could someone explain ai to me?

[jefhatfield]

i have a friend who works with artificial intelligence (degrees from mit and stanford) and since it's for the military, he can't elaborate on it...i respect that

from what i gather, it does not use data, information, processors, or binary the way that regular computer technology does...he even states that it may be beyond quantum computing...and that ai does not compute or use computers (or it won't when they get there)

apparently, the science is many layers deep and involves computers coding themselves and "thinking" for themselves and that this technology, as he explains it, is 200-250 years away... at the rate they are approaching it

i am not a true science head since i don't understand the arguments on why the universe is a donut shaped finite entity that cannot go on forever and we can travel in time in theory, but never in practice???

so for any of you physics types out there, please, in plain english, explain artificial intelligence to me and why:

1) it is not data/information
2) does not use binary
3) and why, a self-realized robot like yul brenner's mad cowboy, cannot exist anytime soon

[jefhatfield]

come on, five views and no answers...i don't know jack about ai so any info would help..he he

and with 5,000+ members, there must be one physics type here

and also, what is a four way gate in quantum computing?, and how does this "decision-science" mechanism replace binary computing of simple 0s and 1s?

is it a yes-no-maybe-not maybe thing?

or yes-no-maybe if you send me chocolate-i don't give a rip thing? ;)

thanks in advance for any explanations

(my friend's boss, the senior ai guy, says there is no such thing as ai and it is an embarassment to even sling that term

...so many scientists i have met like to challenge traditional thinking to the point where they can't see the forest from the trees and on any given day, they are one step from taking out a schoolyard or sending anthrax letters to politicians and then there is that cal prof, the unabomber)

[Mr. Anderson]

You're asking a ton of questions here jef, more than I can answer as well. I've dabbled in a few of these things, but one thing I do know is that there are several ways to approach ai. And quantum computing and computers is not a simple thing to understand - do you know anything about quantum theory?

Even though in theory a quantum computer is possible, making one is difficult. And then progamming one is even harder. There was an article in CNN I think about a scientist in New Jersey who had written the first program for a quantum computer - only there wasn't any qComputer to run it on. More of an academic exercise.

You might be better off looking on line for some of your answers.

D

[jefhatfield]

Originally posted by dukestreet
You're asking a ton of questions here jef, more than I can answer as well. I've dabbled in a few of these things, but one thing I do know is that there are several ways to approach ai. And quantum computing and computers is not a simple thing to understand - do you know anything about quantum theory?

Even though in theory a quantum computer is possible, making one is difficult. And then progamming one is even harder. There was an article in CNN I think about a scientist in New Jersey who had written the first program for a quantum computer - only there wasn't any qComputer to run it on. More of an academic exercise.

You might be better off looking on line for some of your answers.

D

thanks duke

a lot of what i see online is way over my head like the morpho matrix stuff i have looked at and other things scientific

my background is a ba degree from a b-school and i took a grad class in quant analysis and we studied only the tip of the iceberg on game theory and decision science...the stuff i see on the web that is worthwhile is way beyond what i know about ai

i just see anything a computer processes as binary information...and how can a ai "machine" process something that is not data, information, or be some form of computing?

here is an analogy i heard once in a science class that relates to scientific perception and thinking outside the box:

a boy gets hit by a car and gets sent to an er where the doctor says, " i can't work on this patient...he's my son...send him onto the other hospital"

the boy is sent to the other hospital down the block and when he gets sent to er, the doctor there says, "i can't work on this patient, he's my son"

now how is that?

answer in later post ;)

[Mr. Anderson]

that's a simple one, mother and father doctor team

[jefhatfield]

Originally posted by dukestreet
that's a simple one, mother and father doctor team ah, you got it

a whole college freshman biology class of 70 couldn't figure that one out

and now, with gay adoption and marriage (in hawaii), one could be a husband and the other a husband...or one could be a wife and the other a wife, too

i somehow perceive computers using data, and information, and having the computers process the information/data

it's just that quantum computing, as it relates only to ai, is not binary, information, data, and is not processed...am i missing something very obvious like the two MDs example?

also one of my cliets is an AIDS project, and once i sent an email from one of the computers i was fixing and spelled aids with lower case letters and they corrcted me and said aids is something like a hearing aid and AIDS is the disease

so does ai work (not process) something other than data or information on something that is not a computer?

[Ifeelbloated]

Have you checked with NASA? Unmanned probes and satellites that have to think for themselves? I'd imagine they'd be on the cutting edge.

[jefhatfield]

Originally posted by Ifeelbloated
Have you checked with NASA? Unmanned probes and satellites that have to think for themselves? I'd imagine they'd be on the cutting edge.

thanks, i will check that out but doubt they will give me a layman's course on the inner working of ai...but i will check...thanks:D

i went to whitehouse.com and they didn't really tell me much about GW Bush, but they did give me an insight into the inner workings of bill clinton...he he

[wake up Jobs!!!]

I know nothing , I swear!!!!!!:eek:

-GaBe-

[jefhatfield]

Originally posted by wake up Jobs!!!
I know nothing , I swear!!!!!!:eek:

-GaBe-

but you are a tele player like me;)

what? asks the rest of macrumors:rolleyes:

i had '65 maple neck, '68 maple w/bigsby, '78 rosewood, '90 rosewood, '90s squier maple, and Fernandez tele maple copy of butterscotch blond

[wake up Jobs!!!]

I got a new Mexican Standard Telecaster with sunburst paintjob, Im planing on switching the lipstick pickup for a seamour duncan hotrails.
Rock on!

-GaBE-

[Taft]

Originally posted by jefhatfield

1) it is not data/information
2) does not use binary
3) and why, a self-realized robot like yul brenner's mad cowboy, cannot exist anytime soon

From my understanding of AI (which consists mainly of reading articles, Godel Escher Bach: The Eternal Golden Braid, and an undergrad course), the three points you make are not necessarily true.

To define AI you must first define what intelligence is (I'm pulling this from the material in Godel, Esher, Bach--its an amazing book for those of you with the patience to read it). Is intelligence simply the ability to compute? Is it the ability to draw conclusions from seemingly unrelated facts and material? Is it encapsulated in the ability to observe patterns in complex datasets (the patterns in a quilt or generally in the world around us)?? It is the ability to learn based on previous experience?

Once you've got a good definition of intelligence, you'll see why there is a huge gap between current computer intelligence and our intelligence. Now moving onto the question of implementation...the view I've formed of this is that there are tons of ways to go about creating an artificial intelligence, but the holy grail is far from being acheived.

The majority of AI systems currently out there use a system of programmed rules in combination with some ability to adjust their bahavior based on previous user interaction. This can be seen in many video games these days. The logic of how "bots" operate in games is preprogrammed, but the behavior changes based on actions the user makes. Black and White is a good example of this (you actually try to teach your creatures in the game). Also, systems like IBM's Via Voice, use this. As you speak to the system it "learns" how to read your voice.

Now these AI systems are pretty rudementary, but they display the level of today's practical application of AI. Cutting edge research is hoping to take us much farther. But for now we are limited to our current technology and our current understanding of AI, making a magical thinking robot unlikely for quite some time.

There are many viewpoints as to the true way towards artificial intelligence. One is that behavior is preprogrammed, but the system is able to modify its own code to alter its behavior in future situations. Thus learning is really just modifying the systems programming. Another viewpoint is that all behavior and actions of a system can be dictated by data storage and the appropriate links between that data. Thus the system reacts and acts according to what its data and links tell it to do. Prolog was designed with a principal similar to this in mind.

Personally, I think that defining what intelligence is and being able to describe it, is the answer to how AI can be acheived. I think that mathmeticians and logicians will play as big a role as computer scientists in acheiving AI. The current methods of logic in our computers are far too constrained to allow for true AI. Also, figuring out how we think might shed some light into how to program machines to think. Our brain is still in many ways a mystery.

A final thought...I increasingly believe that AI may very well just be stumbled upon. One day, a system might just start acting in ways unpredicted by code. When that happens the system may be deemed to have acheived intelligence independent of the code that was written for the system.

I find this topic very interesting, and there is a lot of great material out there on the subject.

Taft

[wake up Jobs!!!]

I know somthing about AI. The way it works is when a robot or somthing that posses AI comes incontact with any object or situation, it records it to a hard disk presumibly and when it incounters a similar situation , it looks through its database to find what to do in that situation on in that incounter. I found that in a Popular science Magazine article a while ago.

-GaBe-

[G4scott]

Alice Bot (http://www.alicebot.org)

This link may prove useful... What's cool, is that you can talk to a computer. It's hard to have an intelligent conversation, and most of it is just programming the computer to respond in certain ways, but it's still interesting.

[Durandal7]

I can't help you with 1 but I may know a bit about 2 and 3. I have studied various advanced technologies and AI theroies (both as requirements and hobbies) and I may have a bit of insight. I'm no expert so the part about no data/info blows me away.

As for not using binary, there is a type of experimental software/hardware known as fuzzy logic. Since the black and white world of binary does not truly represent the world or the way we think fuzzy logic may be used instead. Fuzzy logic is basically interpreting data with variables between 1 and 0 and having native hardware that understands it without translating it to digital.

The AI being a long ways off is because of the nature of our brain. Our brain interprets billions of pieces of input at any given time and there is simply no hardware out there that can emulate that.

I'm probably wrong on both these counts, after all I'm just a student and know nothing of what the Military Industrial Complex is up to but I hope this helps.

[rainman::|:|]

hey all, i'm not an expert (obviously) but the field interests me. allow me to make a couple of corrections as i know it... first off, AI does not exist. not yet. Learning software isn't AI, it simply follows it's programming and "learns" what it's programmed to. Same for the space probes, they don't "think" any more than your computer does-- again, they simply follow their programming, even if that means making some decisions... they can only do that when software allows, they know the outcome of both decisions, they just pick the appropriate course of action based on the data at hand, and their programming. Basically an IF loop. For a long time, AI was thought of in terms of the "fuzzy switch", basically a transistor that's got a third position, maybe. i think the trend has gone away from that, but that's what scientists thought differenciated between computers and people's brains. Personally, i find that this definition of AI serves better: The ability to expand beyond the programming at will. 'Course, that raises the question of what will really is. a question for philosophers, no doubt. Anyway, that's about the extent of my knowledge...

:)
pnw

[rainman::|:|]

Originally posted by asurace
As for not using binary, there is a type of experimental software/hardware known as fuzzy logic. Since the black and white world of binary does not truly represent the world or the way we think fuzzy logic may be used instead. Fuzzy logic is basically interpreting data with variables between 1 and 0 and having native hardware that understands it without translating it to digital.

Ah, you beat me to it! damn you...

hehe
:)
pnw

[zimv20]

there are, of course, many different views on what AI is and what it can do.

in my mind, there are three popular and distinct views.

1. AI is the ability of a device (likely computer hw/sw) to mimic humans and the way they draw conclusions and make decisions. c.f. alan turing and the litmus test he designed for achieving this (basically, if a human can't tell the difference between human responses and computer responses, then you've got something).

this technology is available today in expert systems, neural nets, etc.

2. AI is the ability of a system/device to modify its own "logic center" to suit its own purposes. just as our brains can conceive of something w/o specific external prodding (i.e. creativity), such systems could grow w/o human's fiddling with it. and, here's the thing, grow in a way it wasn't necessarily designed to do.

i've not studied AI since the late 80s, so i can't speak to current progress in this view.

3. AI is the ability of a system/device to be sentient. (how one tests that i'll leave to the philosophers). we're obviously nowhere near this.

jefhatfield, when you ask your questions, it makes sense to interpret the answers in terms of those views. your friend may be working on AI #3, where maybe you're thinking in terms of AI #2.

that said, what if we asked those questions about our own brains? (the ultimate AI machine)

1. it is not data/information

well, it's not, but it's a mass of cells that stores and processes that. i think that would always be true.

2. does not use binary

hm. our own neurons act as binary devices -- they're either idling or firing. the specifics of how they fire may not, to some people, be binary.

3. yul brenner robot

there are arguments that to achieve AI #3, we're basically becoming god -- creating conscious life from, for the sake of argument, non-animal matter. are we close to that? we're much better at blowing each other up. (oops, got political there :-)

long, but does that help?

[G4scott]

http://crab.rutgers.edu/~goertzel/AIsites.htm

This site might also be useful. You can find tons of information about AI on the net, but some of it is a little out dated, and from the realms of science fiction...

[mischief]

Quantum particles exist in 32 independant states. The idea being that if each particle represents a digit, that digit could be any value between 1 and 32 with heisenberg adding a zero state just to make things fun.

AI in any real sense adds:

"Yes, and"

"Sort of"

and qualified logic statements like " Only when I feel like it"

to traditional logic theory. The whole reason AI is often considered impossible is that math-based AI when used in an essentially mechanical computer lacks a number of basic requirements to "thinking" like the instinctual need for self preservation, procreation and conservation of energy.

[Gelfin]

Artificial Intelligence is a deceptively compelling term. The current state of the art in AI is a lot more boring and mundane than making a machine that is self-aware.

AI, in practice, means building a system which simulates complex human decision-making processes. Generally what this means is programming a large database of patterns. The machine then compares its inputs to those patterns and executes code associated with the patterns that match the inputs. In a sufficiently complex system this can give the illusion of creativity because the system's designers can't easily predict how the machine will react to certain inputs. But it is possible to design the system such that it gives results similar to a human expert in the areas covered by the machine's database. As a result these are called "expert systems." A textbook application of this sort of technology is medical diagnosis.

Fuzzy logic is a refinement of expert systems which became something of a popular buzzword several years ago. Essentially, instead of directly matching inputs to patterns in a binary (match/no-match) sense, the system gives a rating to how well a pattern matches an input. Imagine a machine designed to process temperature readings, with a pattern that fires if a given reading "is hot." A traditional pattern recognition system might have to define a temperature threshold for the "is hot" pattern. A fuzzy logic system, however, would be able to define "is hot" relatively, so that a hundred degrees farenheit would be considered hot if you were taking room temperatures, but not hot if you were taking temperatures in an engine compartment. This allows a machine to generate even more seemingly clever deductions.

Then there are neural networks, which are more theoretically interesting, but generally less practically useful than expert systems. A neural network, as typically implemented, is essentially a big matrix of initially mostly random numbers, coupled with a set of rules that define, first, how the system should produce outputs based on the current state of the matrix, and second, how the system should modify the state of the matrix based on the inputs. The result of this is a system that must be "trained" instead of "programmed." A human looking at the matrix cannot determine how the numbers relate to any information the neural network contains, but the system can be trained to produce consistent output relative to input.

The problem with all these systems, increasing at the neural network end, is that while the machines can produce surprisingly good output in many cases, they do not posess the human capacity for introspection. They can give you conclusions, but they can't tell you HOW they arrived at a given conclusion. This makes some of their good answers mysterious, and more importantly, makes some of their bad answers confounding, and often very difficult to fix. With neural networks in particular, you can't be entirely sure that they're "learning" what you think they're "learning." My AI professor in college had worked on some military projects, and was able to tell me about one where they were training a system to recognize various tanks and the like from photographs. The army took dozens of snapshots of military hardware, which were fed into the neural network. The output results seemed off, and they finally figured out why: The photos of the hardware had all been taken on the same day, in roughly the same area. As a result, when they thought they were training the network to recognize tanks, they were actually training it to recognize sunny days.

That's the kind of thing you set yourself up for when you're doing AI. It seems really neat, and the prospect of building a machine that can hold a conversation with you is the sort of thing that draws people into it, but once you get into it you realize there's no magic. It's just a different way of telling the same dumb machine what to do, and in some cases it's a much more difficult and frustrating way, especially when it occurs to you just how far off that computer you can talk to really is.

[Taft]

Originally posted by paulwhannel
hey all, i'm not an expert (obviously) but the field interests me. allow me to make a couple of corrections as i know it... first off, AI does not exist. not yet. Learning software isn't AI, it simply follows it's programming and "learns" what it's programmed to. Same for the space probes, they don't "think" any more than your computer does-- again, they simply follow their programming, even if that means making some decisions... they can only do that when software allows, they know the outcome of both decisions, they just pick the appropriate course of action based on the data at hand, and their programming.
...



It all depends on your definition of AI and intelligence. You've defined it to be the ability for machines to think like humans. Most define it more loosely as the ability for machines to behave intelligently.

I completely agree that true intelligence has not been acheived by any machine, but that doesn't mean that AI isn't being currently employed in current software at some level.

The programs out there right now that have the ability to adjust themselves have aquired that ability through AI research and ideas. They are representitive of the current state of AI in mainstream/commercial computing. These programs provide us the best examples of AI technology being applied today.

Just because a computer or program doesn't have true intelligence doesn't mean its incapable of some intelligence. AI principles are being applied to more and more programs as the ideas of researchers disseminate throughout commercial programming.

Taft

[Joshlew]

As I understand it, the gateways (paths representing yes(0) or no(1)) in a quantum computer, can be either 1 or 0 not just one of the two. I'm not sure
how the computer decides which one though...........:confused:

Pick up a copy of the latest issue of 'Scientific American' to find out more.

[jefhatfield]

thanks for all the input...it's much better explained than what i have seen from the net in a general search

the "fuzzy" thing does go quite a ways to explaining how terms like data, information, processing, binary, and computing are very strictly not in the vocabulary of AI, or AI #3 (i take it those terms are for the binary, normal computing devices)

i know that nature is not binary, 0s and 1s, thanks paul w for your excellent explanation, and the doctor (neurosugeon friend of mine) says that the brain is not binary...therefore, we can't just attach a binary device like a flash card to one's brain, like in sci fi, and do this instead of schooling...though it would be nice

i know there is the yes, no, and maybe, but i was wondering also (last question) what the fourth "switch" was as it relates to ai and quantum computing...but i will look at your resources and hyperlinks

thanks again everybody:p :D :D

[jefhatfield]

Originally posted by wake up Jobs!!!
I got a new Mexican Standard Telecaster with sunburst paintjob, Im planing on switching the lipstick pickup for a seamour duncan hotrails.
Rock on!

-GaBE-

wow, i didn't see your post

ok, off the subject of ai and into guitars,,,for a second

i had a duncan quarter pounder set in my 1978 tele and both pickups were good for lead and rhythm

the 1965 tele was only good for rhythm and had no balls for lead (same for the 1990s squier tele...too thin)

but the fernandes tele and the 1968 tele with stock pickups and bigsby had decent output for both rhythm and lead

the 1990 american standard tele was good for both but the guitar sounded really muddy and nothing like what i wanted out of a fender...at least the 1978 tele with duncans was designed not to sound like a tele but fender giving me a new 1990 telecaster stock and having it sound like a heavy metal guitar put me off

right now, i have only one guitar, and it's not a tele and not even a fender...it's an esp/ltd viper 301..kind of like a gibson or epiphone SG and priced in the same range as the high end epiphone SG/G-400 and i paid $519 US for my viper guitar

if i was more into playing, i would ditch the viper and get a telecaster and my dream guitar would be the ronnie wood signature series telecaster by esp but the $1400 price tag and waiting forever to even get it deters me...i would much rather buy a mid range ibook for nearly the same money;)

[jefhatfield]

Originally posted by Taft


From my understanding of AI (which consists mainly of reading articles, Godel Escher Bach: The Eternal Golden Braid, and an undergrad course), the three points you make are not necessarily true.


Taft

information, data, and processing are related, by our industry to binary, which means 0 or 1

ai is not that simple and i guess deserves to be in its own field and in no way a subset of computers...as we know them

but the pissed off, egotistical, out of control, robot like yul brenner must be way, way off in the future...right? otherwise, we are in big trouble

i don't know if my friend who works in ai will tell me much more because he hates his job and he spent 12 years studying the topic because he had an abusive father who forced him through mit and stanford and when he came down with a terminal/chronic illness while in school, his dad called him lazy though he has four degrees in the study of ai, three of them advanced degrees

in some way, i feel sorry for him, and he does remind me of that other professor with that cabin in montana

i believe there is a fine line, or no line, between genius and madness...and high iq has been statistically linked to mental illness...has anybody ever heard of a dumb schizophrenic?

and then there was "a beautiful mind"...it seems the best scientists i meet and read about are nut cases:p

[Gelfin]

Originally posted by jefhatfield
...has anybody ever heard of a dumb schizophrenic?

Ever ridden public transit in San Francisco?

Schizophrenia is not absolutely correlated to intelligence. Smart people DO seem to have more mental health problems, in particular depression and anxiety, but have you considered that maybe that's at least in part a result of the way we treat smart people in our society? Adult Americans overtly respect intelligent people, but privately mistrust or resent them. Young Americans don't even worry about that dichotomy. They just kick the smart kids' asses as often as possible, and abuse them psychologically when it's not possible to abuse them physically.

Couple that with the way smart people tend to see through the comforting oversimplifications of life under which most people operate and of course you're going to get smart people with "issues."

More on topic, though, is the notion that the more intelligent our machines are, the more we're going to have to maintain them with psychologists rather than engineers. A neurotic computer would be more indicative of progress in AI than a flawlessly reasonable one. What I'm saying is, the first thing you'd want to do with your fledgling machine intelligence WOULDN'T be to plug it into a planetary defense system. Duh.

[jefhatfield]

Originally posted by Gelfin


Ever ridden public transit in San Francisco?



i can't say i have but i went to school in the SOMA which is full of homeless, drug dealers, hookers, etc but not as bad as neighboring tenderloin district

i was also a social worker in 90-92 and also since last year, working with heroin addicts, as a volunteer

...and so far, i have not met one, dumb or unintelligent schizophrenic

there is no doubt that smart people have more problems with mental illness and i don't think that the correlation is psychological with smart nerds getting picked on and such...so much

but smarter people have shorter synapses in the brain than the average person, but when those synapses get too short, it enters into the realm of schizoprenia...some people who have short path synapses who are schizophrenic are indeed stupid, but i am mentioning a well documented psychological phenomena

where would we be without game theory as it relates to economics..pay thanks to dr. nash, a brilliant schizophrenic mathematician

and where would the art world be without vincent van gogh, a schizophrenic? only picasso is more well known in the art world

and syd barrett, the original founder and leader of pink floyd...also a schizophrenic and inspiration for the movie, "the wall"

and don't forget brian wilson of the beach boys, whose early work was lennon and mccartney's main inspiration for "sgt. pepper's lonely hearts club band" according to paul mccartney in an interview

genius definitely has a major correlation with insanity, specifically schizophrenia

but i am not glorifying schizophrenia...i suspect my friend who is the brilliant ai scientist has some major psychological problems but he is extremely secretive about his sometimes strange behavior and lack of basic hygiene at times

[Gus]

that in order for a computer to make decisions in the manner that a human would, it would have to have sensory equipment similar to a human; otherwise it will always be dependent upon a human feeding it information. A machine cannot become self-aware or dependent without those things being in place first.

I thought about saying that a new kind of language other than binary would have to be developed in order for a machine to make decisions like a human, but this would be false, I think. Every decision we make is a yes/no answer: the only difference is that we reason before drawing the conclusion of yes/no.

The real difference would come to non-subjective decisions. For example: Does that taste good? Is that music good? Is is too looud? What is your favorite food, etc. Of course, with the proper sensory information, I guess that a machine could "grow" or learn which preference they like.

Anyway, I don't know anything scientific about AI, this is just be blabbing.

Gus

[rainman::|:|]

Originally posted by Gelfin
Smart people DO seem to have more mental health problems, in particular depression and anxiety, but have you considered that maybe that's at least in part a result of the way we treat smart people in our society? Adult Americans overtly respect intelligent people, but privately mistrust or resent them. Young Americans don't even worry about that dichotomy. They just kick the smart kids' asses as often as possible, and abuse them psychologically when it's not possible to abuse them physically.

Straying further from the path of AI, i can see where you're coming from, but i must disagree. i think you're quite correct in that depression and anxiety increase with intelligence, i dislike getting this personal, but i've suffered from both for quite some time. i won't say what my IQ is, but it's up there, if you believe in that measurement, though i can't say i do. i'm afflicted with a rather severe social phobia (social anxiety disorder) that leaves me little more than a complete recluse, so i have a lot of time to think the matter over. I think that, at least in the case of anxiety disorders, they occur in highly-intelligent people because of heightened-awareness, sensory overload of the brain, if you will. i don't think nature/god/whatever intended for human brains to be burdened with as much as it is today, and i think this is the result. Even in more enlightened cultures in history, depression, anxiety, and mental illness have never been rampant as today. Interesting historical sidenote, if anyone is interested, read about Abraham Lincoln's melancholia, it's amazing the man lived to do his great deeds, he should have been dead from mercury poisoning (treatments) long before.

Mayhap someday we'll be busy treating AIs for these same illnesses :eek:

:)
pnw

[iwantanewmac]

Originally posted by jefhatfield


but you are a tele player like me;)

what? asks the rest of macrumors:rolleyes:

i had '65 maple neck, '68 maple w/bigsby, '78 rosewood, '90 rosewood, '90s squier maple, and Fernandez tele maple copy of butterscotch blond

That's a nice collection :)

[Nipsy]

AI is in the simplest terms and the easisest definitions, the ability of inorganic objects to learn.

Simple AI consists of rulesets:
OSX conatins an object callled a rulebook server.

This server contains rules, coded by humans, which the OS uses to interact with the computer, and you, the user.

In very simple AI, the computer would code its own rules, in order to preserve itself.

Say you try to do something and have a kernel protection failure. The computer would write that to a log, along with circumstances surrounding the failure.

Say it happened twice more, and the rulebook server says, that after three failures, it will not allow the process that caused the failure to occur, by writing a new rule to the rulebook.

Now if you try the behavior which causes the failure, the computer's rulebook server says, nope, that's dangerous, let's find another way to do the task.

The computer has now created a rule to avoid the task which has caused it to fail in a repeatable fashion, and found a work around.

It has learned.

The problem with AI now, is that the overhead is huge (everything must be logged, and every trio of problems becomes a rule, and every rule must be checked against evey other rule, etc.) only to perform the simplest task of self preservation.

The other problem is that programming is logical, and intelligence is rational. The human computer processes and intakes vast amounts of information, while maintaing all its own subsystems, and making rational decisions. It maintains information it determines important, and then releases it when it is no longer deemed necessary. It processes petabytes of visual input in realtime. It is self monitoring and self operating and self regulating and self repairing. Nowadays, the tend not to crash beyond repair for 60 odd years, and further more, they are, for the most part, self updating.

We don't process a complex set of yes or no variables to come out with a decision, we make decisions by instantaneously consulting the several relational dbs that are the brain, memory, emotion, involutary response programming, etc, at a mind boggling rate.

The computer on the other hand makes computations based on its limited instruction sets, and does it millions of times per second. The logical yes or no is at the heart of every computer's life, and it can make millions on yes or no (0 or 1) decisions per second.

The reason your friend probably talked away from chips and binary results is that binary decisions don't lend themselves to being rational. The only way a computer can make rational decision, is by making many binary decisions, while consulting vast rdbs of rulesets. Even super-computers of the highest order are still slow at this sort of "thinking".

As we gain processing power, we can make more complex rulesets, and larger dbs, and even let the computer decide its own rules, but binary processing still hasn't reached the point where the computer can process [your mother tongue], and respond in kind (whether typed or spoken).

I still don't beleive anyone who thinks AI (in the sense that we envision it [HAL 9000]) is more than 50 years away. 50 years ago, the most credible industry pros predicted a worldwide need for no more than 5 computers.

The real scary part is when AI overtakes human intellect, and I'm curious as to whether that will happen in my lifetime (60 more years if I'm really really lucky).

[jefhatfield]

Originally posted by Nipsy

The real scary part is when AI overtakes human intellect, and I'm curious as to whether that will happen in my lifetime (60 more years if I'm really really lucky).

my professor friend admits his form of computer science is not engineering in the traditional sense but a sort of science that is not accepted by his fellow scientists in biology, mechanical engineering, civil engineering, electronic engineering, etc

the closest thing to what he can be labeled at is a software engineer but he makes the point that using the two terms together makes the "real" engineers very upset since many software gurus don't know the first thing about basic math or engineering

several decades ago, he got his undergrad degree in electronic engineering so he understands the traditionalists and their objection to computer programmers calling themselves engineers

on top of the software engineers, there are now the network engineers, of which 60 percent of the microsoft ones in the field don't even possess a two year degree

and to boot, they don't usually program anything and get paid more money than any type of engineer out there...but this is due to the fact that networks are extremely crucial for business and the inroads microsoft has made into the business world in general

i respect traditional engineers very much but the world also needs the psuedo engineers like programmers, network engineers, and sales engineers

one of the best paid sales engineers is the cisco certified internetworking expert...a designation given to a technically inclined router salesperson who are rewarded a ten percent discount off of wholesale

the CCIE is not a degree, and only has been recognized as a technical certification for network engineers, but it is a sales engineer title that has to be refreshed and retested anytime cisco feels the need

the average CCIE makes 120k a year starting, and these people who embarked on what was a retail sales technical training program have passed a set of tests very few phds in computer science could ever pass...in silicon valley, phds from ivies or MIT are a dime a dozen, but of the millions of people who study cisco technologies, only six thousand or so people in the world have been able to pass the CCIE exam designed for technical sales people

wow...i veered off track...but the unpredictability of the high tech field and the earning pyramid and conventions of taditional education being turned upside down, we may find the answer to AI

[jefhatfield]

Originally posted by jefhatfield


wow...i veered off track...but the unpredictability of the high tech field and the earning pyramid and conventions of taditional education being turned upside down, we may find the answer to AI

super long post, i know...he he

but look how fast things progressed once man discovered agriculture ten thousand years ago...for many tens of thousands of years before that, nothing much changed and basically stayed the same

look what happened once "zero" was used in the numbering system

look what happened to the world of computers as it related to an affordable consumer product with the advent of the apple II

maybe there could be some breakthrough in the not yet discovered field of ai to make ai a reality "worthy" of a label

a label?

like in the current san francisco 49ers with jeff garcia as quarterback, with records and good stats he put down...who could be superbowl champs (but the are IN THE MAKING - rebuilding)...but they have not won a superbowl with garcia as quarterback

when they do, then the 49ers with garcia can be superbowl champs instead of potential superbowl champs

ai could be ai instead of a potential field "in the making"