|
Inverse Icarus
|
 |
flying too low to the ground
May 2001
time: 00:25
|
|
*note: i am defining sentience as the ability to reason, show cognition, and demonstrate higher-level intelligence*
machines will rise to power eventually if AI is done the way i propose it to be, but not like in T2.
for the most obvious anti-t2 reasoning, the human form is not the best form to be. we are rather slow moving, large targets, etc.
if you wanted to make a military-use robot, you'd make bee-like semi-sentient machines with posion in them. then release several swarms and control them from miles away.
i could also see machines much like spiders evolving (ever see sg1, the replicators?) that would work well.
i am currently studying cognative science at one of the leading universities in america, and let me tell you, machines will be tought to think on their own. they wont be bound to "whats programmed into them" as Kontiki said.
the study of the mind is rather young when compared to physics / chemistry / biology. when more is mapped out it will be just a matter of converting it to code.
as it stands right now, one of the most widely held hypothesises as that all human thought is dependent on schemas, cognative frameworks that we use to sort anything and everything. when you take a simple schema and draw it on paper, it resembles a comp-sci style class, easily. all you'd need to make a robotic schema intelligence base is a very advanced program that would allow the sentient ai to dynamically alter/create new classes.
one of the lesser understood parts of the paradox is the neural networking. sure, we would have a program that can take in information and plop it into a schema, but how does thought flow from one shcema to another? there has to be linking of some sort, which isn't hard to code in theory, but when you ponder everything you have a schema for, each schema would have pointers pointing to possibly hundreds of thousands of other schemas. my group is in the process of proposing a design relating to this, where each schema had links to multiple schemas, but they are organized into "teirs" depending on how often they correspond to the related schema. for example, when you think of mountain dew code red, you may think of cherries, but you can also think of the alumunium can, or even the ingredients in it. according to use, each of those things is linked.
the major problem with such a system is that it is utterly impossible to map by hand. there are far too many things to comprehend and sketch out. another problem that arises is the whole sorting thing. most computer programs use several ways to organize data, linearly (arrays, vectors, lists, etc), or tree (binary tree, heap). with the proposed neural netoworking, we'd end up with a "neural net" for lack of a better name, with little or no ordering. we'd have to maintiain a tree seperate from the neural network, but pointing to the nodes in it. therefore we could search the tree for something, and then jump to it in the neural network. then we can explore the teirs of links until we have whatever it is we're looking for exactly.
another thing that pisses me off, and i wrote an essay on this one, is that the connotation of sentience is human-based, with good reason of course. for millena we have been the only sentient thing to exist, so everything we think about sentience comes back to "but is it human?". this is wrong, in my opinon, but this is the way it is, sadly. say you made a completely sentient ai, but all it did was sit in AOL Instant messanger and chat. say it could preform mathetical operations and problem solve. you could give it word problems and it would think about them and solve them. simple text input/output. no other senses / output methods. the general populus may not consider that sentience. it's too un-human. so, to make a humanistically sentient being you'd have to give it several human senses. the major ones are sight and hearing, feeling comming into a secondary position. taste and smell are useless for the purposes of research as of yet. developing a vision system on par with that of a human is a task in and of itself though. forget cognition and reasoning for a second and think of what it takes to make a vision system that can take an image from 2 cameras (eyes), convert it into a 3 dimensional image, and then seperate the pieces and identify objects. how do you know that can of mountain dew is a can of mountain dew, and not just a green cylinader? i do not specialize in vision systems so that is a crappy argument, but you get the gist. in order to make a sentient being in the eyes of the public you would have to give it the power to communicate with humans on a personal level.
about the rising to power points. they will if they are given the chance. and i dont mean the chance to revolt violently. i mean if the people building the robots are ethical, and when true sentience develops, the robots are treated as people. there is a theory in cognative science that states that the purpose of biological life is to give rise to mechanical life.
when you look at it, machines are better than us.
heres an essay i wrote for the class:
The End of our Evolution and the Rise of Machines
A Paper Based on Conjecture from “Can Animals and Machines be Persons?”
“Life” has been evolving since the dawn of Earth, and has no intention of letting up any time soon. Evolution of a biological species takes millions of years, and is the product of specific gene mutations being passed on to next generations. It’s a slow moving process at best, and it’s a statistician’s nightmare, a roll of the die sets the whole thing to chaos. Mechanical evolution, however, can be planned, organized, evaluated, and perhaps more importantly, can be done extremely rapidly. Look at how far computers have come since their conception, only a few decades ago, it’s simply amazing. In a new world, simplicity will begat complexity, chaos will begat order, and biological will begat mechanical.
One of the biggest arguments against artificial intelligence is that its current incantation is essentially hard coded rules in which a machine is being told how to emulate human thought, and essentially, that is true. Today’s code, both in games and research areas, is a nothing but emulation, at best. However, a lot of research has been done into the areas of artificial intelligence, self-evolving code, and robotics recently, and when these areas advance sufficiently enough to meet up, mechanical life will begin.
The only way to develop a truly sentient artificial intelligence, in my opinion, is to not develop it at all. The criticisms based on underlying, hard-coded, humanistic tendencies for artificial intelligence must be addressed, and the artificial intelligence must be allowed to evolve on its own. It can be supplied with the basic structures for learning, such as schemas, and the basic foundations of cognition and data organization. Then, the agent should be run, and left alone to evolve its own intelligence, besides, recreating human intelligence for a non-human host seems ignorant, and honestly, boring. Imagine what could happen regarding artificial agents’ emotions if they weren’t provided any ideas about them. Would they evolve their own emotions? Would they be similar to ours? I wonder what intelligent agents “grown” in isolation would do, if we created “societies” of artificial agents in an isolated environment with no knowledge of humanity, where we could supply them with enough raw materials to do their bidding. Would the original agents look to create other ones, to reproduce, or would they simply look to expand their own functionality? The question comes down to this, what happens to cognition when the biological drives of life are taken away?
i'm hoping to one day be a major player in the field of cognative science, and i await the day when my creations destroy you all 
|
15-01-2003 04:53
nice Gatekeeper 
|