The AI in Carmageddon - Better than you think!

Breakin'Benny · Post by **Breakin'Benny** » Wed Jan 07, 2015 9:19 pm

Pretty sure that the part with the opponent going straight to kill itself just to complete the race for you is a coincidence, the CPU drivers are mentally deranged and wouldn't mind going Berserk-o even if something was obviously in their path. How about picking up Carmageddons 1, Splat Pack and even TDR 2000 while yer at it to find something to compare with? Right-out saying C2 is your favorite game and because you aren't a big gamer is no excuse, piracy is not too concerning for me.

Simply put, AI stupidity.

Cybernettr · Post by **Cybernettr** » Wed Jan 07, 2015 9:20 pm

CzechDeath wrote:Everywhere I look, listening to majority of educated people in this topic (AI, robotics, neurologists, biologists) they are repeting over and over that mind is generated by brain and so when we make computer as complex as human brain, we will get consciousness. Yet nobody knows exactly how this is done in that brain. Exact process however is what we as humanity need to understand to even try to make AI. So far we failed to pinpoint which neurons are responsible for what. We know what is the hardware made of, what part relates to what function of body, but how is all this conscious, that is debate that humans are having since they realised the are conscious. No answare yet...

Well, not exactly. Scanning the human brain and figuring out how it works (the top down approach) is one way of achieving artificial intelligence, while the bottom-up approach, or figuring out specific elements of intelligence and how to replicate them in the computer from scratch (the "bottom-up" approach) is another way to tackle the problem. This is why we have so many examples of "narrow AI," such as optical character recognition, face recognition, language translation, self driving cars, routing of trains and airplanes, stock trading, and a thousand other ways in which AI is used in the world today. As for consciousness, who can exactly say what consciousness is? This is something I don't get into in my videos. And yes, the AI in Carmageddon would be an example of "narrow AI." It's an example of an AI's mastery of a virtual world. Look up "SHRDLU."

Yfrid · Post by **Yfrid** » Wed Jan 07, 2015 9:36 pm

Cybernettr wrote:This is why we have so many examples of "narrow AI," such as optical character recognition, face recognition, language translation, self driving cars, routing of trains and airplanes, stock trading, and a thousand other ways in which AI is used in the world today. As for consciousness, who can exactly say what consciousness is

None of these has anything to do with AI... and i told you in my previous post why consciousness can't exist in nowdays (let alone past) technologies.

Electric Blue · Post by **Electric Blue** » Thu Jan 08, 2015 8:04 am

All aboard, gentlemen! :)

CzechDeath · Post by **CzechDeath** » Thu Jan 08, 2015 10:26 am

Cybernettr wrote:
CzechDeath wrote:Everywhere I look, listening to majority of educated people in this topic (AI, robotics, neurologists, biologists) they are repeting over and over that mind is generated by brain and so when we make computer as complex as human brain, we will get consciousness. Yet nobody knows exactly how this is done in that brain. Exact process however is what we as humanity need to understand to even try to make AI. So far we failed to pinpoint which neurons are responsible for what. We know what is the hardware made of, what part relates to what function of body, but how is all this conscious, that is debate that humans are having since they realised the are conscious. No answare yet...

Well, not exactly. Scanning the human brain and figuring out how it works (the top down approach) is one way of achieving artificial intelligence, while the bottom-up approach, or figuring out specific elements of intelligence and how to replicate them in the computer from scratch (the "bottom-up" approach) is another way to tackle the problem. This is why we have so many examples of "narrow AI," such as optical character recognition, face recognition, language translation, self driving cars, routing of trains and airplanes, stock trading, and a thousand other ways in which AI is used in the world today. As for consciousness, who can exactly say what consciousness is? This is something I don't get into in my videos. And yes, the AI in Carmageddon would be an example of "narrow AI." It's an example of an AI's mastery of a virtual world. Look up "SHRDLU."

So you are talking about AI as AI with it's hardware boundries? And not anything outworldish? Because the game is hardly thinking... it executes predetirmened actions based on what player does. It does not see in to future nor it is reading player's mind =D becasuse it you suggest it does, than we could go in to the whole supernatural and psychadelic realm... I'm not saying things like that are impossible, but I see it as highly unlikely for less complicated explanations offering sutisfying answares...

also take a look on this https://www.youtube.com/watch?v=3hvEiNpZrGU this is pretty grounded view on AI and robotics...(has 2 parts)

Cybernettr · Post by **Cybernettr** » Thu Jan 08, 2015 3:46 pm

So you are talking about AI as AI with it's hardware boundries? And not anything outworldish? Because the game is hardly thinking... it executes predetirmened actions based on what player does. It does not see in to future nor it is reading player's mind =D becasuse it you suggest it does, than we could go in to the whole supernatural and psychadelic realm... I'm not saying things like that are impossible, but I see it as highly unlikely for less complicated explanations offering sutisfying answares...

Nothing supernatural about it. Technology always seems "magical" until another piece of it is figured out. Then it just becomes "routine computation." (Maybe this is why magicians never tell the secret. Because once the secret is known, it doesn't seem like "magic" anymore.)

also take a look on this https://www.youtube.com/watch?v=3hvEiNpZrGU this is pretty grounded view on AI and robotics...(has 2 parts)

Yeah, I've seen that already. Consider mine as sort of an "alternative viewpoint."

I predict that, even if we someday get an AI like in the movie "Her," many people will still insist that it's not "real" intelligence. No matter what A I achieves, it will be called either "imaginary" or "just routine computation." Call it a sort of "neo-Luddism."

Dodo Bizar · Post by **Dodo Bizar** » Thu Jan 08, 2015 4:58 pm

Cybernettr wrote:Okay, then let me ask you this: what would it take to convince you that there is true AI in the game, (within the bounds of the capabilities of the game engine, of course)? Obviously, one of the avatars is not going to climb out of their car and start having a conversation with you, because that is not within the capabilities of the game engine. Maybe we just have a different definition of what constitutes "intelligence," or proof of same. Remember, I am not saying that the AI is as intelligent as a real person, just that it shows SOME intelligence.

I really like this question. See, based on my extensive gaming experience (like it's a job hehe) where you basically get to know the simple set of rules for much gaming AI (where I do not mean consiousness with AI, but just the set of rules it works with) and my work as an engineer where I do actually have to work with or at least study somebody elses neural networks sometime... well all that just gives me a foundation/background which helps me to know how back in 97/98 they programmed C1/C2 most likely. And to me I am very confidend it's the drive-path definitions with some simple rules for aggresive/racing/fleeing behaviour etc. as explained before. Hence your videos and interpretations are to me merely an overestimation of the amount of knowledge the programmers put in their game. When you say some opponent is doing a specific act for a specific reason, I'd say some opponent was just doing something based on a status, perhaps with some randomisation to get varying behaviour. Ok, so based on the different interpretation here it will be very hard if not impossible for you to convince me.

However, would I try to answer your question appropiately I must confess I get stuck. So regardless of the memory monitoring I read here which would not even be foolproof, when you ask me what I would like to see happening in the game to convince me of strong AI algorythms (is that the way I best can put the question?)... I must confess I would not know. The easy way out is probably when I see something that I believe I let you know...

Yfrid · Post by **Yfrid** » Thu Jan 08, 2015 5:41 pm

Cybernettr wrote:Technology always seems "magical" until another piece of it is figured out.

not sure what do you mean? people doesn't figure out technology, people makes it. From scratch.

Then, like in consumer scenario, other people that has no knowledge about how some piece of electronic works, is allowed to buy it as a product for its features and may oversimplify its functionality as "magical cool tech thing" but there's no mystery behind that device. Just lack of knowledge from the end user (most of the times doesn't need to, after all.)

Or did you meant something else?

Cybernettr wrote:I predict that, even if we someday get an AI like in the movie "Her," many people will still insist that it's not "real" intelligence. No matter what A I achieves, it will be called either "imaginary" or "just routine computation." Call it a sort of "neo-Luddism."

if proper AI will be ever done, i'm confident the hardware where will be stored won't have anything to share with ordinary computer technology. That's why i don't buy theories about sentient IA hidden in games or other devices that shows some alleged sparks of free will. Like mentally challenged by low end hardware or something.

AI made from ordinary calculators IS and will ALWAYS be an illusion made of complex routines, because that's as far such technology can go, because of the (well known) way it works.

Maybe one day we will be able to replicate a complex neural network that is self aware and intelligent, but it won't be designed anything like a common pc or any of its derivates.

I have hard times imagining the thing will have an intel cpu, kingston ram and an SSD drive...

skyNET · Post by **skyNET** » Thu Jan 08, 2015 6:28 pm

Cybernettr wrote:
In any event, I don't think that even if you did find the database files on disk of the players past actions that this would prove that the game is, in fact, intelligent.
In some of my videos, the AI clearly had to have at least some kind of short-term memory and an awareness of the current game conditions. (...) It obviously had to have at least a short-term "memory" of where the opponent was and other aspects of the game's conditions.

You don't get it do you?

1. You say that the A.I. displays these traits: intention, foresight, innovativeness, "bending the rules", resourcefulness.

2. At least two of these traits require "engrams" of past experiences.

3. You say that the A.I. must have at least short-term memories.

4. You say that locating these "engrams" would not prove that the A.I. is "conscious".

Can't you see the contradiction you got yourself into?

A.I. is intelligent and has at least short-term memory -> finding it's engrams won't prove that it's intelligent

(but your videos on their own do)

What the fuck man?

You are downplaying a possible hard evidence, because right now it fits your agenda (honestly I think you are afraid that you won't find anything). That's not wise nor scientific. It's just disgusting, and for me it bodes that you're talking shit and you know it. You won't make an effort or give a thought, all the while you wasted countless hours on petty videos which have zero weight added towards your claims.

It's really hard for me at this point not to offend you with some smack-talk.

Addendum

I made a video of my boyo while he combines his toys and the floor, while I say to him "what's the sum total of 3+3?" then he smashes his toy six times into the ground. My 10 months old can already count like a pro, I got proof.

Post scriptum

You are subtly moving the goalpost always further, and are trying to lay the burden of proof on another.

- edited some mis-spelling I found -

Errol · Post by **Errol** » Thu Jan 08, 2015 7:09 pm

Cybernettr wrote:It obviously had to have at least a short-term "memory" of where the opponent was and other aspects of the game's conditions.

See finite state machines.

I can also say with 100% certainty that the game does not construct a database of previous play sessions to inform future AI decision making.

TwIsTeDMiNd · Post by **TwIsTeDMiNd** » Thu Jan 08, 2015 7:30 pm

This thread name should be changed to AI in Carmageddon - Worse than you thought

Espyo · Post by **Espyo** » Thu Jan 08, 2015 7:42 pm

This really reminds me of the AI in Super Smash Bros. Brawl. Some people claimed it learned with experience, others disproved it. I'm not convinced on either side, but I do know that my CPU went from purposely missing Dragoon strikes (sniper shots, pretty much), from missing them if I dodged, to following me mid-dodge and always hit me at the end of it. Though chances are that's more of an AI difficulty ramp-up than it is "learning".

Breakin'Benny · Post by **Breakin'Benny** » Thu Jan 08, 2015 10:07 pm

TwIsTeDMiNd wrote:This thread name should be changed to AI in Carmageddon - Worst than you thought

"Worst" is the superlative term for an Adjective, okay? Just sayin'.

Cybernettr · Post by **Cybernettr** » Thu Jan 08, 2015 10:25 pm

I can also say with 100% certainty that the game does not construct a database of previous play sessions to inform future AI decision making.

You may be right. I'm not making an absolute statement one way or the other. I just said that it might. The AI clearly does show goal setting behavior, and therefore at least short-term memory.

Cybernettr · Post by **Cybernettr** » Thu Jan 08, 2015 10:48 pm

Yfrid wrote:
Cybernettr wrote:Technology always seems "magical" until another piece of it is figured out.

not sure what do you mean? people doesn't figure out technology, people makes it. From scratch.

Then, like in consumer scenario, other people that has no knowledge about how some piece of electronic works, is allowed to buy it as a product for its features and may oversimplify its functionality as "magical cool tech thing" but there's no mystery behind that device. Just lack of knowledge from the end user (most of the times doesn't need to, after all.)

Or did you meant something else?

Okay, what I meant by that is that at one time, people said that if a computer could read printed characters in whatever font, that would be considered "intelligence," but then when it was achieved people said it had nothing to do with intelligence. Then people said that when a computer could understand spoken English, that would be "intelligence" but once that was achieved people, people said it was no big deal. Then people said that when a computer could translate languages…etc.

As soon as a part of the puzzle is figured out, then it stops being considered a part of "intelligence." I'm not saying these things are "intelligence," because I think a big part of intelligence is being able to apply something you've learned in one domain to a problem in a totally different domain, something which computers can't do. All these things fall into the category of "narrow AI."

Yfrid · Post by **Yfrid** » Thu Jan 08, 2015 11:33 pm

i assume those goals were considered sign of intelligence in their time because they expected to achieve such things in some fictional unnecessarily complicated way (say, emulating a real brain understanding of such things).

But then, characters reading and spoken words recognition were accomplished with some "boring" software compiled by some smart guys and today can be computed with ease by a common smartphone with no wow. They are not considered valid argument for intelligence anymore because we got them to work on ordinary HW with no big deal. So they revealed to be no big deal in the end.

Razor · Post by **Razor** » Sat Jan 10, 2015 8:37 pm

What the fuck............................................................

skyNET · Post by **skyNET** » Sun Jan 11, 2015 6:27 am

this thread in a nutshell:

You can't convince a believer of anything; for their belief is not based on evidence, it's based on a deep seated need to believe.

also

You can give a person knowledge, but you can't make them think. Some people want to remain fools, only because the truth requires change.

Sa1cor · Post by **Sa1cor** » Fri Jan 16, 2015 10:25 am

This is one quality troll. Seriously, making so many videos and even narrating them? And it lasted for more than a year even. That's world class. This needs to be preserved in history. Well at least it wasn't all for nothing, I and hopefully some other people that don't know much about programming will now have a basic understanding how AI actually works since Trent did a good job explaining that.

Electric Blue · Post by **Electric Blue** » Fri Jan 16, 2015 10:47 am

Sa1cor wrote:This is one quality troll. Seriously, making so many videos and even narrating them? And it lasted for more than a year even. That's world class. This needs to be preserved in history. Well at least it wasn't all for nothing, I and hopefully some other people that don't know much about programming will now have a basic understanding how AI actually works since Trent did a good job explaining that.

Trolling or not, it's not over yet.

CzechDeath · Post by **CzechDeath** » Fri Jan 16, 2015 1:59 pm

Pfff we are all in virtual reality, processed on quantum computer, everyone who found cheat codes would agree =D

So why the hell not...

Carmageddon2 is sentient beeing...OK? Ok...

The CWA Board