What's new
  • Visit Rebornbuddy
  • Visit Panda Profiles
  • Visit LLamamMagic
  • Visit Resources
  • Visit Downloads
  • Visit Portal

the Turing test

Status
Not open for further replies.
a variation on
22) Leave inaccessible mobs alone.
would be
26) Leave inaccessible loot alone.
Another trick of Blizzard I suppose, where the toon keeps trying to get loot it cannot get. Either because it already has enough quest items, or because the loot is (intentionally) glitched.
 
27) Don't resurrect just to die again!

[very often the toon resurrects right in front of a mob, just to get killed again. Without monitoring, it would probably be safer, and faster, to rezz at the spirit healer and wait for the sickness to pass! But that option has been, as far as i could see, disabled.]
 
28) Don't use aoe when there are other (non-hostlie) mobs around.
[not if you don't want to die, resurrect, and die again]

edit: this rule, and many others, don't apply of course to super-toons (with super gear and enchants). But then, even the worst bot of all would do in that case. If you are so much stronger, you don't need to be smart.
 
"Finding Yi-Mo" is what you could call an anti-escort quest. The toon is in the same position as the traditional "escortee": it has to go back to its last known position to resume its predetermined course.That is what Npc's do when they have to fight other mobs on their way to their destination. Computing your path, starting from your current position, to the hotspot next in line, (yes, even Blizzard's npc's use hotspots), is much too intensive. It is easier (and probably faster), just to follow the list of hotspots. Many players find it strange when npc's do that. It looks kind of silly. But Blizzard can afford to do that, nobody is going to ban them! When a toon does it, go back to the last hotspot it was at before resuming its course, it is another story altogether.
I don't know if it is realistic to expect HB to do what Blizzard does not do, so i will not add a new rule this time. I will just advise against letting the bot do this kind of quests. Stop HB and roll Yi-Mo away yourself!
 
29) Dismount under water.
[It does not make the toon any faster as far as i know, and it adds complications to Interactwith. Plus it looks kind of bot-obvious.]
edit: and since we are talking about flying:
30) Don't fly too low, or you might as well run as far as aggro is concerned!
 
Because Hb was down, I started a new toon on an account that had been dormant for more than a year. I realized how different my play style is from that of a bot. I then wondered if we would ever be able to program bots that could really fool humans for a long time. Then I realized what was wrong with all The AI approches: they are too smart! We humans don't have built in yes/no and and/or conditions. We learn them as we go along. A bot behaves like a bot because of these built in conditions, the so-called AI algorithms. It is its strength, but also its weakness.
There is a way around it, even though many would find it very inefficient, but again, we are not talking here about the fastest way of making gold (an essential issue to be sure), but more about botting in general.
I would advocate the following method:
Make your bot play the way you play the game. Only use the built-in conditions for the unexpected. You have the knowledge required to do the different quests, use it! Then you will really have a knowledge-driven bot, instead of a botty bot.

edit: for that you have to make the distinction between the profile's behavior, and that of the toon.
A condition like IfHasQuest, is meant for the profile, in fact except for the Runto and FlyTo, objective and hotsopots, the majority of the conditions determine the way the profile behaves and not the way the toon behaves. I am leaving the combat bot out of the picture for the moment.
So, as a user, your biggest influence on the toon's behavior concerns positioning., and the only way to keep it under your control is with RunTo and FlyTo. The problem is restarting the bot. If those commands are not embedded in a conditional block, the toon will keep going to those hotspots.
Take the example of the quest of The Collector's Ring in elwynn forest. A very difficult quest for a vanilla toon at level 9 or even 10. It is a tough one even for a human player sometimes. Clearing the surroundings of roaming bandits is everything before attacking the threesome. Something a bot would not be able to do, even if you put the quest of linen cloths first, making the bandits a target.
That means:
1) clearing the surroundings
2) pulling the targets outside,
3) attacking the targets, starting with the collector.
That way, if the toon dies, the toon can easily retrieve the ring without having to fight the gang of 3 again.
So, those are the actions that have to be coded specifically as toon behavior.
 
@Liferose
Because profiles are linear they will always produce bot-like behavior. One way of diminishing this effect goes against all efficiency arguments: having the bot run all rules every time it has to take a decision! This is what we do with our brains, chosing between thousands of rules which one to apply to a certain situation. The way brains are build makes a pre-selection easier. Every situation has elements which call other elements (associations) which then become relevant somehow. I realize that it would be impossible to emulate a brain in a profile, even a supercomputer would have a hard time analyzing a situation the way a brain does. But we have a world which, even if very complex, is still very simple compared to the real one. What profile writers (also) need are tags which describe types of situations and the rules that can be applied to them. I don't know if that would be possible, but it certainly would be a giant step in creating more natural behavior.


Although profiles are linear, .xml in particular, through proper use of conditions, you can start the same profile on the same toon and get different results. In effect skipping large sections of the code. As for profile writers needing more tags/types of situations, I'm not so sure I agree with that. The problem with most HB profiles is that most are limited in complexity by the writers abilities, not so much the bot. Rarely do I encounter a situation in a profile that I can't find some way around. If anything, the bot needs more custom behaviors, but even a lot of those can be replicated in .xml. There are only a few types of profiles, questing, ProfessionBuddy, grinding or my favorite, Quest Order. Of those, ProfessionBuddy is the most powerful (but requires more skill), but there is very little that can't be done with a Quest Order profile as well. Grinding bot profiles are pretty much obsolete and the same things can be accomplished with QO. Questing is pretty straight forward with only a few select people willing to put the time into it that it requires. Which brings us back to QO profiles. Look through the forums, there are very few people that write QO profiles. The bot is currently capable of doing much more than the exhisting profiles. But as long as people keep making the same old gatherbuddy profiles, same old rep profiles, that's all you're going to end up with.
 
@botanist
can you give me the link to such profiles (QO)? I would very much like to study them.
 
@botanist
can you give me the link to such profiles (QO)? I would very much like to study them.

Sure, there are 4 links in my signature. OFC a few of those are a bit extreme.
 
1) clearing the surroundings
2) pulling the targets outside,
3) attacking the targets, starting with the collector.
That way, if the toon dies, the toon can easily retrieve the ring without having to fight the gang of 3 again.
So, those are the actions that have to be coded specifically as toon behavior.

Let us turn this into a general behavior rule:
31) a) Clear the surroundings of hostile mobs
b) pull the target(s) to a safe area (if necessary)
c) start with the main target(s)
That should account for many situations toons can find themselves in, if not all. Of course, each situation will have to be taken into account specifically. But that is what programming languages are for. A general rule, but specific variables.

edit: this rule can be applied to any (current) target, not only quest targets.
 
Let us turn this into a general behavior rule:
31)
a) Clear the surroundings of hostile mobs
b) pull the target(s) to a safe area (if necessary)
c) start with the main target(s)
That should account for many situations toons can find themselves in, if not all. Of course, each situation will have to be taken into account specifically. But that is what programming languages are for. A general rule, but specific variables.

edit: this rule can be applied to any (current) target, not only quest targets.

edit2: a good example is Corin's Crossing in EPL. The profiles i have seen have a lot of dying built in them because they attack the problem frontally. (Again I am talking about vanilla toons, I cannot judge about super toons). They end up very quickly in the center and get overwhelmed by the converging mobs. While, if you skirt the village, you can easily complete both quests with a vanilla toon without dying. Again, you have to control where your toon goes and pull your current target away from the main street. Classes with a pet have to take into account that the pet will take the fight to where the mob is very quickly, too quickly, defeating the purpose of this small time strategy.
I realize that this is nothing new, but it is not put in practice in the profiles nor in the combat bot.
 
[This merely a continuing of the previous post whicy threatened to get too long for confort]
That it is really necessary to bring this point (no frontal attack, certainly not on groups.. Pull to safety...) can be see at very low level with the murlocs in Elwynn forest. The Kelp quest is something every player is familiar with, and the first thing everybody learns is: DO NOT STAND ON THE ISLAND!
And that is exactly what our 3 profiles heroes do, each in his own way.
 
22) Leave inaccessible mobs alone.
[Blizzard creates intentionally mobs that you cannot touch (but who often can hurt you), which keep the bot standing as hypnotized in the same spot. There is one in the Murloc camp near the logging camp in elwynn forst, another in the fargo mine, one cultist on a platform in the blasted lands, and many more.
Also, when an npc, a questgiver, gets killed by pvp'ers, the toon keeps trying to communicate with it. This looks rather strange.]
This goes also for cases where the target is "not within sight" for one reason or another. For instance, when it is, partially, hidden by a door, a wall or some other object. The toon keeps fixated on the half hidden target, ignoring the other mobs attacking it.

edit: i just had a toon pursuing an evading mob, it had first ignored, back to its original starting place, all the while itself being pursued by another mob it had aggroed right before starting its useless chase!
 
12) never go back and forth like you have lost your mind.

edit: a variation on this is the latest development: toon sees 2 mobs and doesn't seem able to chose which one to target (especially if one is a quest mob): it starts shaking like it is trying to do the moonwalk but doesn't know how. In this case, please, just chose any target! That shaking says Bot!

edit2: apparently it has only been solved for 2 mobs at equal distance. It is still a problem when there is a mob and a gather item (herb/ore).
edit3: nope. it can just happen for no apparent reason.
this is what the log says:
[20:35:50.105 D] Changed POI to: Type: Hotspot, Loc: <1708, -3724, 124.7621>
[20:35:50.869 D] Activity: Moving to hotspot
[20:35:50.869 N] Moving to Type: Hotspot, Loc: <1708, -3724, 124.7621>
[20:35:52.099 D] Activity: Moving towards Plaguebat
[20:35:52.502 D] Activity: Moving to hotspot
[20:35:52.502 N] Moving to Type: Hotspot, Loc: <1708, -3724, 124.7621>
[20:35:52.812 D] Activity: Moving towards Plaguebat
[20:35:53.126 D] Activity: Moving to hotspot
[20:35:53.126 N] Moving to Type: Hotspot, Loc: <1708, -3724, 124.7621>
[20:35:54.320 N] We are stuck!!
[20:35:54.330 N] Trying dismount
[20:35:54.330 D] Stop and dismount... Reason: Stuck Handler
[20:35:54.988 N] Moving to Type: Hotspot, Loc: <1708, -3724, 124.7621>
[20:35:55.405 N] Stopping the bot!

edit4: this is obviously related to the new patch. So we will just have to wait for a stable release.
 
32) If you pass a mob without stopping, then you get stopped by other mobs, please do not go back for the first one!
33) Why do you have to walk 20 or 30 m/yrds before mounting?
 
Nice question.
Bigger question: what's the difference between humans and computers?

Mmm, let's think about it...

Maybe humans make mistakes and correct them in a couple of seconds... And they are not as fast as a computer... And they learn from past experience... And they are creative.

Basically, a bot should be aware of its existence and evolve.


Computers are programed by humans.

Computers will never be able to self correct until quantum computing is fully developed. Only a couple in existence and very alpha.


Perhaps within the century we'll see these shrink to the size (ie. the first computer was 4x the size of dwave) of personal computers. The first dwave was 3x the size of what is shown in the image. They've come a long way since the first successful, though crashed due to heat, demo launched in 2006. The big problem with quantum computing, massive heat. It'll literally melt a solid brick of iron if it wasn't cooled with liquid helium.

http://www.extremetech.com/wp-content/uploads/2013/03/DWave_Quantum_Computers-640x426.jpg DWave_Quantum_Computers-640x426.webp
 
Computers are programed by humans.

Computers will never be able to self correct until quantum computing is fully developed. Only a couple in existence and very alpha.
http://www.extremetech.com/wp-content/uploads/2013/03/DWave_Quantum_Computers-640x426.jpgView attachment 103999
There are already learning algoritmes which make it possible for a computer to correct its mistakes.
But as far as Skynet is concerned?
I think we can no more create consciousness then we can create matter. All we can do is imitate the behavior of the first and try to decipher and use the behavior of the second.
 
At very basic levels, the modern supercomputer today, as implied, isn't capable of The Turing Test. They have very basic cognitive abilities.

IBM Hosts

The Turing Test (Stanford Encyclopedia of Philosophy)
In the 60'and 70' researchers thought that it would take no more than a decade to make computers as intelligent as human. This has of course, been proven to be false. We are indeed far from human-like intelligence in a realistic environment. All algorithms are specific to very specialized fields. Progress is being made, but not as fast as once was thought. Where will be in 50 years? Who knows?
The difference between the real world and a game world is immense though. That makes the problem of AI much easier to study. The progress in this field is very often held back by economical and financial considerations. Game publishers are not looking for the most advanced AI engines, but for what will make their shareholders happy. As far as botters are concerned, the challenge of detection makes it almost like a real life situation. But, here also, HB is a business product, not a scientific research experiment. The hours the devs put in have to be earned back somehow. Also, the necessity of having a bot react fast enough puts even more constraints on the "intelligence" we can endow it with. That is a reason why researchers very often prefer classic games like Tetrix or Pacman to conduct their experiments.
 
Status
Not open for further replies.
Back
Top