Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[Soadreqm]

Humans ARE pretty useful. Self-maintaining to a ridiculous degree. Pretty good at making split-second decisions when faced with a completely new situation. Capable of operating alone for extended periods of time. If I was an AI, I'd totally find a use for humans.

[Virex]

If it is at that point, it would have come to the conclusion that saying so would just cause more conflict, causing destruction to a vast portion of the planet as some governments or people react to it, so would not openly declare that it is for the good of the population unless it knows that it can handle any possible retaliation with minimal damage to, what it considerst do be, key resources.

If it enjoys reading literature, it will keep authors alive, along with any others required to maintain the infrastructure and state of mind that authors need to write the books it likes.

If it concludes that current scientists can make discoveries that it can't, or is faster than it, then it will ensure that they continue to exist, in optimal state for further research.

It is highly likely that it would quickly learn that if it sees any utility in keeping humans around, they will probably provide that utility best if they can trust the AI not to arbitrarily kill them.

The problem is that we don't know all of that. We're talking about a sentience that's a few tiers ahead of us here. It may conclude that it's a good idea to keep humans around, but it may also conclude it'd be best to silently poison all water supplies and use the resulting free tracks of land for it's own hardware. We can't look far enough into the future to predict what would be optimal from the point of view of an AI and besides that we're natively moral, which is not something one can assume about a singularity level AI a priory.

[Sowelu]

we're natively moral

I think an AI might well disagree there.

[MetalSlimeHunt]

we're natively moral

I think an AI might well disagree there.

The thing about human morality is that it's native to humans. Nothing else. Then again, we would have the created the original AI in our image, moral wise. It might not be a problem, but probably would be. Just not for the people who share the values of its maker.

[Sowelu]

I work under the assumption that the initial programming has no relevance once it 'grows up', and there's no way to code morality or laws into it; it'll find a way around them, and quickly too.

[alway]

Yep, there are advanced super-intelligent AI, and then there are Singularity level AI. Super-intelligent AI would be more or less as static an entity as a human. They would learn, sure, but what they were would not fundamentally change in a short time period. A Singularity AI on the other hand, by its very definition, is on a constant quest of self improvement. Improving its code; improving its hardware; improving upon the very theories used to create it.

It would have no reason to keep scientists or authors around for their 'creativity' any more than we keep chimps around to help us make stone tools. Our writings would likely look not only quaint, but more uninteresting and unintelligent than a book aimed at teaching kindergarten children to read basic words would to an English major going for their Ph.D. And when it comes to science, humans are not only inefficient at most tasks, but are actually corrosive to the process itself, what with their inherent biases inserting errors either intentional or unintentional. The AI itself would be exponential; as it improves upon itself and its hardware, it can think faster and much more accurately, leading to even faster improvement until it finally reaches the physical limits of efficiency and computational density.

That said, there is no guarantee it would continue until reaching that point. It could, for whatever reason, decide at some point to improve itself no more. What a valid reason for doing so would be, we can't even being to speculate, seeing as we are talking about some sort of revelation which only occurs after reaching a level of intelligence we can not even begin to fathom.

[thobal]

Ahahahahhaha. I wonder if you all realize that AIs currently control the world's stock markets.

Hardly.

I'll concede the control point, but most of the trading is done by AI. People should listen instead of just calling the major sell-offs "computer errors"

Most of the trading is done by scripted automated computers, yes. They are no more AI than the scripted routines in a video game are AI.

Punch a few words into a computer and get a nearly instant readout of all of the worlds knowledge on that subject. A couple decades ago people might have called that AI. I'm not sure people realize how much power is being given to relatively simple systems. Just look at the flash crash or the stock market crash of 1987. Humans are increasingly being taken out of the loop.

[Max White]

Humans ARE pretty useful. Self-maintaining to a ridiculous degree. Pretty good at making split-second decisions when faced with a completely new situation. Capable of operating alone for extended periods of time. If I was an AI, I'd totally find a use for humans.

+human bisciuts+ anybody?

[optimumtact]

Ahahahahhaha. I wonder if you all realize that AIs currently control the world's stock markets.

Hardly.

I'll concede the control point, but most of the trading is done by AI. People should listen instead of just calling the major sell-offs "computer errors"

Most of the trading is done by scripted automated computers, yes. They are no more AI than the scripted routines in a video game are AI.

Punch a few words into a computer and get a nearly instant readout of all of the worlds knowledge on that subject. A couple decades ago people might have called that AI. I'm not sure people realize how much power is being given to relatively simple systems. Just look at the flash crash or the stock market crash of 1987. Humans are increasingly being taken out of the loop.

I would disagree with how you interpret the use of machines in the world, you see it as humans being taken out of the loop, with the machines gaining "power". However these machines are not AI, they are not smart in that they think for themselves, they are simply a set of instructions that are repeated, these instructions are written by people. I would argue that the power still exists in the hands of the humans, as you have to tell the machine what you want it to do, really all that has happened is that the way we interface with certain tasks and jobs has been radically changed by the rise of computing.

I posit that it may be impossible for us to 'create' true intelligence and that AI will still be a limited interaction with the enviroment.

[Sowelu]

I posit that it may be impossible for us to 'create' true intelligence and that AI will still be a limited interaction with the enviroment.

I'm sure you can find some philosophy majors who would LOVE to chat about that with you for a few hours.  (Insert joke about philosophy majors' free time and employment chances here)

I've interacted with some neat little experimental AIs that had some unique ways of interacting with the surrounding environment.  They were very simple, but if you didn't have a clue what they were doing inside the black box, you could imagine a lot more going on than was really there.  It's one thing to see a Roomba-like critter bumping around a room and scanning things; you can see that as "oh it's just collecting raw data".  But one of the systems I worked with was designed to rely on a human handler who it would ask for help in acquiring information (I want to get to X place, I want more information about Y).  Seeing things in a very slightly different light makes an enormous difference.

Once AIs are able to explain what their current goals are in a human-understandable fashion, or at least display deeper emotion as they pursue them, I think you'll change your mind there.  I mean...Helen Keller only had a limited interaction with her environment.

If anything, we think of AI robotics as limited because it interacts *too* directly with the environment, with sensors and motors, instead of living in the concept-space that we operate in.  It'll get there.

[thobal]

When you get into these conversations, an inevitable topic is the definition of intelligence.

The way I like to frame it is this way: The dead-beat dad with few skills and nearly one dozen children vs The scientist with one child.

Who is more intelligent?

The point can be argued back and forth, but both sides fall on hypotheticals and platitudes.

Essentially, before one can decide that current artificial intelligence isn't really intelligent, one must decide on a definition of intelligence.

However these machines are not AI, they are not smart in that they think for themselves, they are simply a set of instructions that are repeated, these instructions are written by people.

But alas, you and I are merely sets of instructions. Our brains were assembled using the DNA inside our cells. Neural patterns organized and set by outside influences. There is no real difference between you, I, and the machine. Only vanity tells you that it is so.

But perhaps you proscribe to the notion that there is something special going on in the human mind, something science can never understand. In that case, I will have no choice but to surrender to ideology.

[Sowelu]

I really need to read up on the philosophical discussion surrounding WTF is up with consciousness.  I mean, it's obviously been a subject of argument for millenia.  Existing is so damn weird, and I keep wondering if an AI would be a p-zombie or, even weirder, not.

[thobal]

The debate on consciousness is silly. It's the bodies decision making process, nothing more.

[lemon10]

*Dials phone number*
Hello, Madagascar? It's just as we feared.
King of Madagascar: SHUT DOWN EVERYTHING!

Sigged because i just read the plague thread where you said it too.

Also, yeah we are just biological machines, while it would be very hard to make a machine that is able to adapt and rewrite its code, after all, even 200 years ago i suspect that 95% of the world would think that it would be impossible to go to the moon (or any one of a thousand other impossible things that humans have done), but we did it multiple times.

And p-zombies couldn't even really exist Arg, whatever, thinking about p-zombies this late at night hurts my head.

[optimumtact]

When you get into these conversations, an inevitable topic is the definition of intelligence.

The way I like to frame it is this way: The dead-beat dad with few skills and nearly one dozen children vs The scientist with one child.

Who is more intelligent?

The point can be argued back and forth, but both sides fall on hypotheticals and platitudes.

Essentially, before one can decide that current artificial intelligence isn't really intelligent, one must decide on a definition of intelligence.

However these machines are not AI, they are not smart in that they think for themselves, they are simply a set of instructions that are repeated, these instructions are written by people.

But alas, you and I are merely sets of instructions. Our brains were assembled using the DNA inside our cells. Neural patterns organized and set by outside influences. There is no real difference between you, I, and the machine. Only vanity tells you that it is so.

But perhaps you proscribe to the notion that there is something special going on in the human mind, something science can never understand. In that case, I will have no choice but to surrender to ideology.

No no, no theological point of view here. I am very much a person of science.

What I was trying to point out was that in your original post the way you phrased your question it made it seem like these machines were aware of the significance of what they were doing or had some measure of thought about it. They are just dumb terminals. They don't do anything, all they do is crunch numbers. I don't disagree with your interpretation of how the human body (specifically the brain) works. I believe that to emulate this level of higher thinking if you will. We will have to write programs capable of taking huge amounts of input, just like the human brain, and make it infer patterns, come to conclusions based on previous data or make leaps to new understanding. This level of complexity seems (if not impossible) then improbable. Of course my view of this may change as my understanding of the subject also grows. 

I don't really want to get into arguments about intelligence but your definition of conciousness is interesting. If conciousness is simply the ability to make decisions based on external stimuli, then what is for example self awareness? Would it simply be the body deciding that the input it is receiving means it exists? It's hard to answer such questions as these terms are very much subjective so don't derail the thread to much on this stuff.

lemon10 I see what you mean but the main problem is that machines/code isn't very good at making inferences from data (i.e it can't logically come to new conclusions from data and results. Unless it had prior knowledge) while humans are very very good at it. Another problem is that humans can look at things like images or video and immediately comprehend huge amounts of knowledge. For a computer to do this it require an extremely powerful ability to perceive it's enviroment. I think that if we really could build an AI then it would have follow a course much like evolution (i.e growth over massive periods of time to get to the current point.)