think it just sits there randomly rolling dice until it lands on an action that is possible and then does it, can you explain why the AI sometimes just stops acting altogether? If you think that's the AI just completely breaking, it isn't. The AI is still running, and somehow electing to do nothing. I actually asked the devs a clarifying question about this, because a) the answer should be obvious but b) the deep dive said something else that confused me about this, thus the need to clarify.
This is not a trivial thing to unravel.
So what did they tell (re:last para)? Or can you not tell it at this time?
The last time I posted about AI, Crashed said I’d hit the nail on the head. Interestingly none of you people have any critique on any content, just a reflexive derision to anything anti Kabam… Very interesting.
No critique because there's no point in providing any since the content of your post shows you dont know how programming works therefore anything anyone would say you would not understand and subsequently deny out of plain ignorance
If you believe that every change that has happened to the AI is unintended then you’re really bad at this game. Have you ever place celestial or play at the top of war? If you knew my work history you wouldn’t make assumptions that make you seem disingenuous.
I’ll simplify it for the people who aren’t being disingenuous. Programming is hard and we don’t expect perfection but to tell us to wait for AI 2.0 while there are major problems right now is absurd. And no not every single AI change has been inintended. That’s not how programming works unless you’re completely inept (which seems to be your defense of Kabam).
It doesn't actually sound like you understand how the AI works. The Deep Dive is a very simplified version, to give the average player a basic overview of what's going on. But you seem to be inferring things no technically competent person would assume in this situation. For example: "I've said on many occasions that the AI appears to be frame perfect (just frame) and this means that 60 times per second (used to be 30 times a second), the AI can read your input and react accordingly." You seem to be inferring that every single frame the AI picks a random action to perform in that specific frame. But that's not how the AI works. In fact, no game AI could possibly work that way.
What's missing from the Deep Dive, because it would mostly cloud the discussion with technical details almost no player would have any use for, is that all game AI systems are state engines. The AI chooses to transition from state to state, and within each state there are valid actions it can perform. Because of that, the choices it can make at any given moment in time change over time - you can even see a hint of that in the animated display in the Deep Dive. The AI is not deciding sixty times per second whether or not to intercept you. If that was the case, it would basically never fail to do so. It would simply have too many opportunities to select that action.
You're also getting hung up on "frame accurate." Most AI systems in video games are "frame accurate." They kind of have to be, and the more simple they are, the more likely this is the case, because as the Deep Dive states, the AI has no memory. As a result, it can only make decisions now based on the things it sees now. In other words, all of its actions involve what it sees in a single physics and animation frame, and by definition it must take action now because it will not remember what it was supposed to do one frame later. Because the AI has no memory, it has always been frame accurate going back to the beginning of the game. That just means it is not programmed to deliberately whiff or make other similar errors. Why is a more complex discussion, but the bottom line is that an AI that did that would be too easy to cheese without sophisticated mechanisms the AI simply doesn't have. It can't remember, it can't predict, and retrofitting such awareness into the state engine would be non-trivial.
The Deep Dive is meant to be a way to communicate the issues to a non-technical audience. But if you're going to use the Deep Dive to make critiques about how the AI is being reworked, you would need far more extensive experience dealing with such systems, enough to infer what MCOC looks like given the limited information the Deep Dive contains.
If you think you understand how the AI works, if you think it just sits there randomly rolling dice until it lands on an action that is possible and then does it, can you explain why the AI sometimes just stops acting altogether? If you think that's the AI just completely breaking, it isn't. The AI is still running, and somehow electing to do nothing. I actually asked the devs a clarifying question about this, because a) the answer should be obvious but b) the deep dive said something else that confused me about this, thus the need to clarify.
This is not a trivial thing to unravel.
As usual, this is a ridiculous post with very little substance. In order to understand what they are doing, I'd need to see the code which is obviously not happening. I'm talking as a player not as a coder. Have you seen the code? Or are you just guessing at what Kabam is doing? Eitherway, it's not worth my time to read this.
When my customers criticize my work, I figure out a way to fix it. It must be nice, for Kabam, that enablers such as yourself will rise up to defend them at all costs because obviously, the assertion that we shouldn't have to wait till AI 2.0 is an unreasonable one.
Kabam themselves said that players aren't buying revives bcos it used to be easy and even after the AI became difficult, pro players are using less revives
There's no reason for Kabam to roll back the AI bcos it was too easy back then and even now, it's getting easier for me
I learnt to adapt and play around the new AI
The current AI has plenty of loopholes actually
Yup same here. People who are extra salty about the AI are clearly revive spamming through content, unable to adapt. There are some annoyances for sure, but nothing to get super worked up about like some drama queen...
I’m not tier shaming but wow what a straw man.
The people that really care about bad AI and outta sync frame rates are typically people who play or have played high tier competitive BGs or AW, competitive modes where a single mistake or bugged or Ai manifested error has catastrophic outcomes 9 times outta 10.
If you think it’s people revive spamming through solo content… then I don’t know how much weight anything you have to say on the topic truly has.
OP is arguing about AI being a problem for everyone and we shouldn’t have to wait for a fix. Random people are chiming in saying they do or don’t see a problem in whatever area of game they play at.
While I agree that who you are responding to is a bit too flippant, I disagree with your gatekeeping stance. You are currently arguing that this is a high level BG problem that only high level BG players know what they are talking about, even though OP’s post makes no mention about BGs.
And then if we were to accept the “high level BG” talking point, why should there be a major change for everyone (trying to change the AI) based on the experiences of a niche few high level players of a niche game mode?
If there are two options: (1) Maintain current AI, warts and all, while waiting for AI 2.0, that inconveniences mainly just high level BG players.* (2) Try to fiddle around with current AI to potentially benefit just high level BG players who noticing a problem or break it make the game worse for everyone
As long as Kabam continues to make money and BGs continues to be a niche game mode, I can totally understand which option Kabam pursues.
*Side note: If this is a high level BG issue, every high level player is dealing with the same AI, so it should all end up a wash. You might get light intercepted one time, they might experience it another time.
Kabam themselves said that players aren't buying revives bcos it used to be easy and even after the AI became difficult, pro players are using less revives
There's no reason for Kabam to roll back the AI bcos it was too easy back then and even now, it's getting easier for me
I learnt to adapt and play around the new AI
The current AI has plenty of loopholes actually
Yup same here. People who are extra salty about the AI are clearly revive spamming through content, unable to adapt. There are some annoyances for sure, but nothing to get super worked up about like some drama queen...
I’m not tier shaming but wow what a straw man.
The people that really care about bad AI and outta sync frame rates are typically people who play or have played high tier competitive BGs or AW, competitive modes where a single mistake or bugged or Ai manifested error has catastrophic outcomes 9 times outta 10.
If you think it’s people revive spamming through solo content… then I don’t know how much weight anything you have to say on the topic truly has.
OP is arguing about AI being a problem for everyone and we shouldn’t have to wait for a fix. Random people are chiming in saying they do or don’t see a problem in whatever area of game they play at.
While I agree that who you are responding to is a bit too flippant, I disagree with your gatekeeping stance. You are currently arguing that this is a high level BG problem that only high level BG players know what they are talking about, even though OP’s post makes no mention about BGs.
And then if we were to accept the “high level BG” talking point, why should there be a major change for everyone (trying to change the AI) based on the experiences of a niche few high level players of a niche game mode?
If there are two options: (1) Maintain current AI, warts and all, while waiting for AI 2.0, that inconveniences mainly just high level BG players.* (2) Try to fiddle around with current AI to potentially benefit just high level BG players who noticing a problem or break it make the game worse for everyone
As long as Kabam continues to make money and BGs continues to be a niche game mode, I can totally understand which option Kabam pursues.
*Side note: If this is a high level BG issue, every high level player is dealing with the same AI, so it should all end up a wash. You might get light intercepted one time, they might experience it another time.
It wasn't elegantly stated by Blue. It's more like it's a bigger issue for top level players because it's a bigger pain point. The AI is not a factor when doing Thronebreaker EQ or SQ for a higher level player, it really doesn't matter what happens there. Where it really matters is BGs and War and the higher you are, they higher the stakes. Doesn't mean it's not an issue for everyone (it is), but the reason certain people complain is that it affects the tremendous effort and investment to push for C6+ and/or Masters.
That doesn't mean that it isn't a problem for non-top tier players, it's a problem for everyone and we should advocate for better as customers (many of us paying).
Kabam themselves said that players aren't buying revives bcos it used to be easy and even after the AI became difficult, pro players are using less revives
There's no reason for Kabam to roll back the AI bcos it was too easy back then and even now, it's getting easier for me
I learnt to adapt and play around the new AI
The current AI has plenty of loopholes actually
Yup same here. People who are extra salty about the AI are clearly revive spamming through content, unable to adapt. There are some annoyances for sure, but nothing to get super worked up about like some drama queen...
I’m not tier shaming but wow what a straw man.
The people that really care about bad AI and outta sync frame rates are typically people who play or have played high tier competitive BGs or AW, competitive modes where a single mistake or bugged or Ai manifested error has catastrophic outcomes 9 times outta 10.
If you think it’s people revive spamming through solo content… then I don’t know how much weight anything you have to say on the topic truly has.
OP is arguing about AI being a problem for everyone and we shouldn’t have to wait for a fix. Random people are chiming in saying they do or don’t see a problem in whatever area of game they play at.
While I agree that who you are responding to is a bit too flippant, I disagree with your gatekeeping stance. You are currently arguing that this is a high level BG problem that only high level BG players know what they are talking about, even though OP’s post makes no mention about BGs.
And then if we were to accept the “high level BG” talking point, why should there be a major change for everyone (trying to change the AI) based on the experiences of a niche few high level players of a niche game mode?
If there are two options: (1) Maintain current AI, warts and all, while waiting for AI 2.0, that inconveniences mainly just high level BG players.* (2) Try to fiddle around with current AI to potentially benefit just high level BG players who noticing a problem or break it make the game worse for everyone
As long as Kabam continues to make money and BGs continues to be a niche game mode, I can totally understand which option Kabam pursues.
*Side note: If this is a high level BG issue, every high level player is dealing with the same AI, so it should all end up a wash. You might get light intercepted one time, they might experience it another time.
It wasn't elegantly stated by Blue. It's more like it's a bigger issue for top level players because it's a bigger pain point. The AI is not a factor when doing Thronebreaker EQ or SQ for a higher level player, it really doesn't matter what happens there. Where it really matters is BGs and War and the higher you are, they higher the stakes. Doesn't mean it's not an issue for everyone (it is), but the reason certain people complain is that it affects the tremendous effort and investment to push for C6+ and/or Masters.
That doesn't mean that it isn't a problem for non-top tier players, it's a problem for everyone and we should advocate for better as customers (many of us paying).
Exactly this. 40% revives and 20% revives have never been more widely an available than they are now. (Excepting farming 🪦) so I’m simply assuming that players experiencing AI shenanigans in solo content aren’t the ones getting mad enough to post about it here. Or are more likely to dismiss AI based complaints flippantly due to playing with lower stakes and not understanding why others would be more frustrated since they’ve never experienced the same levels of competition in the contest.
The last time I posted about AI, Crashed said I’d hit the nail on the head. Interestingly none of you people have any critique on any content, just a reflexive derision to anything anti Kabam… Very interesting.
No critique because there's no point in providing any since the content of your post shows you dont know how programming works therefore anything anyone would say you would not understand and subsequently deny out of plain ignorance
If you believe that every change that has happened to the AI is unintended then you’re really bad at this game. Have you ever place celestial or play at the top of war? If you knew my work history you wouldn’t make assumptions that make you seem disingenuous.
I’ll simplify it for the people who aren’t being disingenuous. Programming is hard and we don’t expect perfection but to tell us to wait for AI 2.0 while there are major problems right now is absurd. And no not every single AI change has been inintended. That’s not how programming works unless you’re completely inept (which seems to be your defense of Kabam).
It doesn't actually sound like you understand how the AI works. The Deep Dive is a very simplified version, to give the average player a basic overview of what's going on. But you seem to be inferring things no technically competent person would assume in this situation. For example: "I've said on many occasions that the AI appears to be frame perfect (just frame) and this means that 60 times per second (used to be 30 times a second), the AI can read your input and react accordingly." You seem to be inferring that every single frame the AI picks a random action to perform in that specific frame. But that's not how the AI works. In fact, no game AI could possibly work that way.
What's missing from the Deep Dive, because it would mostly cloud the discussion with technical details almost no player would have any use for, is that all game AI systems are state engines. The AI chooses to transition from state to state, and within each state there are valid actions it can perform. Because of that, the choices it can make at any given moment in time change over time - you can even see a hint of that in the animated display in the Deep Dive. The AI is not deciding sixty times per second whether or not to intercept you. If that was the case, it would basically never fail to do so. It would simply have too many opportunities to select that action.
You're also getting hung up on "frame accurate." Most AI systems in video games are "frame accurate." They kind of have to be, and the more simple they are, the more likely this is the case, because as the Deep Dive states, the AI has no memory. As a result, it can only make decisions now based on the things it sees now. In other words, all of its actions involve what it sees in a single physics and animation frame, and by definition it must take action now because it will not remember what it was supposed to do one frame later. Because the AI has no memory, it has always been frame accurate going back to the beginning of the game. That just means it is not programmed to deliberately whiff or make other similar errors. Why is a more complex discussion, but the bottom line is that an AI that did that would be too easy to cheese without sophisticated mechanisms the AI simply doesn't have. It can't remember, it can't predict, and retrofitting such awareness into the state engine would be non-trivial.
The Deep Dive is meant to be a way to communicate the issues to a non-technical audience. But if you're going to use the Deep Dive to make critiques about how the AI is being reworked, you would need far more extensive experience dealing with such systems, enough to infer what MCOC looks like given the limited information the Deep Dive contains.
If you think you understand how the AI works, if you think it just sits there randomly rolling dice until it lands on an action that is possible and then does it, can you explain why the AI sometimes just stops acting altogether? If you think that's the AI just completely breaking, it isn't. The AI is still running, and somehow electing to do nothing. I actually asked the devs a clarifying question about this, because a) the answer should be obvious but b) the deep dive said something else that confused me about this, thus the need to clarify.
This is not a trivial thing to unravel.
As usual, this is a ridiculous post with very little substance. In order to understand what they are doing, I'd need to see the code which is obviously not happening. I'm talking as a player not as a coder. Have you seen the code? Or are you just guessing at what Kabam is doing? Eitherway, it's not worth my time to read this.
When my customers criticize my work, I figure out a way to fix it. It must be nice, for Kabam, that enablers such as yourself will rise up to defend them at all costs because obviously, the assertion that we shouldn't have to wait till AI 2.0 is an unreasonable one.
Why would you want to see the code? You don't have experience programming for gaming remember? So what would you be able to do?
The last time I posted about AI, Crashed said I’d hit the nail on the head. Interestingly none of you people have any critique on any content, just a reflexive derision to anything anti Kabam… Very interesting.
No critique because there's no point in providing any since the content of your post shows you dont know how programming works therefore anything anyone would say you would not understand and subsequently deny out of plain ignorance
If you believe that every change that has happened to the AI is unintended then you’re really bad at this game. Have you ever place celestial or play at the top of war? If you knew my work history you wouldn’t make assumptions that make you seem disingenuous.
I’ll simplify it for the people who aren’t being disingenuous. Programming is hard and we don’t expect perfection but to tell us to wait for AI 2.0 while there are major problems right now is absurd. And no not every single AI change has been inintended. That’s not how programming works unless you’re completely inept (which seems to be your defense of Kabam).
It doesn't actually sound like you understand how the AI works. The Deep Dive is a very simplified version, to give the average player a basic overview of what's going on. But you seem to be inferring things no technically competent person would assume in this situation. For example: "I've said on many occasions that the AI appears to be frame perfect (just frame) and this means that 60 times per second (used to be 30 times a second), the AI can read your input and react accordingly." You seem to be inferring that every single frame the AI picks a random action to perform in that specific frame. But that's not how the AI works. In fact, no game AI could possibly work that way.
What's missing from the Deep Dive, because it would mostly cloud the discussion with technical details almost no player would have any use for, is that all game AI systems are state engines. The AI chooses to transition from state to state, and within each state there are valid actions it can perform. Because of that, the choices it can make at any given moment in time change over time - you can even see a hint of that in the animated display in the Deep Dive. The AI is not deciding sixty times per second whether or not to intercept you. If that was the case, it would basically never fail to do so. It would simply have too many opportunities to select that action.
You're also getting hung up on "frame accurate." Most AI systems in video games are "frame accurate." They kind of have to be, and the more simple they are, the more likely this is the case, because as the Deep Dive states, the AI has no memory. As a result, it can only make decisions now based on the things it sees now. In other words, all of its actions involve what it sees in a single physics and animation frame, and by definition it must take action now because it will not remember what it was supposed to do one frame later. Because the AI has no memory, it has always been frame accurate going back to the beginning of the game. That just means it is not programmed to deliberately whiff or make other similar errors. Why is a more complex discussion, but the bottom line is that an AI that did that would be too easy to cheese without sophisticated mechanisms the AI simply doesn't have. It can't remember, it can't predict, and retrofitting such awareness into the state engine would be non-trivial.
The Deep Dive is meant to be a way to communicate the issues to a non-technical audience. But if you're going to use the Deep Dive to make critiques about how the AI is being reworked, you would need far more extensive experience dealing with such systems, enough to infer what MCOC looks like given the limited information the Deep Dive contains.
If you think you understand how the AI works, if you think it just sits there randomly rolling dice until it lands on an action that is possible and then does it, can you explain why the AI sometimes just stops acting altogether? If you think that's the AI just completely breaking, it isn't. The AI is still running, and somehow electing to do nothing. I actually asked the devs a clarifying question about this, because a) the answer should be obvious but b) the deep dive said something else that confused me about this, thus the need to clarify.
This is not a trivial thing to unravel.
As usual, this is a ridiculous post with very little substance. In order to understand what they are doing, I'd need to see the code which is obviously not happening. I'm talking as a player not as a coder. Have you seen the code? Or are you just guessing at what Kabam is doing? Eitherway, it's not worth my time to read this.
When my customers criticize my work, I figure out a way to fix it. It must be nice, for Kabam, that enablers such as yourself will rise up to defend them at all costs because obviously, the assertion that we shouldn't have to wait till AI 2.0 is an unreasonable one.
Why would you want to see the code? You don't have experience programming for gaming remember? So what would you be able to do?
If he's not satisfied with DNA's knowledge and experience in this case, nothing will do.
The last time I posted about AI, Crashed said I’d hit the nail on the head. Interestingly none of you people have any critique on any content, just a reflexive derision to anything anti Kabam… Very interesting.
No critique because there's no point in providing any since the content of your post shows you dont know how programming works therefore anything anyone would say you would not understand and subsequently deny out of plain ignorance
If you believe that every change that has happened to the AI is unintended then you’re really bad at this game. Have you ever place celestial or play at the top of war? If you knew my work history you wouldn’t make assumptions that make you seem disingenuous.
I’ll simplify it for the people who aren’t being disingenuous. Programming is hard and we don’t expect perfection but to tell us to wait for AI 2.0 while there are major problems right now is absurd. And no not every single AI change has been inintended. That’s not how programming works unless you’re completely inept (which seems to be your defense of Kabam).
It doesn't actually sound like you understand how the AI works. The Deep Dive is a very simplified version, to give the average player a basic overview of what's going on. But you seem to be inferring things no technically competent person would assume in this situation. For example: "I've said on many occasions that the AI appears to be frame perfect (just frame) and this means that 60 times per second (used to be 30 times a second), the AI can read your input and react accordingly." You seem to be inferring that every single frame the AI picks a random action to perform in that specific frame. But that's not how the AI works. In fact, no game AI could possibly work that way.
What's missing from the Deep Dive, because it would mostly cloud the discussion with technical details almost no player would have any use for, is that all game AI systems are state engines. The AI chooses to transition from state to state, and within each state there are valid actions it can perform. Because of that, the choices it can make at any given moment in time change over time - you can even see a hint of that in the animated display in the Deep Dive. The AI is not deciding sixty times per second whether or not to intercept you. If that was the case, it would basically never fail to do so. It would simply have too many opportunities to select that action.
You're also getting hung up on "frame accurate." Most AI systems in video games are "frame accurate." They kind of have to be, and the more simple they are, the more likely this is the case, because as the Deep Dive states, the AI has no memory. As a result, it can only make decisions now based on the things it sees now. In other words, all of its actions involve what it sees in a single physics and animation frame, and by definition it must take action now because it will not remember what it was supposed to do one frame later. Because the AI has no memory, it has always been frame accurate going back to the beginning of the game. That just means it is not programmed to deliberately whiff or make other similar errors. Why is a more complex discussion, but the bottom line is that an AI that did that would be too easy to cheese without sophisticated mechanisms the AI simply doesn't have. It can't remember, it can't predict, and retrofitting such awareness into the state engine would be non-trivial.
The Deep Dive is meant to be a way to communicate the issues to a non-technical audience. But if you're going to use the Deep Dive to make critiques about how the AI is being reworked, you would need far more extensive experience dealing with such systems, enough to infer what MCOC looks like given the limited information the Deep Dive contains.
If you think you understand how the AI works, if you think it just sits there randomly rolling dice until it lands on an action that is possible and then does it, can you explain why the AI sometimes just stops acting altogether? If you think that's the AI just completely breaking, it isn't. The AI is still running, and somehow electing to do nothing. I actually asked the devs a clarifying question about this, because a) the answer should be obvious but b) the deep dive said something else that confused me about this, thus the need to clarify.
This is not a trivial thing to unravel.
As usual, this is a ridiculous post with very little substance. In order to understand what they are doing, I'd need to see the code which is obviously not happening. I'm talking as a player not as a coder. Have you seen the code? Or are you just guessing at what Kabam is doing? Eitherway, it's not worth my time to read this.
When my customers criticize my work, I figure out a way to fix it. It must be nice, for Kabam, that enablers such as yourself will rise up to defend them at all costs because obviously, the assertion that we shouldn't have to wait till AI 2.0 is an unreasonable one.
Why would you want to see the code? You don't have experience programming for gaming remember? So what would you be able to do?
If he's not satisfied with DNA's knowledge and experience in this case, nothing will do.
I’m glad that Kabam is proactively communicating more often but they are dealing with many players who have a decade’s worth of experience with the AI. Our thousands of hours of focused interaction with the product give us quite a lot of knowledge about what it does, what the AI’s norms are / were, etc. Kabam is going to have to do a lot more than provide a basic primer on how the AI works before we just accept it without question or comment.
The last time I posted about AI, Crashed said I’d hit the nail on the head. Interestingly none of you people have any critique on any content, just a reflexive derision to anything anti Kabam… Very interesting.
No critique because there's no point in providing any since the content of your post shows you dont know how programming works therefore anything anyone would say you would not understand and subsequently deny out of plain ignorance
If you believe that every change that has happened to the AI is unintended then you’re really bad at this game. Have you ever place celestial or play at the top of war? If you knew my work history you wouldn’t make assumptions that make you seem disingenuous.
I’ll simplify it for the people who aren’t being disingenuous. Programming is hard and we don’t expect perfection but to tell us to wait for AI 2.0 while there are major problems right now is absurd. And no not every single AI change has been inintended. That’s not how programming works unless you’re completely inept (which seems to be your defense of Kabam).
It doesn't actually sound like you understand how the AI works. The Deep Dive is a very simplified version, to give the average player a basic overview of what's going on. But you seem to be inferring things no technically competent person would assume in this situation. For example: "I've said on many occasions that the AI appears to be frame perfect (just frame) and this means that 60 times per second (used to be 30 times a second), the AI can read your input and react accordingly." You seem to be inferring that every single frame the AI picks a random action to perform in that specific frame. But that's not how the AI works. In fact, no game AI could possibly work that way.
What's missing from the Deep Dive, because it would mostly cloud the discussion with technical details almost no player would have any use for, is that all game AI systems are state engines. The AI chooses to transition from state to state, and within each state there are valid actions it can perform. Because of that, the choices it can make at any given moment in time change over time - you can even see a hint of that in the animated display in the Deep Dive. The AI is not deciding sixty times per second whether or not to intercept you. If that was the case, it would basically never fail to do so. It would simply have too many opportunities to select that action.
You're also getting hung up on "frame accurate." Most AI systems in video games are "frame accurate." They kind of have to be, and the more simple they are, the more likely this is the case, because as the Deep Dive states, the AI has no memory. As a result, it can only make decisions now based on the things it sees now. In other words, all of its actions involve what it sees in a single physics and animation frame, and by definition it must take action now because it will not remember what it was supposed to do one frame later. Because the AI has no memory, it has always been frame accurate going back to the beginning of the game. That just means it is not programmed to deliberately whiff or make other similar errors. Why is a more complex discussion, but the bottom line is that an AI that did that would be too easy to cheese without sophisticated mechanisms the AI simply doesn't have. It can't remember, it can't predict, and retrofitting such awareness into the state engine would be non-trivial.
The Deep Dive is meant to be a way to communicate the issues to a non-technical audience. But if you're going to use the Deep Dive to make critiques about how the AI is being reworked, you would need far more extensive experience dealing with such systems, enough to infer what MCOC looks like given the limited information the Deep Dive contains.
If you think you understand how the AI works, if you think it just sits there randomly rolling dice until it lands on an action that is possible and then does it, can you explain why the AI sometimes just stops acting altogether? If you think that's the AI just completely breaking, it isn't. The AI is still running, and somehow electing to do nothing. I actually asked the devs a clarifying question about this, because a) the answer should be obvious but b) the deep dive said something else that confused me about this, thus the need to clarify.
This is not a trivial thing to unravel.
As usual, this is a ridiculous post with very little substance. In order to understand what they are doing, I'd need to see the code which is obviously not happening. I'm talking as a player not as a coder. Have you seen the code? Or are you just guessing at what Kabam is doing? Eitherway, it's not worth my time to read this.
When my customers criticize my work, I figure out a way to fix it. It must be nice, for Kabam, that enablers such as yourself will rise up to defend them at all costs because obviously, the assertion that we shouldn't have to wait till AI 2.0 is an unreasonable one.
Why would you want to see the code? You don't have experience programming for gaming remember? So what would you be able to do?
If he's not satisfied with DNA's knowledge and experience in this case, nothing will do.
I’m glad that Kabam is proactively communicating more often but they are dealing with many players who have a decade’s worth of experience with the AI. Our thousands of hours of focused interaction with the product give us quite a lot of knowledge about what it does, what the AI’s norms are / were, etc. Kabam is going to have to do a lot more than provide a basic primer on how the AI works before we just accept it without question or comment.
Not about OP, but experience with the AI really doesn't mean that they can or they will understand how the AI works. There are still people who are rejecting Kabam's statement that AI does not know game effects because of their "experience" with different game effects like power stings and unblockable.
The last time I posted about AI, Crashed said I’d hit the nail on the head. Interestingly none of you people have any critique on any content, just a reflexive derision to anything anti Kabam… Very interesting.
No critique because there's no point in providing any since the content of your post shows you dont know how programming works therefore anything anyone would say you would not understand and subsequently deny out of plain ignorance
If you believe that every change that has happened to the AI is unintended then you’re really bad at this game. Have you ever place celestial or play at the top of war? If you knew my work history you wouldn’t make assumptions that make you seem disingenuous.
I’ll simplify it for the people who aren’t being disingenuous. Programming is hard and we don’t expect perfection but to tell us to wait for AI 2.0 while there are major problems right now is absurd. And no not every single AI change has been inintended. That’s not how programming works unless you’re completely inept (which seems to be your defense of Kabam).
It doesn't actually sound like you understand how the AI works. The Deep Dive is a very simplified version, to give the average player a basic overview of what's going on. But you seem to be inferring things no technically competent person would assume in this situation. For example: "I've said on many occasions that the AI appears to be frame perfect (just frame) and this means that 60 times per second (used to be 30 times a second), the AI can read your input and react accordingly." You seem to be inferring that every single frame the AI picks a random action to perform in that specific frame. But that's not how the AI works. In fact, no game AI could possibly work that way.
What's missing from the Deep Dive, because it would mostly cloud the discussion with technical details almost no player would have any use for, is that all game AI systems are state engines. The AI chooses to transition from state to state, and within each state there are valid actions it can perform. Because of that, the choices it can make at any given moment in time change over time - you can even see a hint of that in the animated display in the Deep Dive. The AI is not deciding sixty times per second whether or not to intercept you. If that was the case, it would basically never fail to do so. It would simply have too many opportunities to select that action.
You're also getting hung up on "frame accurate." Most AI systems in video games are "frame accurate." They kind of have to be, and the more simple they are, the more likely this is the case, because as the Deep Dive states, the AI has no memory. As a result, it can only make decisions now based on the things it sees now. In other words, all of its actions involve what it sees in a single physics and animation frame, and by definition it must take action now because it will not remember what it was supposed to do one frame later. Because the AI has no memory, it has always been frame accurate going back to the beginning of the game. That just means it is not programmed to deliberately whiff or make other similar errors. Why is a more complex discussion, but the bottom line is that an AI that did that would be too easy to cheese without sophisticated mechanisms the AI simply doesn't have. It can't remember, it can't predict, and retrofitting such awareness into the state engine would be non-trivial.
The Deep Dive is meant to be a way to communicate the issues to a non-technical audience. But if you're going to use the Deep Dive to make critiques about how the AI is being reworked, you would need far more extensive experience dealing with such systems, enough to infer what MCOC looks like given the limited information the Deep Dive contains.
If you think you understand how the AI works, if you think it just sits there randomly rolling dice until it lands on an action that is possible and then does it, can you explain why the AI sometimes just stops acting altogether? If you think that's the AI just completely breaking, it isn't. The AI is still running, and somehow electing to do nothing. I actually asked the devs a clarifying question about this, because a) the answer should be obvious but b) the deep dive said something else that confused me about this, thus the need to clarify.
This is not a trivial thing to unravel.
So what did they tell (re:last para)? Or can you not tell it at this time?
Simplifying a bit, I asked a question that boiled down to how the system determines if an action is valid, and the broad answer was that it determines this locally not globally.
It is thus theoretically possible for the AI to enter a state where there is no valid option to exit that state. Most likely, in my experience this happens when an action sequencer fouls with an animation sequencer.
Or to put it more colloquially, the AI has an implicit understanding that certain actions happen in certain orders, but the animation engine of the game (or the mechanics of the game in general) can sometimes fail to obey that in certain weird corner cases. If the game finds itself expecting something to happen that will never happen, it waits forever.
I should point out this sounds like a contradiction, as I mentioned earlier I don’t believe the AI makes predictions. “Expects” is a colloquialism here, it means the AI enters a state where the proscribed possible actions the random weighted table offers have been designed by some human to be what the human believed would be valid options at that moment, because the human then expected that after the AI took one of those actions it would then be put into a different state that opened the door to other options that were logical continuations of the fight. But if the AI cannot do any of those things, it cannot continue the fight and it cannot transition to a state where other options become available. It is simply continuing to do what it was programmed to do, which is to do one of the things it is told are valid options, which in this case is nothing, and continuing to transition to new states determined by those actions, which is also nothing. I’m anthropomorphizing in this as “expecting something and waiting for it to happen.” It doesn’t expect and it doesn’t wait. It just keeps rolling dice that will never come up sevens.
That’s also why we can “wake it up.” If the AI was broken, it would stay broken no matter what we did. But when we act, *we* transition the state engine to a state where the locally valid actions now allow the AI to both act and continue to transition to other new states.
As long as I’m explaining this, I might as well explain the other question I asked (since it echoes things the devs have already said in the past publicly, I don’t think it gives away anything of a sensitive nature). I long assumed there was a set of tuning parameters for the AI, something that would explain how the devs could say there were very few AI profiles but in my experience there were far more AI “tendencies.” I suspected there was some initialization parameters - something where at the start of the fight the game sort of rolled dice to give the fight some “character” and those random initial settings could strongly influence the “personality” of the AI, even if the AI itself was identical from fight to fight. According to the devs, there is no such thing happening. However, they did confirm another one of my suspicions, something I wrote about several years ago when AI first started getting discussed a lot on the forums. Because the AI is a state engine (so I assumed back then), its moment by moment behavior is implicitly influenced by the past., The AI doesn’t have a memory per se, but the state engine does change the behavior of the AI with each action we do, and then it does (randomly), and so on. So just like a Chess game that can be strongly influenced by the opening moves of the game, the “personality” of the AI can, in some cases, be strongly influenced by the very first thing it does, which triggers the players to react in a very specific way to that, and so on. Moment by moment, the AI is very simple. But because it is also a state engine, it can become caught in feedback loops. For example passive AI is often a positive feedback loop, because the AI tends to take actions in response to the player. When it gets passive, the players don’t have openings to attack, so the players get passive. This can then get reinforced to the point where it gets difficult to break out of.
If the state engines were implemented in a generic engine using data, the state engines could be tinkered with relatively easily, at least to a degree. But I suspect it isn’t: I believe the AI state engines is implemented directly in code. And that’s why they have to write a completely new one from scratch: tampering with any part of the behavior involves directly tampering with the AI code. This is a non-trivial thing to do, especially because it is unlikely to be some module you can rip out of Unity. It is more likely to be intertwined with the rest of the engine, because if MCOC is a cleanly modularized Unity project, it will be the first one I’ve ever heard of. Unity doesn’t encourage that sort of thing. If you’re going to do that sort of thing anyway, you might as well write an AI engine that does what you want from scratch.
The last time I posted about AI, Crashed said I’d hit the nail on the head. Interestingly none of you people have any critique on any content, just a reflexive derision to anything anti Kabam… Very interesting.
No critique because there's no point in providing any since the content of your post shows you dont know how programming works therefore anything anyone would say you would not understand and subsequently deny out of plain ignorance
If you believe that every change that has happened to the AI is unintended then you’re really bad at this game. Have you ever place celestial or play at the top of war? If you knew my work history you wouldn’t make assumptions that make you seem disingenuous.
I’ll simplify it for the people who aren’t being disingenuous. Programming is hard and we don’t expect perfection but to tell us to wait for AI 2.0 while there are major problems right now is absurd. And no not every single AI change has been inintended. That’s not how programming works unless you’re completely inept (which seems to be your defense of Kabam).
It doesn't actually sound like you understand how the AI works. The Deep Dive is a very simplified version, to give the average player a basic overview of what's going on. But you seem to be inferring things no technically competent person would assume in this situation. For example: "I've said on many occasions that the AI appears to be frame perfect (just frame) and this means that 60 times per second (used to be 30 times a second), the AI can read your input and react accordingly." You seem to be inferring that every single frame the AI picks a random action to perform in that specific frame. But that's not how the AI works. In fact, no game AI could possibly work that way.
What's missing from the Deep Dive, because it would mostly cloud the discussion with technical details almost no player would have any use for, is that all game AI systems are state engines. The AI chooses to transition from state to state, and within each state there are valid actions it can perform. Because of that, the choices it can make at any given moment in time change over time - you can even see a hint of that in the animated display in the Deep Dive. The AI is not deciding sixty times per second whether or not to intercept you. If that was the case, it would basically never fail to do so. It would simply have too many opportunities to select that action.
You're also getting hung up on "frame accurate." Most AI systems in video games are "frame accurate." They kind of have to be, and the more simple they are, the more likely this is the case, because as the Deep Dive states, the AI has no memory. As a result, it can only make decisions now based on the things it sees now. In other words, all of its actions involve what it sees in a single physics and animation frame, and by definition it must take action now because it will not remember what it was supposed to do one frame later. Because the AI has no memory, it has always been frame accurate going back to the beginning of the game. That just means it is not programmed to deliberately whiff or make other similar errors. Why is a more complex discussion, but the bottom line is that an AI that did that would be too easy to cheese without sophisticated mechanisms the AI simply doesn't have. It can't remember, it can't predict, and retrofitting such awareness into the state engine would be non-trivial.
The Deep Dive is meant to be a way to communicate the issues to a non-technical audience. But if you're going to use the Deep Dive to make critiques about how the AI is being reworked, you would need far more extensive experience dealing with such systems, enough to infer what MCOC looks like given the limited information the Deep Dive contains.
If you think you understand how the AI works, if you think it just sits there randomly rolling dice until it lands on an action that is possible and then does it, can you explain why the AI sometimes just stops acting altogether? If you think that's the AI just completely breaking, it isn't. The AI is still running, and somehow electing to do nothing. I actually asked the devs a clarifying question about this, because a) the answer should be obvious but b) the deep dive said something else that confused me about this, thus the need to clarify.
This is not a trivial thing to unravel.
As usual, this is a ridiculous post with very little substance. In order to understand what they are doing, I'd need to see the code which is obviously not happening. I'm talking as a player not as a coder. Have you seen the code? Or are you just guessing at what Kabam is doing? Eitherway, it's not worth my time to read this.
When my customers criticize my work, I figure out a way to fix it. It must be nice, for Kabam, that enablers such as yourself will rise up to defend them at all costs because obviously, the assertion that we shouldn't have to wait till AI 2.0 is an unreasonable one.
Why would you want to see the code? You don't have experience programming for gaming remember? So what would you be able to do?
If he's not satisfied with DNA's knowledge and experience in this case, nothing will do.
I’m glad that Kabam is proactively communicating more often but they are dealing with many players who have a decade’s worth of experience with the AI. Our thousands of hours of focused interaction with the product give us quite a lot of knowledge about what it does, what the AI’s norms are / were, etc. Kabam is going to have to do a lot more than provide a basic primer on how the AI works before we just accept it without question or comment.
As an explanation, it works fine. But if you expect Kabam to prove they need to do what they say they are currently doing, by the time they did that they would probably be done with AI 2.0.
The last time I posted about AI, Crashed said I’d hit the nail on the head. Interestingly none of you people have any critique on any content, just a reflexive derision to anything anti Kabam… Very interesting.
No critique because there's no point in providing any since the content of your post shows you dont know how programming works therefore anything anyone would say you would not understand and subsequently deny out of plain ignorance
If you believe that every change that has happened to the AI is unintended then you’re really bad at this game. Have you ever place celestial or play at the top of war? If you knew my work history you wouldn’t make assumptions that make you seem disingenuous.
I’ll simplify it for the people who aren’t being disingenuous. Programming is hard and we don’t expect perfection but to tell us to wait for AI 2.0 while there are major problems right now is absurd. And no not every single AI change has been inintended. That’s not how programming works unless you’re completely inept (which seems to be your defense of Kabam).
It doesn't actually sound like you understand how the AI works. The Deep Dive is a very simplified version, to give the average player a basic overview of what's going on. But you seem to be inferring things no technically competent person would assume in this situation. For example: "I've said on many occasions that the AI appears to be frame perfect (just frame) and this means that 60 times per second (used to be 30 times a second), the AI can read your input and react accordingly." You seem to be inferring that every single frame the AI picks a random action to perform in that specific frame. But that's not how the AI works. In fact, no game AI could possibly work that way.
What's missing from the Deep Dive, because it would mostly cloud the discussion with technical details almost no player would have any use for, is that all game AI systems are state engines. The AI chooses to transition from state to state, and within each state there are valid actions it can perform. Because of that, the choices it can make at any given moment in time change over time - you can even see a hint of that in the animated display in the Deep Dive. The AI is not deciding sixty times per second whether or not to intercept you. If that was the case, it would basically never fail to do so. It would simply have too many opportunities to select that action.
You're also getting hung up on "frame accurate." Most AI systems in video games are "frame accurate." They kind of have to be, and the more simple they are, the more likely this is the case, because as the Deep Dive states, the AI has no memory. As a result, it can only make decisions now based on the things it sees now. In other words, all of its actions involve what it sees in a single physics and animation frame, and by definition it must take action now because it will not remember what it was supposed to do one frame later. Because the AI has no memory, it has always been frame accurate going back to the beginning of the game. That just means it is not programmed to deliberately whiff or make other similar errors. Why is a more complex discussion, but the bottom line is that an AI that did that would be too easy to cheese without sophisticated mechanisms the AI simply doesn't have. It can't remember, it can't predict, and retrofitting such awareness into the state engine would be non-trivial.
The Deep Dive is meant to be a way to communicate the issues to a non-technical audience. But if you're going to use the Deep Dive to make critiques about how the AI is being reworked, you would need far more extensive experience dealing with such systems, enough to infer what MCOC looks like given the limited information the Deep Dive contains.
If you think you understand how the AI works, if you think it just sits there randomly rolling dice until it lands on an action that is possible and then does it, can you explain why the AI sometimes just stops acting altogether? If you think that's the AI just completely breaking, it isn't. The AI is still running, and somehow electing to do nothing. I actually asked the devs a clarifying question about this, because a) the answer should be obvious but b) the deep dive said something else that confused me about this, thus the need to clarify.
This is not a trivial thing to unravel.
As usual, this is a ridiculous post with very little substance. In order to understand what they are doing, I'd need to see the code which is obviously not happening. I'm talking as a player not as a coder. Have you seen the code? Or are you just guessing at what Kabam is doing? Eitherway, it's not worth my time to read this.
When my customers criticize my work, I figure out a way to fix it. It must be nice, for Kabam, that enablers such as yourself will rise up to defend them at all costs because obviously, the assertion that we shouldn't have to wait till AI 2.0 is an unreasonable one.
When my customers ask me to explain what I'm doing, I'm more than happy to do so. But if they ask me to prove what I'm doing is necessary, I tell them they can trust me, or they can do it themselves. I can explain why I do what I do. But to prove it is necessary would require the kind of training and knowledge transfer comparable to what I need to deliver to a new employee over a period of years.
Saying I'm defending Kabam when all I'm doing is explaining how this stuff works - how it basically works *everywhere* - is just further proof you don't understand how this stuff works, but want really badly to still assert you have a valid opinion about it. You think this is stuff only Kabam would know, and only the people in Kabam who have actually looked at the code. There are specifics that only someone who looked at the implementation would know, but there's lots of stuff that a) anyone who's worked on games with AI modules would reasonably know and b) anyone with sufficient systems design would be able to reasonably infer.
The very first course in most programming degrees is called "Data Structures and Algorithms." It certainly doesn't teach every single data structure and algorithm in existence, of course, but it is called that in large part because (software) systems design is fundamentally about how you store data and what algorithms you use to process that data. When you do enough of it, you learn what the cosmos of techniques is out there. I would not need to explain how lootboxes worked, or how animation sequencers worked, or how game AI works to someone with sufficient expertise in general systems design. They would be able to infer, given just a few pieces of information, because there are only a few standard ways to tackle certain problems that everyone uses (unless they are crazy or stupid).
Any sufficiently experienced systems designer would probably deduce that a) most game AI uses state engines, and b) oh my god they use state engines. Because state engines are simultaneously very common, specifically because they can encapsulate a lot of complex behavior in a relatively simple implementation, and extremely prone to subtle problems, specifically because they can encapsulate a lot of complex behavior in a relatively simple implementation. I've seen religious arguments far more civil than the fights that break out over the use and abuse of state engines.
This is all common knowledge among the people actually in the field.
The last time I posted about AI, Crashed said I’d hit the nail on the head. Interestingly none of you people have any critique on any content, just a reflexive derision to anything anti Kabam… Very interesting.
No critique because there's no point in providing any since the content of your post shows you dont know how programming works therefore anything anyone would say you would not understand and subsequently deny out of plain ignorance
If you believe that every change that has happened to the AI is unintended then you’re really bad at this game. Have you ever place celestial or play at the top of war? If you knew my work history you wouldn’t make assumptions that make you seem disingenuous.
I’ll simplify it for the people who aren’t being disingenuous. Programming is hard and we don’t expect perfection but to tell us to wait for AI 2.0 while there are major problems right now is absurd. And no not every single AI change has been inintended. That’s not how programming works unless you’re completely inept (which seems to be your defense of Kabam).
It doesn't actually sound like you understand how the AI works. The Deep Dive is a very simplified version, to give the average player a basic overview of what's going on. But you seem to be inferring things no technically competent person would assume in this situation. For example: "I've said on many occasions that the AI appears to be frame perfect (just frame) and this means that 60 times per second (used to be 30 times a second), the AI can read your input and react accordingly." You seem to be inferring that every single frame the AI picks a random action to perform in that specific frame. But that's not how the AI works. In fact, no game AI could possibly work that way.
What's missing from the Deep Dive, because it would mostly cloud the discussion with technical details almost no player would have any use for, is that all game AI systems are state engines. The AI chooses to transition from state to state, and within each state there are valid actions it can perform. Because of that, the choices it can make at any given moment in time change over time - you can even see a hint of that in the animated display in the Deep Dive. The AI is not deciding sixty times per second whether or not to intercept you. If that was the case, it would basically never fail to do so. It would simply have too many opportunities to select that action.
You're also getting hung up on "frame accurate." Most AI systems in video games are "frame accurate." They kind of have to be, and the more simple they are, the more likely this is the case, because as the Deep Dive states, the AI has no memory. As a result, it can only make decisions now based on the things it sees now. In other words, all of its actions involve what it sees in a single physics and animation frame, and by definition it must take action now because it will not remember what it was supposed to do one frame later. Because the AI has no memory, it has always been frame accurate going back to the beginning of the game. That just means it is not programmed to deliberately whiff or make other similar errors. Why is a more complex discussion, but the bottom line is that an AI that did that would be too easy to cheese without sophisticated mechanisms the AI simply doesn't have. It can't remember, it can't predict, and retrofitting such awareness into the state engine would be non-trivial.
The Deep Dive is meant to be a way to communicate the issues to a non-technical audience. But if you're going to use the Deep Dive to make critiques about how the AI is being reworked, you would need far more extensive experience dealing with such systems, enough to infer what MCOC looks like given the limited information the Deep Dive contains.
If you think you understand how the AI works, if you think it just sits there randomly rolling dice until it lands on an action that is possible and then does it, can you explain why the AI sometimes just stops acting altogether? If you think that's the AI just completely breaking, it isn't. The AI is still running, and somehow electing to do nothing. I actually asked the devs a clarifying question about this, because a) the answer should be obvious but b) the deep dive said something else that confused me about this, thus the need to clarify.
This is not a trivial thing to unravel.
As usual, this is a ridiculous post with very little substance. In order to understand what they are doing, I'd need to see the code which is obviously not happening. I'm talking as a player not as a coder. Have you seen the code? Or are you just guessing at what Kabam is doing? Eitherway, it's not worth my time to read this.
When my customers criticize my work, I figure out a way to fix it. It must be nice, for Kabam, that enablers such as yourself will rise up to defend them at all costs because obviously, the assertion that we shouldn't have to wait till AI 2.0 is an unreasonable one.
When my customers ask me to explain what I'm doing, I'm more than happy to do so. But if they ask me to prove what I'm doing is necessary, I tell them they can trust me, or they can do it themselves. I can explain why I do what I do. But to prove it is necessary would require the kind of training and knowledge transfer comparable to what I need to deliver to a new employee over a period of years.
Saying I'm defending Kabam when all I'm doing is explaining how this stuff works - how it basically works *everywhere* - is just further proof you don't understand how this stuff works, but want really badly to still assert you have a valid opinion about it. You think this is stuff only Kabam would know, and only the people in Kabam who have actually looked at the code. There are specifics that only someone who looked at the implementation would know, but there's lots of stuff that a) anyone who's worked on games with AI modules would reasonably know and b) anyone with sufficient systems design would be able to reasonably infer.
The very first course in most programming degrees is called "Data Structures and Algorithms." It certainly doesn't teach every single data structure and algorithm in existence, of course, but it is called that in large part because (software) systems design is fundamentally about how you store data and what algorithms you use to process that data. When you do enough of it, you learn what the cosmos of techniques is out there. I would not need to explain how lootboxes worked, or how animation sequencers worked, or how game AI works to someone with sufficient expertise in general systems design. They would be able to infer, given just a few pieces of information, because there are only a few standard ways to tackle certain problems that everyone uses (unless they are crazy or stupid).
Any sufficiently experienced systems designer would probably deduce that a) most game AI uses state engines, and b) oh my god they use state engines. Because state engines are simultaneously very common, specifically because they can encapsulate a lot of complex behavior in a relatively simple implementation, and extremely prone to subtle problems, specifically because they can encapsulate a lot of complex behavior in a relatively simple implementation. I've seen religious arguments far more civil than the fights that break out over the use and abuse of state engines.
This is all common knowledge among the people actually in the field.
Yeah not going to bother reading this either. You refuse to understand that I'm complaining as a customer so there's no need to pay you any mind. I'll go back to ignoring your posts. GL.
The last time I posted about AI, Crashed said I’d hit the nail on the head. Interestingly none of you people have any critique on any content, just a reflexive derision to anything anti Kabam… Very interesting.
No critique because there's no point in providing any since the content of your post shows you dont know how programming works therefore anything anyone would say you would not understand and subsequently deny out of plain ignorance
If you believe that every change that has happened to the AI is unintended then you’re really bad at this game. Have you ever place celestial or play at the top of war? If you knew my work history you wouldn’t make assumptions that make you seem disingenuous.
I’ll simplify it for the people who aren’t being disingenuous. Programming is hard and we don’t expect perfection but to tell us to wait for AI 2.0 while there are major problems right now is absurd. And no not every single AI change has been inintended. That’s not how programming works unless you’re completely inept (which seems to be your defense of Kabam).
It doesn't actually sound like you understand how the AI works. The Deep Dive is a very simplified version, to give the average player a basic overview of what's going on. But you seem to be inferring things no technically competent person would assume in this situation. For example: "I've said on many occasions that the AI appears to be frame perfect (just frame) and this means that 60 times per second (used to be 30 times a second), the AI can read your input and react accordingly." You seem to be inferring that every single frame the AI picks a random action to perform in that specific frame. But that's not how the AI works. In fact, no game AI could possibly work that way.
What's missing from the Deep Dive, because it would mostly cloud the discussion with technical details almost no player would have any use for, is that all game AI systems are state engines. The AI chooses to transition from state to state, and within each state there are valid actions it can perform. Because of that, the choices it can make at any given moment in time change over time - you can even see a hint of that in the animated display in the Deep Dive. The AI is not deciding sixty times per second whether or not to intercept you. If that was the case, it would basically never fail to do so. It would simply have too many opportunities to select that action.
You're also getting hung up on "frame accurate." Most AI systems in video games are "frame accurate." They kind of have to be, and the more simple they are, the more likely this is the case, because as the Deep Dive states, the AI has no memory. As a result, it can only make decisions now based on the things it sees now. In other words, all of its actions involve what it sees in a single physics and animation frame, and by definition it must take action now because it will not remember what it was supposed to do one frame later. Because the AI has no memory, it has always been frame accurate going back to the beginning of the game. That just means it is not programmed to deliberately whiff or make other similar errors. Why is a more complex discussion, but the bottom line is that an AI that did that would be too easy to cheese without sophisticated mechanisms the AI simply doesn't have. It can't remember, it can't predict, and retrofitting such awareness into the state engine would be non-trivial.
The Deep Dive is meant to be a way to communicate the issues to a non-technical audience. But if you're going to use the Deep Dive to make critiques about how the AI is being reworked, you would need far more extensive experience dealing with such systems, enough to infer what MCOC looks like given the limited information the Deep Dive contains.
If you think you understand how the AI works, if you think it just sits there randomly rolling dice until it lands on an action that is possible and then does it, can you explain why the AI sometimes just stops acting altogether? If you think that's the AI just completely breaking, it isn't. The AI is still running, and somehow electing to do nothing. I actually asked the devs a clarifying question about this, because a) the answer should be obvious but b) the deep dive said something else that confused me about this, thus the need to clarify.
This is not a trivial thing to unravel.
As usual, this is a ridiculous post with very little substance. In order to understand what they are doing, I'd need to see the code which is obviously not happening. I'm talking as a player not as a coder. Have you seen the code? Or are you just guessing at what Kabam is doing? Eitherway, it's not worth my time to read this.
When my customers criticize my work, I figure out a way to fix it. It must be nice, for Kabam, that enablers such as yourself will rise up to defend them at all costs because obviously, the assertion that we shouldn't have to wait till AI 2.0 is an unreasonable one.
When my customers ask me to explain what I'm doing, I'm more than happy to do so. But if they ask me to prove what I'm doing is necessary, I tell them they can trust me, or they can do it themselves. I can explain why I do what I do. But to prove it is necessary would require the kind of training and knowledge transfer comparable to what I need to deliver to a new employee over a period of years.
Saying I'm defending Kabam when all I'm doing is explaining how this stuff works - how it basically works *everywhere* - is just further proof you don't understand how this stuff works, but want really badly to still assert you have a valid opinion about it. You think this is stuff only Kabam would know, and only the people in Kabam who have actually looked at the code. There are specifics that only someone who looked at the implementation would know, but there's lots of stuff that a) anyone who's worked on games with AI modules would reasonably know and b) anyone with sufficient systems design would be able to reasonably infer.
The very first course in most programming degrees is called "Data Structures and Algorithms." It certainly doesn't teach every single data structure and algorithm in existence, of course, but it is called that in large part because (software) systems design is fundamentally about how you store data and what algorithms you use to process that data. When you do enough of it, you learn what the cosmos of techniques is out there. I would not need to explain how lootboxes worked, or how animation sequencers worked, or how game AI works to someone with sufficient expertise in general systems design. They would be able to infer, given just a few pieces of information, because there are only a few standard ways to tackle certain problems that everyone uses (unless they are crazy or stupid).
Any sufficiently experienced systems designer would probably deduce that a) most game AI uses state engines, and b) oh my god they use state engines. Because state engines are simultaneously very common, specifically because they can encapsulate a lot of complex behavior in a relatively simple implementation, and extremely prone to subtle problems, specifically because they can encapsulate a lot of complex behavior in a relatively simple implementation. I've seen religious arguments far more civil than the fights that break out over the use and abuse of state engines.
This is all common knowledge among the people actually in the field.
Yeah not going to bother reading this either. You refuse to understand that I'm complaining as a customer so there's no need to pay you any mind. I'll go back to ignoring your posts. GL.
I understand you're complaining as a customer. An ignorant customer who believes their opinion of the best way to solve problems should be in any way acknowledged by the company. Just because you're a customer, does not mean you have a take me seriously pass.
Feel free to ignore me. My replies are really more for neutral readers than you. I know I'm not going to convince you of anything. The irony is that this is how your own posts are being treated by anyone in a position to either do anything about anything, or even say anything interesting about anything.
The last time I posted about AI, Crashed said I’d hit the nail on the head. Interestingly none of you people have any critique on any content, just a reflexive derision to anything anti Kabam… Very interesting.
No critique because there's no point in providing any since the content of your post shows you dont know how programming works therefore anything anyone would say you would not understand and subsequently deny out of plain ignorance
If you believe that every change that has happened to the AI is unintended then you’re really bad at this game. Have you ever place celestial or play at the top of war? If you knew my work history you wouldn’t make assumptions that make you seem disingenuous.
I’ll simplify it for the people who aren’t being disingenuous. Programming is hard and we don’t expect perfection but to tell us to wait for AI 2.0 while there are major problems right now is absurd. And no not every single AI change has been inintended. That’s not how programming works unless you’re completely inept (which seems to be your defense of Kabam).
It doesn't actually sound like you understand how the AI works. The Deep Dive is a very simplified version, to give the average player a basic overview of what's going on. But you seem to be inferring things no technically competent person would assume in this situation. For example: "I've said on many occasions that the AI appears to be frame perfect (just frame) and this means that 60 times per second (used to be 30 times a second), the AI can read your input and react accordingly." You seem to be inferring that every single frame the AI picks a random action to perform in that specific frame. But that's not how the AI works. In fact, no game AI could possibly work that way.
What's missing from the Deep Dive, because it would mostly cloud the discussion with technical details almost no player would have any use for, is that all game AI systems are state engines. The AI chooses to transition from state to state, and within each state there are valid actions it can perform. Because of that, the choices it can make at any given moment in time change over time - you can even see a hint of that in the animated display in the Deep Dive. The AI is not deciding sixty times per second whether or not to intercept you. If that was the case, it would basically never fail to do so. It would simply have too many opportunities to select that action.
You're also getting hung up on "frame accurate." Most AI systems in video games are "frame accurate." They kind of have to be, and the more simple they are, the more likely this is the case, because as the Deep Dive states, the AI has no memory. As a result, it can only make decisions now based on the things it sees now. In other words, all of its actions involve what it sees in a single physics and animation frame, and by definition it must take action now because it will not remember what it was supposed to do one frame later. Because the AI has no memory, it has always been frame accurate going back to the beginning of the game. That just means it is not programmed to deliberately whiff or make other similar errors. Why is a more complex discussion, but the bottom line is that an AI that did that would be too easy to cheese without sophisticated mechanisms the AI simply doesn't have. It can't remember, it can't predict, and retrofitting such awareness into the state engine would be non-trivial.
The Deep Dive is meant to be a way to communicate the issues to a non-technical audience. But if you're going to use the Deep Dive to make critiques about how the AI is being reworked, you would need far more extensive experience dealing with such systems, enough to infer what MCOC looks like given the limited information the Deep Dive contains.
If you think you understand how the AI works, if you think it just sits there randomly rolling dice until it lands on an action that is possible and then does it, can you explain why the AI sometimes just stops acting altogether? If you think that's the AI just completely breaking, it isn't. The AI is still running, and somehow electing to do nothing. I actually asked the devs a clarifying question about this, because a) the answer should be obvious but b) the deep dive said something else that confused me about this, thus the need to clarify.
This is not a trivial thing to unravel.
As usual, this is a ridiculous post with very little substance. In order to understand what they are doing, I'd need to see the code which is obviously not happening. I'm talking as a player not as a coder. Have you seen the code? Or are you just guessing at what Kabam is doing? Eitherway, it's not worth my time to read this.
When my customers criticize my work, I figure out a way to fix it. It must be nice, for Kabam, that enablers such as yourself will rise up to defend them at all costs because obviously, the assertion that we shouldn't have to wait till AI 2.0 is an unreasonable one.
Why would you want to see the code? You don't have experience programming for gaming remember? So what would you be able to do?
If he's not satisfied with DNA's knowledge and experience in this case, nothing will do.
I’m glad that Kabam is proactively communicating more often but they are dealing with many players who have a decade’s worth of experience with the AI. Our thousands of hours of focused interaction with the product give us quite a lot of knowledge about what it does, what the AI’s norms are / were, etc. Kabam is going to have to do a lot more than provide a basic primer on how the AI works before we just accept it without question or comment.
This is what's so surprising to me. Even the smallest changes in the AI are very noticeable.
I haven't seen a situation where unhappy paying customers are shushed by other paying customers...
The last time I posted about AI, Crashed said I’d hit the nail on the head. Interestingly none of you people have any critique on any content, just a reflexive derision to anything anti Kabam… Very interesting.
No critique because there's no point in providing any since the content of your post shows you dont know how programming works therefore anything anyone would say you would not understand and subsequently deny out of plain ignorance
If you believe that every change that has happened to the AI is unintended then you’re really bad at this game. Have you ever place celestial or play at the top of war? If you knew my work history you wouldn’t make assumptions that make you seem disingenuous.
I’ll simplify it for the people who aren’t being disingenuous. Programming is hard and we don’t expect perfection but to tell us to wait for AI 2.0 while there are major problems right now is absurd. And no not every single AI change has been inintended. That’s not how programming works unless you’re completely inept (which seems to be your defense of Kabam).
It doesn't actually sound like you understand how the AI works. The Deep Dive is a very simplified version, to give the average player a basic overview of what's going on. But you seem to be inferring things no technically competent person would assume in this situation. For example: "I've said on many occasions that the AI appears to be frame perfect (just frame) and this means that 60 times per second (used to be 30 times a second), the AI can read your input and react accordingly." You seem to be inferring that every single frame the AI picks a random action to perform in that specific frame. But that's not how the AI works. In fact, no game AI could possibly work that way.
What's missing from the Deep Dive, because it would mostly cloud the discussion with technical details almost no player would have any use for, is that all game AI systems are state engines. The AI chooses to transition from state to state, and within each state there are valid actions it can perform. Because of that, the choices it can make at any given moment in time change over time - you can even see a hint of that in the animated display in the Deep Dive. The AI is not deciding sixty times per second whether or not to intercept you. If that was the case, it would basically never fail to do so. It would simply have too many opportunities to select that action.
You're also getting hung up on "frame accurate." Most AI systems in video games are "frame accurate." They kind of have to be, and the more simple they are, the more likely this is the case, because as the Deep Dive states, the AI has no memory. As a result, it can only make decisions now based on the things it sees now. In other words, all of its actions involve what it sees in a single physics and animation frame, and by definition it must take action now because it will not remember what it was supposed to do one frame later. Because the AI has no memory, it has always been frame accurate going back to the beginning of the game. That just means it is not programmed to deliberately whiff or make other similar errors. Why is a more complex discussion, but the bottom line is that an AI that did that would be too easy to cheese without sophisticated mechanisms the AI simply doesn't have. It can't remember, it can't predict, and retrofitting such awareness into the state engine would be non-trivial.
The Deep Dive is meant to be a way to communicate the issues to a non-technical audience. But if you're going to use the Deep Dive to make critiques about how the AI is being reworked, you would need far more extensive experience dealing with such systems, enough to infer what MCOC looks like given the limited information the Deep Dive contains.
If you think you understand how the AI works, if you think it just sits there randomly rolling dice until it lands on an action that is possible and then does it, can you explain why the AI sometimes just stops acting altogether? If you think that's the AI just completely breaking, it isn't. The AI is still running, and somehow electing to do nothing. I actually asked the devs a clarifying question about this, because a) the answer should be obvious but b) the deep dive said something else that confused me about this, thus the need to clarify.
This is not a trivial thing to unravel.
As usual, this is a ridiculous post with very little substance. In order to understand what they are doing, I'd need to see the code which is obviously not happening. I'm talking as a player not as a coder. Have you seen the code? Or are you just guessing at what Kabam is doing? Eitherway, it's not worth my time to read this.
When my customers criticize my work, I figure out a way to fix it. It must be nice, for Kabam, that enablers such as yourself will rise up to defend them at all costs because obviously, the assertion that we shouldn't have to wait till AI 2.0 is an unreasonable one.
When my customers ask me to explain what I'm doing, I'm more than happy to do so. But if they ask me to prove what I'm doing is necessary, I tell them they can trust me, or they can do it themselves. I can explain why I do what I do. But to prove it is necessary would require the kind of training and knowledge transfer comparable to what I need to deliver to a new employee over a period of years.
Saying I'm defending Kabam when all I'm doing is explaining how this stuff works - how it basically works *everywhere* - is just further proof you don't understand how this stuff works, but want really badly to still assert you have a valid opinion about it. You think this is stuff only Kabam would know, and only the people in Kabam who have actually looked at the code. There are specifics that only someone who looked at the implementation would know, but there's lots of stuff that a) anyone who's worked on games with AI modules would reasonably know and b) anyone with sufficient systems design would be able to reasonably infer.
The very first course in most programming degrees is called "Data Structures and Algorithms." It certainly doesn't teach every single data structure and algorithm in existence, of course, but it is called that in large part because (software) systems design is fundamentally about how you store data and what algorithms you use to process that data. When you do enough of it, you learn what the cosmos of techniques is out there. I would not need to explain how lootboxes worked, or how animation sequencers worked, or how game AI works to someone with sufficient expertise in general systems design. They would be able to infer, given just a few pieces of information, because there are only a few standard ways to tackle certain problems that everyone uses (unless they are crazy or stupid).
Any sufficiently experienced systems designer would probably deduce that a) most game AI uses state engines, and b) oh my god they use state engines. Because state engines are simultaneously very common, specifically because they can encapsulate a lot of complex behavior in a relatively simple implementation, and extremely prone to subtle problems, specifically because they can encapsulate a lot of complex behavior in a relatively simple implementation. I've seen religious arguments far more civil than the fights that break out over the use and abuse of state engines.
This is all common knowledge among the people actually in the field.
Yeah not going to bother reading this either. You refuse to understand that I'm complaining as a customer so there's no need to pay you any mind. I'll go back to ignoring your posts. GL.
I understand you're complaining as a customer. An ignorant customer who believes their opinion of the best way to solve problems should be in any way acknowledged by the company. Just because you're a customer, does not mean you have a take me seriously pass.
Feel free to ignore me. My replies are really more for neutral readers than you. I know I'm not going to convince you of anything. The irony is that this is how your own posts are being treated by anyone in a position to either do anything about anything, or even say anything interesting about anything.
I ignore you because you're condescending and write a lot without saying anything concrete. So there's no point to read your posts or respond; I don't view you as a neutral party or even pro-player based on prior interactions. Your entire take is an appeal to authority. You seem angry for some reason. A reason I cannot comprehend.
I put average desi in the same bucket as you all the time but in this thread, they've been quite normal and thus we had a normal interaction even though they don't agree with me. I can accept that they feel the AI is fine no problem.
Nothing I've said is anything different from what the top players say on streams. Like I said, even employees are saying the same things in frustration (a point you refuse to acknowledge). So I'm not in the wrong here (from a player perspective). I'm probably wrong about the technicalities but that's to be expected as it's an attempt to get a point across.
Ultimately, if the AI was in a good place, there wouldn't be a need for AI 2.0 and I'll say it again, if you cannot see that the changes appear to be deliberate in some case, then you're not a good enough player.
I don't view you as a neutral party or even pro-player based on prior interactions.
Calling the person who pushed for 2 tokens for winning in Battlegrounds,starting season in a high brackets instead of from the Zero , roster based matchups till D2 as well as free revives for AW as "not pro player" is a terrible take. You yourself said you play high war and BG, so you could appreciate what DNA has pushed for. Although I think DNA might to refer to themself as Pro Game(?)
Your entire take is an appeal to authority. You seem angry for some reason.
DNA is the authority. Also Appeal to authority does not apply here. Appeal to Authority fallacy is used when you substitute an expertise in one field as being relevant crebility for an opinion in another.
I put average desi in the same bucket as you
I don't know I should be happy or DNA should be insulted.
Ultimately, if the AI was in a good place, there wouldn't be a need for AI 2.0 and I'll say it again, if you cannot see that the changes appear to be deliberate in some case, then you're not a good enough player.
Changes happening does not mean they are deliberate. As much as it may seem to players who think Kabam is out to get them
I don't view you as a neutral party or even pro-player based on prior interactions.
Calling the person who pushed for 2 tokens for winning in Battlegrounds,starting season in a high brackets instead of from the Zero , roster based matchups till D2 as well as free revives for AW as "not pro player" is a terrible take. You yourself said you play high war and BG, so you could appreciate what DNA has pushed for. Although I think DNA might to refer to themself as Pro Game(?)
Your entire take is an appeal to authority. You seem angry for some reason.
DNA is the authority. Also Appeal to authority does not apply here. Appeal to Authority fallacy is used when you substitute an expertise in one field as being relevant crebility for an opinion in another.
I put average desi in the same bucket as you
I don't know I should be happy or DNA should be insulted.
Ultimately, if the AI was in a good place, there wouldn't be a need for AI 2.0 and I'll say it again, if you cannot see that the changes appear to be deliberate in some case, then you're not a good enough player.
Changes happening does not mean they are deliberate. As much as it may seem to players who think Kabam is out to get them
I guess you're back to the norm. It was good while it lasted. Maybe someday you'll join us in reality.
I do not think that every change is deliberate. I wonder if you can admit that some changes are likely deliberate...
Comments
What's missing from the Deep Dive, because it would mostly cloud the discussion with technical details almost no player would have any use for, is that all game AI systems are state engines. The AI chooses to transition from state to state, and within each state there are valid actions it can perform. Because of that, the choices it can make at any given moment in time change over time - you can even see a hint of that in the animated display in the Deep Dive. The AI is not deciding sixty times per second whether or not to intercept you. If that was the case, it would basically never fail to do so. It would simply have too many opportunities to select that action.
You're also getting hung up on "frame accurate." Most AI systems in video games are "frame accurate." They kind of have to be, and the more simple they are, the more likely this is the case, because as the Deep Dive states, the AI has no memory. As a result, it can only make decisions now based on the things it sees now. In other words, all of its actions involve what it sees in a single physics and animation frame, and by definition it must take action now because it will not remember what it was supposed to do one frame later. Because the AI has no memory, it has always been frame accurate going back to the beginning of the game. That just means it is not programmed to deliberately whiff or make other similar errors. Why is a more complex discussion, but the bottom line is that an AI that did that would be too easy to cheese without sophisticated mechanisms the AI simply doesn't have. It can't remember, it can't predict, and retrofitting such awareness into the state engine would be non-trivial.
The Deep Dive is meant to be a way to communicate the issues to a non-technical audience. But if you're going to use the Deep Dive to make critiques about how the AI is being reworked, you would need far more extensive experience dealing with such systems, enough to infer what MCOC looks like given the limited information the Deep Dive contains.
If you think you understand how the AI works, if you think it just sits there randomly rolling dice until it lands on an action that is possible and then does it, can you explain why the AI sometimes just stops acting altogether? If you think that's the AI just completely breaking, it isn't. The AI is still running, and somehow electing to do nothing. I actually asked the devs a clarifying question about this, because a) the answer should be obvious but b) the deep dive said something else that confused me about this, thus the need to clarify.
This is not a trivial thing to unravel.
When my customers criticize my work, I figure out a way to fix it. It must be nice, for Kabam, that enablers such as yourself will rise up to defend them at all costs because obviously, the assertion that we shouldn't have to wait till AI 2.0 is an unreasonable one.
While I agree that who you are responding to is a bit too flippant, I disagree with your gatekeeping stance. You are currently arguing that this is a high level BG problem that only high level BG players know what they are talking about, even though OP’s post makes no mention about BGs.
And then if we were to accept the “high level BG” talking point, why should there be a major change for everyone (trying to change the AI) based on the experiences of a niche few high level players of a niche game mode?
If there are two options:
(1) Maintain current AI, warts and all, while waiting for AI 2.0, that inconveniences mainly just high level BG players.*
(2) Try to fiddle around with current AI to potentially benefit just high level BG players who noticing a problem or break it make the game worse for everyone
As long as Kabam continues to make money and BGs continues to be a niche game mode, I can totally understand which option Kabam pursues.
*Side note: If this is a high level BG issue, every high level player is dealing with the same AI, so it should all end up a wash. You might get light intercepted one time, they might experience it another time.
That doesn't mean that it isn't a problem for non-top tier players, it's a problem for everyone and we should advocate for better as customers (many of us paying).
I’m glad that Kabam is proactively communicating more often but they are dealing with many players who have a decade’s worth of experience with the AI. Our thousands of hours of focused interaction with the product give us quite a lot of knowledge about what it does, what the AI’s norms are / were, etc. Kabam is going to have to do a lot more than provide a basic primer on how the AI works before we just accept it without question or comment.
It is thus theoretically possible for the AI to enter a state where there is no valid option to exit that state. Most likely, in my experience this happens when an action sequencer fouls with an animation sequencer.
Or to put it more colloquially, the AI has an implicit understanding that certain actions happen in certain orders, but the animation engine of the game (or the mechanics of the game in general) can sometimes fail to obey that in certain weird corner cases. If the game finds itself expecting something to happen that will never happen, it waits forever.
I should point out this sounds like a contradiction, as I mentioned earlier I don’t believe the AI makes predictions. “Expects” is a colloquialism here, it means the AI enters a state where the proscribed possible actions the random weighted table offers have been designed by some human to be what the human believed would be valid options at that moment, because the human then expected that after the AI took one of those actions it would then be put into a different state that opened the door to other options that were logical continuations of the fight. But if the AI cannot do any of those things, it cannot continue the fight and it cannot transition to a state where other options become available. It is simply continuing to do what it was programmed to do, which is to do one of the things it is told are valid options, which in this case is nothing, and continuing to transition to new states determined by those actions, which is also nothing. I’m anthropomorphizing in this as “expecting something and waiting for it to happen.” It doesn’t expect and it doesn’t wait. It just keeps rolling dice that will never come up sevens.
That’s also why we can “wake it up.” If the AI was broken, it would stay broken no matter what we did. But when we act, *we* transition the state engine to a state where the locally valid actions now allow the AI to both act and continue to transition to other new states.
As long as I’m explaining this, I might as well explain the other question I asked (since it echoes things the devs have already said in the past publicly, I don’t think it gives away anything of a sensitive nature). I long assumed there was a set of tuning parameters for the AI, something that would explain how the devs could say there were very few AI profiles but in my experience there were far more AI “tendencies.” I suspected there was some initialization parameters - something where at the start of the fight the game sort of rolled dice to give the fight some “character” and those random initial settings could strongly influence the “personality” of the AI, even if the AI itself was identical from fight to fight. According to the devs, there is no such thing happening. However, they did confirm another one of my suspicions, something I wrote about several years ago when AI first started getting discussed a lot on the forums. Because the AI is a state engine (so I assumed back then), its moment by moment behavior is implicitly influenced by the past., The AI doesn’t have a memory per se, but the state engine does change the behavior of the AI with each action we do, and then it does (randomly), and so on. So just like a Chess game that can be strongly influenced by the opening moves of the game, the “personality” of the AI can, in some cases, be strongly influenced by the very first thing it does, which triggers the players to react in a very specific way to that, and so on. Moment by moment, the AI is very simple. But because it is also a state engine, it can become caught in feedback loops. For example passive AI is often a positive feedback loop, because the AI tends to take actions in response to the player. When it gets passive, the players don’t have openings to attack, so the players get passive. This can then get reinforced to the point where it gets difficult to break out of.
If the state engines were implemented in a generic engine using data, the state engines could be tinkered with relatively easily, at least to a degree. But I suspect it isn’t: I believe the AI state engines is implemented directly in code. And that’s why they have to write a completely new one from scratch: tampering with any part of the behavior involves directly tampering with the AI code. This is a non-trivial thing to do, especially because it is unlikely to be some module you can rip out of Unity. It is more likely to be intertwined with the rest of the engine, because if MCOC is a cleanly modularized Unity project, it will be the first one I’ve ever heard of. Unity doesn’t encourage that sort of thing. If you’re going to do that sort of thing anyway, you might as well write an AI engine that does what you want from scratch.
Saying I'm defending Kabam when all I'm doing is explaining how this stuff works - how it basically works *everywhere* - is just further proof you don't understand how this stuff works, but want really badly to still assert you have a valid opinion about it. You think this is stuff only Kabam would know, and only the people in Kabam who have actually looked at the code. There are specifics that only someone who looked at the implementation would know, but there's lots of stuff that a) anyone who's worked on games with AI modules would reasonably know and b) anyone with sufficient systems design would be able to reasonably infer.
The very first course in most programming degrees is called "Data Structures and Algorithms." It certainly doesn't teach every single data structure and algorithm in existence, of course, but it is called that in large part because (software) systems design is fundamentally about how you store data and what algorithms you use to process that data. When you do enough of it, you learn what the cosmos of techniques is out there. I would not need to explain how lootboxes worked, or how animation sequencers worked, or how game AI works to someone with sufficient expertise in general systems design. They would be able to infer, given just a few pieces of information, because there are only a few standard ways to tackle certain problems that everyone uses (unless they are crazy or stupid).
Any sufficiently experienced systems designer would probably deduce that a) most game AI uses state engines, and b) oh my god they use state engines. Because state engines are simultaneously very common, specifically because they can encapsulate a lot of complex behavior in a relatively simple implementation, and extremely prone to subtle problems, specifically because they can encapsulate a lot of complex behavior in a relatively simple implementation. I've seen religious arguments far more civil than the fights that break out over the use and abuse of state engines.
This is all common knowledge among the people actually in the field.
Feel free to ignore me. My replies are really more for neutral readers than you. I know I'm not going to convince you of anything. The irony is that this is how your own posts are being treated by anyone in a position to either do anything about anything, or even say anything interesting about anything.
I haven't seen a situation where unhappy paying customers are shushed by other paying customers...
Your entire take is an appeal to authority. You seem angry for some reason. A reason I cannot comprehend.
I put average desi in the same bucket as you all the time but in this thread, they've been quite normal and thus we had a normal interaction even though they don't agree with me. I can accept that they feel the AI is fine no problem.
Nothing I've said is anything different from what the top players say on streams. Like I said, even employees are saying the same things in frustration (a point you refuse to acknowledge). So I'm not in the wrong here (from a player perspective). I'm probably wrong about the technicalities but that's to be expected as it's an attempt to get a point across.
Ultimately, if the AI was in a good place, there wouldn't be a need for AI 2.0 and I'll say it again, if you cannot see that the changes appear to be deliberate in some case, then you're not a good enough player.
I guess you're back to the norm. It was good while it lasted. Maybe someday you'll join us in reality.
I do not think that every change is deliberate. I wonder if you can admit that some changes are likely deliberate...
If the Mods's want later on, they can re-open this, but there isn’t anything new being discussed.