The Singularity

Spoonist · Post by **Spoonist** » 2003-01-16 02:10pm

The problem with your ascertion(sp?) is that even if computers would get 100-1000 times faster then they are today, that still wouldn't put them close to a biological brain.
Not even by a far shot.

Biological brains can make logical jumps and ascosiations that an AI can not.
The biological brain has evolved for eons to the state where we are today.

The only thing that favors AI is that we can make them space born before we ourselves can go out there. So the first true colonizers of space could be AI.

The human brain
http://www.med.harvard.edu/AANLIB/home.html

Zoink · Post by **Zoink** » 2003-01-16 02:46pm

kojikun wrote:tharkun, if the speed of computer advancement is different with different computers (which its likely to be) then youre right, it wont spiral off. Atleast not if humans are developing the technology. But if you took a quantum computer and had it design an even better quantum computer, the result is still the same. faster progression at ever increasing rates.

kojikun, I was simply stating a computer can only design the fastest computer of a known type. The part about intel and gates was simply a personal attack on how they will react to the development to this fastest pentium.

The current silicon computer will reach a plateau. You then need to jump to optical, molecular, or quantum computers. Quantum computers haven't been developed yet. Making the jump will require new tech. A really fast pentium can't design a quantum computer without the information on how to design and build one. We don't have that info, and can't make an assumption that we will have that info. Maybe they are impracticle for an as-of-yet unkown reason. Doing otherwise is to slip into the Trekkie fantasy world were warp engines will happen "just because".

tharkûn · Post by **tharkûn** » 2003-01-16 02:59pm

Atleast not if humans are developing the technology
I don't care who is designing the bloody technology, be it human or computer. The upper bound is physical, not software.

But if you took a quantum computer and had it design an even better quantum computer, the result is still the same.
Not really. The limiting factor in quantum computing is keeping the damn thing isolated. The bigger you make it, the hard it is to act upon the whole works uniformly without some stray photon screwing things to hell and back.

faster progression at ever increasing rates.
Bull. It takes time to manufacture the physical chip. Even with a processor with 10^99999 Hz it can't make the physical system come into being any faster. Eventually you reach a point where there is a rate limiting step that processing can do jack to make faster.

No matter what process you work with you still have these problems:
1. Information moves at c (maximum).
2. There is a finite amount of information that can be stored on a given physical object (electronic switch, spin on an atom, whatever).
3. Information gets corrupted by outside noise.

Between them you end up with a speed limit for computing. Using chip technology we are on track to hit it soon and reach the point where the only way to go faster is to:
1. Do something else (i.e. quantum, non-binary, etc.)
2. Make the computer bigger

Currently we are just making our processors (even our quantum processors) denser. This trend has very real physical limitations.

I remain skeptical that AI will ever live up to these predictions. That a computer will ever be able to design a newer, better computer autonomously.

Arrow · Post by **Arrow** » 2003-01-16 08:48pm

Arrow, youre assuming that the brain and emotions is the result of perfectly identical situations. it would incredibly simple to simulate neurons, the issue would be their ARRANGEMENT. that we cant do yet, but if a computer can literally simulate the things that make your brain work, then they can simulate a brain that is not preprogrammed.

Ok, so now your simulating chemical interactions. Big deal, we can do that today. My point is that we lack the ability to define imagination and instinct well enough to define an AI that is a true self-aware being (we can barely define it well enough for us to understand it). Hell, we can't even define every math problem well enough to be completely computed; in many situations the best we can do is 'good enough.' Division, for instance, has no algorithm, only a heuristic. And that's a simple example.

kojikun · Post by **kojikun** » 2003-01-16 09:17pm

Ok, so now your simulating chemical interactions. Big deal, we can do that today. My point is that we lack the ability to define imagination and instinct well enough to define an AI that is a true self-aware being (we can barely define it well enough for us to understand it). Hell, we can't even define every math problem well enough to be completely computed; in many situations the best we can do is 'good enough.' Division, for instance, has no algorithm, only a heuristic. And that's a simple example.

Nature did not need a definition of imagination. If nature can cause imagination to exist without defining it, then near perfect simulation of nature will also give rise to imagination.

Arrow · Post by **Arrow** » 2003-01-16 10:14pm

Computers only run programs. Programs are algorithms and heuristics. The best you can ever do with a computer is approximate human behavior. You can make it act like has emotions and feelings or work toward a goal, but all its ever going to do is run increasing complex programs. And the more complex a system becomes, the harder it is maintain it.

kojikun · Post by **kojikun** » 2003-01-17 01:03pm

Arrow Mk84 wrote:Computers only run programs. Programs are algorithms and heuristics. The best you can ever do with a computer is approximate human behavior. You can make it act like has emotions and feelings or work toward a goal, but all its ever going to do is run increasing complex programs. And the more complex a system becomes, the harder it is maintain it.

You're thinking incorrectly. Youre assuming an AI program must have a subroutine for this, a subroutine for that. TRUE AI will not be accomplished by saying "this algorithm controls emotions". True AI will be achieved by digitizing the neurons of the brain so that theres no knowledge of whats going on just knowledge of which digineurons are doing what. the patterns that arise from this are not fake, they are not pre programmed, not algorithmic, not predictable, atleast no more then your brain is. These patterns ARE thought and emotion and imagination. The medium is irrelevant, the patterns are what make them real.

Admiral Valdemar · Post by **Admiral Valdemar** » 2003-01-17 01:07pm

People gave up on top-down programming in the eighties, it was then seen that AIs may never be as good as us and many an English professor gloated this fact. Simply put, they tried to write a program which had rukes governing everything, every single possibility was coded for and before long you had programs that were millions of lines long and impossible to debug.

So they instead went for bottom-up programming, letting an AI learn from experiences and get through situations by itself using data input from learning. Distributed management systems are the best form of AI so far e.g. the Seven Dwarfs.

kojikun · Post by **kojikun** » 2003-01-17 01:14pm

Yes. Bottom up is good, but theyre starting at mind not neuron level. What they need to do is write a program that simulates rawr neuron functions. With modern bottom-up AI you can still look at their thoughts, so to speak, because its all numbers in a progream describing base instincts. if they have a program that literally makes a brain only out of numbers in the computer then it will exhibit identical patterns as any given brain. it could even go beyond normal brains thanks to calculation speed of the computer, and the ability to vastly interconnect more so then real brains (because the interconnect would be mere numbers saying "this neuron connects to this neuron")

Arrow · Post by **Arrow** » 2003-01-17 01:17pm

I've been looking at this from a software system engineering perspective (which is top down and the way I'm trained to think). Valdemar, what your talking about is pattern matching and database searching. Its still a long way from anything resembling true AI, but it is a different line of thought from where I was going. So I guess your right in that you could a get a true AI, given enough computation power. Kojikun's way of getting the point across just didn't register.

Korvan · Post by **Korvan** » 2003-01-17 03:13pm

Larry Nivin's "Known Space" has an interesting take on self-aware computers. The computers are only good for about six months. After that, they go insane, retreating into a self-computed fantasy world.

AFAIK, Nivin doesn't go into too much detail on why this happens, but I have a couple of speculations. Imagine your own conscious running at a speed that is thousands of time faster than normal. But your body, and the world around you is still running at normal spped.

There'd be an eternity between each usable sensory input and even if the rate of input was increased to match your thought speed, it'd still take forever for anything "interesting" to happen.

Your starship captain would give you a calculation to work out, and you'd have to spend thousands of years (subjectively) doing some pretty damn repetive math to solve it.

And as for the suppossed "utopia" that will result when machines take over all of our work... Here in Canada, the economic assistance given to our native population by treaty has done more to destroy their culture than nearly anything else. (All in my own opinion, of course).

After saving up a fair bit of dough, I fullfilled a life long dream of taking a year off to do nothing but play computer games. I was showing signs of insanity after only a few months and had to return to work. Work is good.

Zoink · Post by **Zoink** » 2003-01-17 04:26pm

Korvan wrote:Larry Nivin's "Known Space" has an interesting take on self-aware computers. The computers are only good for about six months. After that, they go insane, retreating into a self-computed fantasy world.

I would surmise that these self-aware computers lack the instincts we take for granted. Because of a billion years of evolution, we possess a desire for self-preservation, reproduction, self-betterment, etc.

The self-aware computer hasn't gone through this process. Would such a computer value its own existance? Would a self-aware computer care wether it solved any problems, or dealt with the real world vs simulated reality?.... and could a computer that lacked these instincts comprehend human morality? I think we'd have to hardwire this things into the computer to keep it in check.

His Divine Shadow · Post by **His Divine Shadow** » 2003-01-17 04:50pm

Darth Wong wrote:If computers become more like humans, they will also lose some of the traits which make them useful. A truly sentient machine won't be content to simply sit there forever, slaving away for its masters.

Sure it will, once we've threatened it's life and family

Ofcourse then the nukes will start going off and a T-800 is going to rip you a new one

His Divine Shadow · Post by **His Divine Shadow** » 2003-01-17 04:52pm

Enlightenment wrote:If the singularity optimists are right, in fifty years we might have to deal with the consequences of developing an AI with an IQ of 60,000 that regards humanity in the same light as we view ants.

Oh, joy.

Then we better hope Transcend has perfected his mind transfer device

Admiral Valdemar · Post by **Admiral Valdemar** » 2003-01-17 04:55pm

Totally sentient machines will only be useful as talkative companions or intelligent sex dolls, but then who wants a talking piece of ass when getting it on?

Post by **NecronLord** » 2003-01-17 05:03pm

Admiral Valdemar wrote:Totally sentient machines will only be useful as talkative companions or intelligent sex dolls, but then who wants a talking piece of ass when getting it on?

As long as they're compliant.

As for Prof Warrick. I believe a better phase would be "He's got too much publicity" He comes across as a publicity hound. (and gets free money for it

)

Admiral Valdemar · Post by **Admiral Valdemar** » 2003-01-17 05:05pm

NecronLord wrote:
Admiral Valdemar wrote:Totally sentient machines will only be useful as talkative companions or intelligent sex dolls, but then who wants a talking piece of ass when getting it on?
As long as they're compliant.

As for Prof Warrick. I believe a better phase would be "He's got too much publicity" He's comes across as a publicity hound. (and gets free money for it)

Yeah, but what's the point in having brains if you can't make a little extra cash?

Besides, he and his wife have neural chips allowing one anothers feelings to be felt. Great in sex probably but not so good if one has diarrhoea...

Post by **NecronLord** » 2003-01-17 05:10pm

Admiral Valdemar wrote:
NecronLord wrote:
Admiral Valdemar wrote:Totally sentient machines will only be useful as talkative companions or intelligent sex dolls, but then who wants a talking piece of ass when getting it on?
As long as they're compliant.

As for Prof Warrick. I believe a better phase would be "He's got too much publicity" He's comes across as a publicity hound. (and gets free money for it)
Yeah, but what's the point in having brains if you can't make a little extra cash?

Besides, he and his wife have neural chips allowing one anothers feelings to be felt. Great in sex probably but not so good if one has diarrhoea...

I'm not sure how effective that was. I thought they were in their arms? They must have really freaky wiring then.

Admiral Valdemar · Post by **Admiral Valdemar** » 2003-01-17 05:17pm

NecronLord wrote:
Admiral Valdemar wrote:
NecronLord wrote: As long as they're compliant.

As for Prof Warrick. I believe a better phase would be "He's got too much publicity" He's comes across as a publicity hound. (and gets free money for it)
Yeah, but what's the point in having brains if you can't make a little extra cash?

Besides, he and his wife have neural chips allowing one anothers feelings to be felt. Great in sex probably but not so good if one has diarrhoea...
I'm not sure how effective that was. I thought they were in their arms? They must have really freaky wiring then.

Prof. Warwick had two, one in his arm controlled his house and the ones in his spinal column sensed and transmitted limited sensation to his partner.

It was quite a feat, no real use just yet, but it shows we're moving along with cybernetics.

His Divine Shadow · Post by **His Divine Shadow** » 2003-01-17 05:30pm

Oh yes, on that area there is also a similar method, implanting a sort of device in the brain on the area in the brain responsible for whatever part you're interested in has been shown to work.

This one guy who was paralyzed got one and he learned to controll a robotic hand by thought almost as good as his real hand, ofcourse they where not connected to the actual neruons reponsible so he had to learn all over again, took him like 1 year, but the human brain sure is resiliant and adaptable.

This chip also only interfaced with 5-30 or so neurons, they plan to make another one that interfaces with 300.

Admiral Valdemar · Post by **Admiral Valdemar** » 2003-01-17 05:53pm

His Divine Shadow wrote:Oh yes, on that area there is also a similar method, implanting a sort of device in the brain on the area in the brain responsible for whatever part you're interested in has been shown to work.

This one guy who was paralyzed got one and he learned to controll a robotic hand by thought almost as good as his real hand, ofcourse they where not connected to the actual neruons reponsible so he had to learn all over again, took him like 1 year, but the human brain sure is resiliant and adaptable.

This chip also only interfaced with 5-30 or so neurons, they plan to make another one that interfaces with 300.

Thought they just connected the neurons in the arm to the prosthesis as best they could.

His Divine Shadow · Post by **His Divine Shadow** » 2003-01-18 07:02am

Admiral Valdemar wrote:Thought they just connected the neurons in the arm to the prosthesis as best they could.

There aren't any neurons in the arm that I know of, you must be speaking about nerves, but no, that would have been rather useless in this case as the guy was paralyzed from the neck down.

It was instead directly into the brain.

Arrow · Post by **Arrow** » 2003-01-18 08:01am

Shadow, is this guy you're talking about danish? I've seen something like what your discussing on Discovery once, with a danish guy paralyized from the neck down; really cool stuff.

His Divine Shadow · Post by **His Divine Shadow** » 2003-01-18 08:35am

Arrow Mk84 wrote:Shadow, is this guy you're talking about danish? I've seen something like what your discussing on Discovery once, with a danish guy paralyized from the neck down; really cool stuff.

Doesn't say the name, only that it was a 53 year old man, and the number of neurons was only 5 connected ones..

Eleas · Post by **Eleas** » 2003-01-18 11:47am

His Divine Shadow wrote:
Arrow Mk84 wrote:Shadow, is this guy you're talking about danish? I've seen something like what your discussing on Discovery once, with a danish guy paralyized from the neck down; really cool stuff.
Doesn't say the name, only that it was a 53 year old man, and the number of neurons was only 5 connected ones..

An acquaintance of mine works in a computer lab with neural networks. Those would seem to be our best bet to achieve computerized sentience, if that ever happens.