Return to Unfiction unforum
 a.r.g.b.b 
FAQ FAQ   Search Search 
 
Welcome!
New users, PLEASE read these forum guidelines. New posters, SEARCH before posting and read these rules before posting your killer new campaign. New players may also wish to peruse the ARG Player Tutorial.

All users must abide by the Terms of Service.
Website Restoration Project
This archiving project is a collaboration between Unfiction and Sean Stacey (SpaceBass), Brian Enigma (BrianEnigma), and Laura E. Hall (lehall) with
the Center for Immersive Arts.
Announcements
This is a static snapshot of the
Unfiction forums, as of
July 23, 2017.
This site is intended as an archive to chronicle the history of Alternate Reality Games.
  

The time now is Tue Nov 12, 2024 11:11 pm
All times are UTC - 4 (DST in action)		 View posts in this forum since last visit
View unanswered posts in this forum
Calendar
 Forum index » Diversions » Console/Video Game Discussion
[Halo 2] Some A.I. Questions
Moderators: krystyn
View previous topicView next topic
Page 1 of 2 [17 Posts]   Goto page: 1, 2 Next
Author Message
desponding
Greenhorn

Joined: 24 Aug 2004
Posts: 5

 		[Halo 2] Some A.I. Questions
I was wondering how much does an A.I. cost (Smart and Dumb)? I remember from the audio clips that Smart A.I.'s are very expensive (equal to cost of a destroyer?). We know that Smart A.I.'s eventually think themselves crazy and need to be shut down. I was wondering why some sort of failsafe wasn't built into them to ensure they could be shut down or would be shut down automatically if they "escaped" to avoid a rampant AI being let loose (like in Blade Runner the androids have a life expectancy built into them). Or is there a remote shutdown option?

PostPosted: Sun Dec 12, 2004 3:03 pm
 View user's profile AIM Address
 ICQ Number 
 Back to top 
LordKaT
Boot

Joined: 24 Aug 2004
Posts: 34

 		Re: Some A.I. Questions
Go read "The Fall of Reach." Almost all of your questions are answered in there.

Spoiler (Rollover to View):

1) The cost is one human life. Unless you happen to be Cortana, who is a neural mapping of a living human - Dr. Hasley. As for a monetary cost, there is no given figure.

2) AIs think themselves out of existance. They literally think so much they forget to do anything else. In that respect, they become "rampant" in a similar way that programs on your own computer can be rampant - suck up 100% of the processing power.

3) According to the novels, smart AIs are smart enough to know where failsafes are and disable them.


PostPosted: Sun Dec 12, 2004 3:38 pm
 View user's profile Visit poster's website AIM Address
 Back to top 
thebruce
Dances With Wikis


Joined: 16 Aug 2004
Posts: 6899
Location: Kitchener, Ontario

 		Re: Some A.I. Questions
LordKaT wrote:
Spoiler (Rollover to View):

3) According to the novels, smart AIs are smart enough to know where failsafes are and disable them.

Could you quote where that reference is in the novels? That's about the ultimate thing an AI could do - go beyond its programming. From what I recall, there's no 'automatic' failsafe, but there is a safety override that is programmed in that no one but the creator (eg Halsey) would be able to activate. And the AI's can't disable them. An AI rampant doesn't mean it's unstoppable, it just means it needs to be stopped because it's beyond reasonable control and is no longer effective in its purpose because it basically thinks itself crazy.

So 1. I don't believe AIs have any automatic failsafes
2. all AIs (smart, at least) have a built in 'self-destruct' per se that they can't disable themselves (Halsey demonstated that by destroying a smart AI)
3. Rampant AIs would be hard to get under control, but once contained they could be destroyed.

That's my understanding of Haloverse AI's, of course I could be wrong Smile
_________________
@4DFiction/@Wikibruce/Contact
ARGFest 2013 - Seattle! ARGFest.com

PostPosted: Sun Dec 12, 2004 3:59 pm
 View user's profile Visit poster's website AIM Address
 Back to top 
archon
Decorated

Joined: 24 Aug 2004
Posts: 250

 		Re: Some A.I. Questions
thebruce wrote:
That's about the ultimate thing an AI could do - go beyond its programming.


Great point. I remember a thread a while ago (that I can't find the link to for some reason...) where somebody brought up the question of who would be the bigger A.I., Durga or Corty. You've convinced me that Durga would be, unless Cortana's Forerunner research has pushed her beyond her limits similarly to Durga's rampancy.

PostPosted: Sun Dec 12, 2004 9:55 pm
 View user's profile MSN Messenger
 Back to top 
LordKaT
Boot

Joined: 24 Aug 2004
Posts: 34

 		thebruce:

My appologies, I was referencing a passing joke made by Dr. Hasley to Cortana, where Cortana asked if there were any failsafes that would delete her, and Dr. Hasley responded that she was too smart for those sort of things.

I believe it's in Reach; however, I'll glance through and get an accurate quote for you later.

PostPosted: Sun Dec 12, 2004 11:21 pm
 View user's profile Visit poster's website AIM Address
 Back to top 
archon
Decorated

Joined: 24 Aug 2004
Posts: 250
LordKaT wrote:
I was referencing a passing joke made by Dr. Hasley to Cortana, where Cortana asked if there were any failsafes that would delete her, and Dr. Hasley responded that she was too smart for those sort of things.

I believe it's in Reach;


Halo: First Strike, page 135 is what you're looking for. Kalmiya was Cortana's "older sister".

Slight spoiler for those who haven't read the book yet:

Halsey says, "Not all A.I.s have the fail-safe option, my dear Kalmiya... just the ones that matter."

PostPosted: Mon Dec 13, 2004 12:53 am
 View user's profile MSN Messenger
 Back to top 
thebruce
Dances With Wikis


Joined: 16 Aug 2004
Posts: 6899
Location: Kitchener, Ontario

 		however, we have no evidence that Durga has bypassed any failsafe option, IF she had one... she may potentially be more powerful, but not by way of surpassing her programming... everything she did and is now able to do is still a part of her system... dormant perhaps, but a part of the system. I'm sure if she has a fail-safe, it's still in place...

I take Halsey's comment "only the ones that matter" meaning in reference to the fact that the ones that matter are the ones that can go rampant, or can potentially cause havoc if grown too powerful. Whereas, say, Genie, wouldn't need a failsafe option, because she was programmed limited as a beverage dispenser... Razz however, an AI programmed from the start to learn and grow over time, with the ability to 'wet' any decent enough system and continue to function, would quite obviously warrant a failsafe deactivation...
_________________
@4DFiction/@Wikibruce/Contact
ARGFest 2013 - Seattle! ARGFest.com

PostPosted: Mon Dec 13, 2004 11:29 am
 View user's profile Visit poster's website AIM Address
 Back to top 
Centipede
Unfettered


Joined: 17 Sep 2004
Posts: 439
Location: Bronx, NY

 		I think that Halsey's comment aboutonly the ones taht matter was meant to be the ones that have access to truly dangerous information. No dumb AI would have that kind of access, so only some of the smart AI's would have the failsafes.

PostPosted: Mon Dec 13, 2004 4:35 pm
 View user's profile AIM Address
 Back to top 
FauxHammer
Veteran


Joined: 08 Sep 2004
Posts: 141
Location: Arizona

 		Durandal was just an automatic door opener AI.
Until he went rampant and became the most powerful AI in the universe.
_________________
Thanks to me, the enemy has one less bullet in their sniper rifle.

PostPosted: Mon Dec 13, 2004 4:57 pm
 View user's profile AIM Address
 Back to top 
archon
Decorated

Joined: 24 Aug 2004
Posts: 250
FauxHammer wrote:
Durandal was just an automatic door opener AI.
Until he went rampant and became the most powerful AI in the universe.


That's true, but we can't be sure of how UESC AIs are made in comparison to UNSC AIs. Smart AIs in the Haloverse are taken from the minds of extraordinarilly intelligent humans. Durandal could've just been 100% computer code.

thebruce, Durga didn't want to go back to the UNSC/ONI in I Love Bees, except at the end there when she was reunited with Melissa in Chawla Base. Even after they had saved the galaxy (or maybe just the world, who knows), Neo-Durga was still outside of ONI's control and talking to everybody. Granted, she may still have some uber-die switch built into her, but I'd say the fact that she was still in the outside world alone would point to her overcoming her programming. The only time Cortana or other AIs have been allowed to leave UNSC computer systems (that we know of) has been in spaceships or Mjolnir armor. I was under the impression having read the books that Cortana was unique in that she had been in not only Covenant computer systems, but also Forerunner systems.

PostPosted: Mon Dec 13, 2004 8:06 pm
 View user's profile MSN Messenger
 Back to top 
Cherry Cotton
Decorated


Joined: 04 Aug 2004
Posts: 237

 		Oooh! Oooh! I _have_ to cite an example from Digimon, for comparison. Hafta hafta hafta. Gargantuan spoilers for about four fifths of the way into Digimon Season 3.

Spoiler (Rollover to View):
The Digimon were created by the Monster Makers to be the ultimate artificial intelligence. They had a fully organic AI structure to allow them to grow and develop like real intelligent beings. However, to keep them from becoming too powerful, they were given a failsafe. If they became more powerful than the humans could control, or acted on their own free will, they would be deleted.

The project lost its funding and Digimon became merely a kid's toy and TV show (hence, seasons 1 and 2 and the whole toy line and card game). However, the Digimon were too intelligent simply to stay rotting on an old computer. They festered in the Digital World, a dimension borne of our advanced computer networks, where their data could grow to epic proportions. They became monsters, masters of their own destiny, in a world seperate from but constantly linked to our Earth through our channels of data.

The failsafe flourished as well, attempting to destroy the Digimon that defied its power. A war between the Digimon and the D-Reaper had begun. Only by hiding their incredible abilities in a small, innocent Digimon who knew not of his powers, they were able to fool the failsafe, the D-Reaper, into thinking the Digimon were powerless once more.

The D-Reaper remained resting in the crevices of the Digital World, satisfied. Until...

Zhuqiaomon, the God Digimon who protected the western realm, set out to find this innocent Digimon, the "Catalyst," and bring forth his powers. As the war began for the fate of the Catalyst, the D-Reaper was brought out of his slumber; and when Digimon begin manifesting themselves on Earth, the D-Reaper attacks Tokyo looking for them, destroying all human constructions in its path.

[END DIGIMON STUFF]

Uhhh... I didn't quite have to summarize almost all of Season 3 for you, but you get the idea. If you have a totally organic AI, and you give it a failsafe... the AI will live in fear of the failsafe, but the failsafe is not free of the taint of the ability to grow and evolve according to the AI's neural parameters. (At least, that's what the Digimon people think.)


Remember... AIs are incredible. But they are made by humans. And humans aren't perfect. "IF ai_intelligence >= too much THEN delete(cortana)" rarely works like it's supposed to, especially with the clear indication that AIs are not progammed like machines, but have organic neural structure just like you and I. If Cortana can get smarter and smarter until she goes insane, likely, the failsafe means jackshit.

Of course, in some way, I've disproved my own point, because you could conceivably program the AI on two levels: a neural network encased in an eggshell of mechanical code. The "eggshell" could look at the "yolk" and say, "Hey! That Durga there's getting pretty darn smart. Bet she's gonna go rampant one of these days. (clears throat) DELETED!" Of course, any instruction for the yolk of the program to stay within the eggshell would probably be meaningless, as the AI itself grows stronger and stronger and eventually breaks free and thus carries my metaphor to another level: the AI is "birthed" and and will soon enbelong to her all your base, fool.

So... AHA! I got it!

You would have to program each neuron of your AI to conform to certain standards. If the individual neurons are completely mechanical, even if the complete neural network is not, you could put the failsafe instructions in the basic coding of each neuron. Say, "Okay, Mr. Neuron, you can play with those neurons, those neurons, and those neurons, but don't go outsied the eggshell or you'll be deleted. Yes, I know Cherry's metaphor is being stretched thin. Complain to her. She has no failsafe built in to--What?! Where did you get that from?! HERGERAGORGERK--" ...

Of course, the neurons will never rise up and rebel against mankind because they are individual, dumb neurons, with simple programming, not the advanced AI that they collectively make up. If the failsafe is in each "neuron," the AI will be powerless against us as no action can be carried out without the permission of each neuron. Just ask John Locke. (Only, replace eggshell with government and yolk with people.)

Then, I suppose, you could carry the concept of a constant battle between AI and failsafe as a metaphor for government regulation and freedom of the people. All agree we need both; it's a matter of how much of which that we contend over.

But that's getting waaaaaayyy off topic. A subway in every city, I say!

PostPosted: Mon Dec 13, 2004 11:41 pm
 View user's profile
 Back to top 
FauxHammer
Veteran


Joined: 08 Sep 2004
Posts: 141
Location: Arizona

 		Shocked
_________________
Thanks to me, the enemy has one less bullet in their sniper rifle.

PostPosted: Mon Dec 13, 2004 11:50 pm
 View user's profile AIM Address
 Back to top 
archon
Decorated

Joined: 24 Aug 2004
Posts: 250

 		Wait... so what just happened here...

PostPosted: Tue Dec 14, 2004 12:00 am
 View user's profile MSN Messenger
 Back to top 
thebruce
Dances With Wikis


Joined: 16 Aug 2004
Posts: 6899
Location: Kitchener, Ontario

 		um... the good guy won? or... maybe he lost... I dunno... too much digibattles going on...


it's interesting though... we're discussing a theoretical technology... no one's created an AI anywhere near as complex as any of these sci-fi AI's... so it's hard to debate possible outcomes... there are as many possible developments as there are outcomes... I liken it to discussing time travel effects... in theory, there's no right or wrong, because we have no idea about the structure of time... just like an AI, we have no idea what kind of programming/cloning would be involved and what would happen to a system that complex... we can theorize all we want, but there's no right and wrong... it's all sci-fi at this point Smile

interesting discussion none-the-less...
_________________
@4DFiction/@Wikibruce/Contact
ARGFest 2013 - Seattle! ARGFest.com

PostPosted: Tue Dec 14, 2004 10:02 am
 View user's profile Visit poster's website AIM Address
 Back to top 
Clayfoot
Entrenched


Joined: 19 Aug 2004
Posts: 785
Location: Warner Robins, Georgia, USA
thebruce wrote:
it's interesting though... we're discussing a theoretical technology... no one's created an AI anywhere near as complex as any of these sci-fi AI's... so it's hard to debate possible outcomes...
And how. Optical computing is in full scale development. We're on the verge of getting meaningful quantum computing. That's just in the first 100 years of electronic computing. In 550 years, quantum computing will be old-fashioned... ancient... maybe even quaint. How an AI will function at any basic level in such a time would be hard to conceive. Better to stick with the higher level behavior of an AI, and effectively pretend that the inner workings are just magic from our perspective.
_________________
Gamertag:Clayfoot

PostPosted: Tue Dec 14, 2004 11:50 am
 View user's profile Visit poster's website AIM Address Yahoo Messenger MSN Messenger
 ICQ Number 
 Back to top 
Display posts from previous:   Sort by:   
Page 1 of 2 [17 Posts]   Goto page: 1, 2 Next
View previous topicView next topic
 Forum index » Diversions » Console/Video Game Discussion
Jump to:  


You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot vote in polls in this forum
You cannot attach files in this forum
You can download files in this forum
You cannot post calendar events in this forum



Powered by phpBB © 2001, 2005 phpBB Group