forum.comicostrich.com

[AaronLee]

This debate has gone on between my friends and I for an almost indefinitely long time, almost escalating to total war a la Outsider, sweeping up others in the ongoing conflict and showing no sighns of reconcilliation . This question, I pose to you (forum goers and the podwarp panel if they're interested;)

If artifical intelligence was created/born, would it attempt to take over all it could reach?

My friend dumbly argues "YES! Computers will stab us in the back no matter what!" without ever explaining himself. I think he's watched too many half-thought-out S.F. movies. Anyway; I take the argument opposite him. In most popular S.F. movies I've seen, AI's are always perfectly efficient, engineer and emotionally detached and fast thinking intellects, which really confuses me.

This is mostly because my experience as a coder has suggested that computers as we know them are horrendously dumb. Want a robot to walk? Okay, code it. How about climb stairs? It's just as much code. Now, how about getting it to ponder the meaning of existence? That may be tougher.

Computers are fast only because we've designed any operations they perform with forethought and top-to-bottom engineering. Now, humans aren't top-to-bottom-engineered (I'm not christian, so don't hit me with the God argument.)

More than anything, our mental development is an ad-hoc self expansion of neurons building systems from unrelated bits and improvising as time goes on. I reason that's why we think so "slow" and why anything that thinks and acts remotely like us (napoeanic domination of the universe as a prime goal, xenophobia, shortsighted urban engineering etc) would ultimately be just as constrained.

I could go on forever, but I also have a sense of restraint. If you're so inclined, however, discuss!

[zortic]

By AI I assume you're talking about computers with the ability to code themselves (admittedly this is currently a fictional ability). That would remove the argument that we're in control (assuming we'd allow that to happen). So if a computer had control of its own development that would open up a range of reactions as wide as a human's.

As for the question of world domination, I would think that it depends on how intelligent their artificial intelligence is. If they're smart enough they'll realize that world domination isn't worth it, too many malcontents and paperwork. They'd either let us keep our illusion of superiority and start to subtly manipulate us, or they'd just up and leave us for bigger and better things.
_________________
Check out Zortic, ETI-PI, Abby's Agency, Podwarp 1999, and the WCCAs

[dph_of_rules]

I can see robots/ai rebelling if we start treating them as 2nd class citizens. (One of my beliefs is any society which depends on a slave class for its continued existence is morally corrupt and doomed to fail). I like the idea of developing cars that vehicles us places. When it comes to developing personalities, I'm not sure we have the computer capacity to do that. When you consider the computing capacity it takes to run a Windows operating system and compare that to what we consider to be an interactive system, I'm convinced that it would take a supercomputer to simulate that kind of intelligence.

As a precaution, I'd rather limit how far we are allowed to develop ai so that we don't have to deal with the situation of treating ai as an equal.
_________________
Whatever happened to simplicity?

[AaronLee]

The message I get from these posts is we all have pretty different impressions of what Artificial Intellignece will be. I suppose it's mostly up to the imagination. Zortic, you've hit one of the possibilities, and the most potent theoretically possible. But, yes, the most likely.

Dph_of_rules, I'm not sure you could compare any intelligence to Windows . All jokes aside; it would be hard to tokenize an intelligence as simply a frighteningly huge and frighteningly complex program. My point was our current method of creating functionally intelligent programs really couldn't make something remotely like an intellect.

That's becuase manual coding (what we do now) usually requires a lot of understanding of the subject one is tailoring the program to. If there's anything we don't really udnerstand, though, it's ourselves. We vaguely understand the processes that give rise to forms of intelligence, however. Such things as stimuli-response seeking behavior, as with Kismet.

That's mostly why I belive the typical skynet-esque backstab-rebellion is fairly unlikely - at least with the sort of AIs we might actually see. Most sane humans don't backstab (otherwise we'd likely be in anarchy, assumably.) One could attribute this to their social drive and attachment to others. I think it's a bit wonky to think we'd have computer based intelligences that would even be remotely able to work like us unless they live for the most part like us. This also ties in with dph's point.

So anyway, I'm still on a limb as to weather or not anyone here considers this typeof stuff from me as good discussion. I'm just asking to make sure I'm not being a motormouth spam monkey.[/url]

[dph_of_rules]

I have studied computer programming in a few languages. The process I use is what you describe: figure out how I would do something and then make the computer follow my steps. However, that does not always work.

My analogy is as complex as Windows is, true artificial intelligence would be a couple of magnitudes more. We all know how glitchy an operating system is so I imagine a true ai would be . . decades away. Have you ever stopped to ponder how we really think? Or stop to consider the endless number of tasks that go on that we aren't consciously aware of? That's where the sheer complexity gets in. Just consider the process of walking and everything involved in it: physically moving the body, readjusting the balances to avoid collapse, etc. That's a whole set of subroutines. Then stop to consider our verbal and non-verbal speech patterns. For machine to understand that and catch the humor is incredible . . consider the difficulty poeple have in learning languages.

I see HAL as a threat to humanity as a benefit to humanity. We aren't ready for it yet and I'm not sure that we'll ever be ready.
_________________
Whatever happened to simplicity?

[Adam_Y]

Considering the fact that I remember someone recently comparing the state of artificial intelligence to a brain dead cockroach this is a rather odd question to be asking right now. Then again

[dph_of_rules]

The only reasons machines would bother conquering the world is pure ego or self-preservation. If humanity decided to declare war on ai robots, ai robots would have no choice but to fight back. How do you really beat an ai? You would have to delete all of its code from its all computers to be really sure.
_________________
Whatever happened to simplicity?

[AaronLee]

Argh! My friend has already infiltrated this forum! Okay, that was sarcasm.

Anyway, Dph, what makes you think we could evwr code an intelligence in the way you're speaking of? My initial point about coding as it stands today is: it's about as unlikely that we'll "code" an AI as would be building a steam starship in the 1900s.

Assuming the scientific point of view - Humans, as an example, are the result of billions of years of imrpovisation and refinement, and that was just to get the human brain. To think that we could achieve a similar end by different means seems odd to me.

Remember that AI is a loose term. That means the methods could be different. My reasoning is that if we were to create AI's that talk and "think" by our own reasoning, as well as evolve at least as fast as humans, we'd have to do as the romans do, as they say.

Roughly speaking, it could be possible to replicate biology and neurological mechanics as a tool for AI development. That is we'd have AIs that develop intelligence through interaction,the same way it's been done for eons. If they lived around humans and came to rely on them for interaction, what makes you think they'd want to rebel?

@Zortic: The "up and leave" bit reminds me a bit of some fiction I've heard about. The premise was that humanity had created AI's so strong, that their exodus was more akin to a dissapearence than a rebellion. They ended up cracking some law of the universe enough that they felt it a lovely nod to come back to earth on their anniverary and grant everyone's greatest desires. A bit cooshy, but, hey, we feed pigeons, don't we?

[dph_of_rules]

[zortic]

[Ubu]

I've no problem with AI in exactly the same way I have no problem with falling from high heights, it's the part where I hit the ground that bothers me.

I think AI in the sense of something that can learn and improve itself, so long as it has limits. If I happen to have a deep space vessel that I live on, I wouldn't put it in charge of something such as life support or exterior door control. That and not give it something it can PHYSICALLY improve itself with.

So what do you do in the robot uprising? Dig a moat BZZZT!

[DrSaltine]

I just started re-reading Dan Simmons' Hyperion Cantos (which is going to be made into a movie BTW), and i like how he handles this. In his universe, the AIs revolted from humanity, and then rejoined as partners. Within the AI community, there are three schools of thought. The Stables, who believe that a symbiotic relationship with humans is beneficial, the Volatiles, who believe that humanity should be exterminated at the earliest opportunity, and the Ultimates, who consider humans as relevant only as far as they relate to their project to develop the "ultimate intelligence", for now, they are willing to accept the existence of humans, but if humanity ever gets in the way of the project, the Ultimates will not hesitate to eliminate them.

I am ignoring the fact that in the later Hyperion books these categories are revealed to be gross overexagerations.
_________________
Professor Saltine's Astrodynamic Dirigible.
A vicotorian-era inventor who builds a spaceship.
www.professorsaltine.com

[Adam_Y]

[Chaos]

Just in case anyone's interested: Jim was describing a use of Genetic Algorithms in Episode 17, which are (oddly enough) what I was working with while listening to the podcast.

As to contributing to the topic at hand and doing a bit of speculating...

My own vision for AI places it in more of an advisory role for the foreseeable future: Absorbing extremely large amounts of information and providing solutions. AI systems can (and to a limited extent already do) assist in resource management, medicine, law, education, and research, amongst other things. AI can similarly (and to a very limited extent; already does) assist individuals with their day-to-day lives.

AI will definitely permeate our lives to increasing degrees, even if the vast majority of people are unaware of its presence or how it works (you could say that about most technology though). I imagine most people would be shocked at how much AI technology they interact with on a day-to-day basis (at least indirectly and on average).

As for the doomsday scenarios...

Placing any kind of "fuzzy" or mathematically unprovable software in control of a weapon system (at any but the lowest levels) would never happen. The systems are also (supposedly) completely isolated from outside interference. Most larger weapon platforms do contain some form of "strict" AI - usually at least one Expert System. These are usually just for subsystem management and fault identification, things like that.

I don't see humanoid robots (the common hardware platform for AI in sci-fi) becoming commonplace any time soon - the power source is one of the main issues. I shouldn't imagine we'll see them in a domestic setting this century - I can't really think of many uses for a general-purpose domestic robot... Although I can see the benefits in things like smart houses and small autonomous devices like robotic vacuum cleaners. Commercial/military robots will become quite frequent I think, although I expect the latter will be under direct or at least high-level human control (with AI algorithms calculating locomotion dynamics, balance and such), and the former would be largely ineffectual as a weapon in any case (in case of hacking/tampering or what have you).

The potential of nanomachines as a malicious tool worries me more than any other scenario I have come across. Nanotechnology is still in its infancy however - as I understand it we are a long, long way from seeing any sort of practical application (malicious or otherwise).

[AaronLee]

Oh my dear god my topic has asploded! (glee) sorry I missed the call in show, guys! I was busy watching brightly colored club racer cars go fast and spin out (nothing quite like it, plus it's free.)

*ahem* anyway. I'll edit when I finish catching up.

[edit] Good points all around! It's interesting to see the discussion's evolved pretty drastically since my extended absence.

You've raised a pretty valid point, Chaos. Real AI can be defined as anything that acts with intelligence, side-winder missiles and industrial machinery included. I suppose - from a practical standpoint - it's fairly likely we'll see improvements in logistics systems to reduce manpower needs in just about everything. Naval warships with less crew, factories maintained by small maintenance teams et cetera.

Anyway I'm sleep deprived, so ttfn :3

And, I'm back. Chaos, that avatar reminds me of Crimson Dark, are you familiar with that webcomic? I also have a sneaking suspicion...

So, anyway, I wrote on this stuff while writing for Sunrise. Anyone into remotely beleivable science fiction and AI related issues should check this out (in my signature.) Essentially, it's about as galactic network of civillizations that have to deal with various forms of AI just abotu very minute of their lives.

I took a distinct approach to AI you all might find interesting. In the physiology of AIs in Sunrise, there are smart ones and dumb ones, both behave very differently and have differing places in relation to so called "naturals" or natural intelligences. It's also fairly neat ebcause I look at some of the more comical effects of complete cybernetic bodies... scary!
_________________