Thread: What is AI?

IS IT POSSIBLE TO BUILD A MACHINE WHO CAN THINK LIKE A MAN ?

Originally Posted by circuitgram

IS IT POSSIBLE TO BUILD A MACHINE WHO CAN THINK LIKE A MAN ?

no

Actually you can . But then it would not be a machine. It will be a man with metal/carbon nanotube wiring instead of nuerons

Yes, the only thing that separates today's AI and human intelligence is that human intelligence has a much more complicated system which runs many more processes simultaneously, that's the only difference, not free will or any of that mumbo jumbo (please excuse my use of this overly complicated scientific term).

Artificial Intelligence is the intelligence of machines and the branch of computer science that aims to create it. It is an area of computer science that can focus on the behaviors that humans consider intelligent.

Consider that the human brain has several billion cells, each with many thousands of interconnections, and all these cells performing computations simultaneously (that is, by parallel processing).

Scientists understand how these brain cells perform their calculations (enough to simulate a small group of them with a computer by using ANNs – Artificial Neural Networks), but they are not yet clear (as far as I know) on how these cells "learn" or on how they are all interconnected or what their "starting parameters" are (meaning, the various parameters for each brain cell when, for example, a baby is born).

If we knew this for billions of brain cells with their thousands of connections, and if we had a computer that could "think" fast enough in parallel, then we could produce a human within a computer, which would be very interesting indeed along with something of a moral dilemma for some people.

I am certain computers will develop consciousness in the, not too distant, future.
In normal circumstances I would be willing to back my opinion with a decent bet but as it is unlikely to happen, in my lifetime, that would seem to be a fairly pointless act.

Yes it is possible, but it probably cannot be accomplished today or before a certain number of years(depending on how intensive the R&D for such a program would be).


One approach is to simulate neuron interaction, key biological mechanisms and sensory stimulus; the brain simulation would likely not be able to see initially, like a baby, but it would record patterns of sensory input until it could eventually see, and would recognize sound patterns and relate them with other sensory perceptions until it could start talking, like a baby. Potentially a life experience simulation program could be designed to speedup the learning process by uninterrrupted stimulus generation 24/7 for a certain amount of time (so the AI would learn to see, to talk, the words for various things, and read, etc) or have a person's brain scanned so that a certain proportions of learned patterns and memories can be transfered (so the AI would remember things that the person experienced as if it had experienced them).

And consider a computer so big and so fast and so well connected to the outside world that it can "know" and "think" much more than one person ever could, and could do this for much longer than a person could live (and not even get senile). This is the stuff that science fiction is made of.

Thank you For your response.I am a poor boy form a small city of the Nepal and my english is not so good,so please don't mind if there is any mistake in my language.

I have lots of question in my mind.Here is not any technical institute or school where i can learn.So i am learning electronics and programming from internet.

All your answesr help me to understand the actual meaning of AI.
I don't know more right now but i think AI is possible if we will able to built such a technology which can increase it capacity and speed of processing and memory handling in some period interval of time.
Then we can train a simple robot to a advance robot in few years.

I know many engineers are trying to build AI or Give NI to a machine, Even i am not a engineer right now but i will also be one of them in future.

Originally Posted by circuitgram

IS IT POSSIBLE TO BUILD A MACHINE WHO CAN THINK LIKE A MAN ?

I think men would not want to create such a "machine". They do think themselves.

Steve

Originally Posted by Steve Miller

Originally Posted by circuitgram

IS IT POSSIBLE TO BUILD A MACHINE WHO CAN THINK LIKE A MAN ?

I think men would not want to create such a "machine". They do think themselves.

Steve

A little corny don't you think?

Very little modern AI research is done on trying to make a human-level machine brain. Most of it is done on trying to understand how humans solve various problems and why its so different from what machines usually do. The idea is to better understand what "human-level machine brain" would even mean before trying to build one. (Some people think that we've been researching the various sub-problems long enough and that we need to get back to trying to integrate them soon.)

BlueBook,

no why? What for this was good for? The only reason you would get up in the morning to accomplish such a mission will, presumably, make you face the same 'difficulties' that you run in anyway.

But you'll want to get some computer device going so that it moves like a human does. The human like behavior will need lots of unprecedented computing power. It's not going to go any further than this. Lots of your mental and bodily resource are needed only to keep you, on an ever moving earth, in balance, I would say.

The crew on second to latest NASA mission (STS 133 I think ) took an android to space. He was named Robonaut.

http://www.nasa.gov/topics/technolog.../robonaut.html

There are other information available on Wikipedia and other places.

Steve

I don't know much about it but. . .
While I don't buy his theory of consciousness, I thought Penrose raised some interesting points regarding the plausibility of strong AI in The Emperor's New Mind and Shadows of the Mind. You might dig these books. If the former is intelligible and interesting to you I'd then move on to the latter. Although a different kind of book, The Road to Reality deserves mention simply for being epic.

Originally Posted by circuitgram

Thank you For your response.I am a poor boy form a small city of the Nepal and my english is not so good,so please don't mind if there is any mistake in my language.

I have lots of question in my mind.Here is not any technical institute or school where i can learn.So i am learning electronics and programming from internet.

All your answesr help me to understand the actual meaning of AI.
I don't know more right now but i think AI is possible if we will able to built such a technology which can increase it capacity and speed of processing and memory handling in some period interval of time.
Then we can train a simple robot to a advance robot in few years.

I know many engineers are trying to build AI or Give NI to a machine, Even i am not a engineer right now but i will also be one of them in future.

start with
http://en.wikipedia.org/wiki/Logic_programming
http://en.wikipedia.org/wiki/Prolog
http://en.wikipedia.org/wiki/Object-...ed_programming
http://en.wikipedia.org/wiki/Fuzzy_logic

The brain's speed is limited by the physical laws, just like pouring soda in a glass. The soda doesn't need to calculate anything, its motion is purely physical. It just occurs, by the nature of this universe. But simulating such a phenomenon is very complex and requires considerable computation time (indeed, fluid motion is usually measured experimentally when it comes to things such as the flow of air over the wings of an aircraft. Our computational models aren't good enough).

I can't see true AI in any near future, if at all.

Our models for fluid dynamics are fine, it's just that they take days to compute a few seconds of flow, so most of the time, people trade off accuracy for speed and use it as a first approximation.

I think there's a possibility AI's in the same boat, but on a scale we can't really test.

There's a second problem in that a "brain in a jar" AI might never turn out like a human no matter how accurate it is since it doesn't have the same inputs and outputs a human brain has access to. This is a fairly fundamental problem, and one that we don't have much way of testing experimentally either (but notice how sensory defects can effect childhood learning).

i have just started myown AI project not too long ago, my findings as of yet are that you will require at the very least 80-120 billion kb for the base AI program, each kb will be modled after a neuron. the initial projection of the program diagram will consist of a hexagonal shape with 8 courners. each courner will be modled after a fascet of the human psyche as a subroutine with one on top and bottom of the center, top modled after the brain stem for main control and bottom modled after the motor cortex incase you want the AI to have motor control of some kind which is optional. the knowledge base should start very light, modeling infency. the ability to model brain development is key to making a strong AI because the reasoning skills cannot be programed, they must be taught. building an AI on text knowledge alone has been done, but its also considered a Weak AI, because the resoning skills are severly limited. teaching an AI would be a full time job, at the verry least 8 hours a day every day for probably 5-6 years of manual interaction to develop the reasoning skills needed to proform most all tasks aslong as the AI has the required resources and time. at wich point it would be possible to do a main memmory hard copy and mass produce strong AI's.

Originally Posted by Mbush54

the knowledge base should start very light, modeling infency..

This seems very ambitious. A newborn baby has had functioning brain tissue for most of the nine month gestation period. That brain is not sitting idle, but is exposed to a variety of sensory inputs, internal and external. Surely your modelling should begin in that period?

Which facets of the human psyche will you be modelling at the corners?

As a side note, will your ultimate AI be able to spell correctly, or will it be modelled on yourself?

Originally Posted by Mbush54

the ability to model brain development is key to making a strong AI because the reasoning skills cannot be programed, they must be taught.

I read something the other day that shows that babies have good logical reasoning skills from birth. Most people lose them as children and then have to re-learn them.

you will require at the very least 80-120 billion kb for the base AI program, each kb will be modled after a neuron

I assume that is data rather than program? And doesn't that (roughly) translate to what we normally call 80-120 Tb? What sort of hardware architecture are you planning to use to access that sort of data in a reasonable time; I assume it will have to be distributed data and parallel processing?

Originally Posted by John Galt

Originally Posted by Mbush54

the knowledge base should start very light, modeling infency..

This seems very ambitious. A newborn baby has had functioning brain tissue for most of the nine month gestation period. That brain is not sitting idle, but is exposed to a variety of sensory inputs, internal and external. Surely your modelling should begin in that period?

Which facets of the human psyche will you be modelling at the corners?

As a side note, will your ultimate AI be able to spell correctly, or will it be modelled on yourself?

there is no need to model anything prebirth because it has little to no effect on reasoning skills. reasoning and knowledgebase dont start untill the ages of 4-12 months old arguably. but even then, you do not need to model infency in such a litteral sense. for instance, vocabulary and most basic and advanced speech skills can be programmed to shave years off the developement process. speech will be a main subroutine, because it is one of the most important and extensive keys to understanding. asfar as the fascets of human psyche that will be simulated, that is still in the R&D phase. but i can say that it will be designed to combine into a model of human cognitance. also i find it necissary to point out that the model of the brain will physically be designed as fully mature. the area that will model brain development is the growwing knowledge base.

Right. Thank you for that. I was intrigued by this comment: "most humans only use 3-5% of there cognitant(sic) function". This sounds dangerously like the old wife's tale that we only use 10% of our brain. Please assure me you meant something else and explain what that something else is.

Originally Posted by Strange

Originally Posted by Mbush54

the ability to model brain development is key to making a strong AI because the reasoning skills cannot be programed, they must be taught.

I read something the other day that shows that babies have good logical reasoning skills from birth. Most people lose them as children and then have to re-learn them.

you will require at the very least 80-120 billion kb for the base AI program, each kb will be modled after a neuron

I assume that is data rather than program? And doesn't that (roughly) translate to what we normally call 80-120 Tb? What sort of hardware architecture are you planning to use to access that sort of data in a reasonable time; I assume it will have to be distributed data and parallel processing?

well, i would consider every single kb involved in the project as a single program. the subroutines, main memmory,aswell as the main control/interaction program. because all of it would be insepperable and indestinguishable while the program is running. every aspect of the project will be in constant communication with every other aspect. as you can imagine this presents a set of possible hardware problems. the processing power and sheer information fetching will require a very expensive custom machine indeed. at first thought it will require at the very least 3-5 liquid cooled hardlinked quad core towers with 10, 3 terrabyte solid state hard drives each. agian, thats just a preliminary guess.

By intelligence do you mean on a sentient/sapient level of thought? I think machines can certainly "think" in a way, but not with the awareness humans have (not as of today at least). Artificial intelligence is the such made by man (an obvious point)... And not by the natural forces which brought about complex life. And our evolution into humans.

So the point is that we don't have to necessarily use "machines", but if we can replicate the process (or one that is similar) that made us, we can possibly produce intelligent organisms. Though biological, they would still be technically AI because it involved the work of humans.

But of course, androids and superbots are much cooler. And if we can find the secrets to human consciousness, maybe we can create consciousness in fully mechanical bodies... or maybe a mix of steel and biological tissues (like cyborgs).

mbush, I see you have edited out the reference in post #22 to the percentage of the brain which is used by the average person. You have then ignored my question about that point. That is bad forum etiqutte. It make you seem as if you have been caught in a mistake to which you do not wish to admit. Could you please correct that impression by answering my question.

The concepts of AI is C++ or java coding/scripting that the cpu would be able to make on it's own with out human intervention.Correction's on this please ^.^

Todays technology does not have that. I have an idea on how to do it tho but it would use older scripting previously used into a data base,kinda like a search engine does.Keystroke Automated Codeing or something of that sort.

Hrm... I must admit I overlooked the thread.

I've studied a subject on AI. Based on Turing's prove on what is an AI(you know the experiment?) You got an AI when you can't distinguis it form a human. That test could be done on a chat, and due the IQ of the population that wouldn't be very difficult, if you know what a mean...

Basically I think that if we consider the Brain a very sophisticated machine(which from what a read, that's the idea), the first thing we should do is to understand how it works so we can recreate it. And we are far away from that or that it seems, we got something in our head that stills a mystery! How does the frontal lobe work??

Once we had the "NEW BRAIN" we just need it to learn, and I think the only way for it to do that is to have the need. Us humans have the need to live, if we didn't we would stand still(Yeah that's an opinion, but suppose you are immortal, in a life time nothing would have sense, and if you are born that way...). Well, we could also see it from the point that we have the need of reproducing, that would lead to the need to live(That would make you start KB(Knowledge Base)). So any of this would lead the need to learn. And supposedly we would have a new being, an intelligent one(and wiser with the ages).

Anyhow, I think that doesn't add much on this threat, but I wanted to mention the Gödel Escher y Batch book. Is a great book, and basically tries to analyze if our thoughts have anything to do with maths and logic(Which currently is in what AI is based). Read it and you can get your own conclusions, it is really a great book a must read.

My personal though is that currently you can't make an AI, why? Because computers have limitations, as well as the math language(that's Gödel incompleteness theorem). BUT! Maybe Quantum mechanics have the solutions to this limitations...

Hi everyone,
In my point of view,its impossible.Eventhough we tried to make a machine which can think like a man,it cant do whatever a man can do..

I think it's possible, but in good conscience we would not go so far as "like a man" that would be a really screwed up pitiful mechanical man. Stop at something "like a cat" that doesn't break your heart to know it's doomed to dwell in a laboratory cabinet and touch the world with nothing gentler than steel pincers.

Our digital, sequential computers are unsuited to simulating analog, synchronous neural networks. It may be easier to get every person reading and writing numbers in binary.

there are two type of AI... top-bottom, and bottom-top

first is type question-answer, so the program detect question and search the answer in database, where the database create by human/user .... in more big scale... the AI can search in www and fine the most great answer... you can find this type in web...

second is type that more like human, starting from blank... the program start detect question and the answer... again and again.... database created by AI from what it learn and create conclusion, when find new question it'll start find search all releated answer... and create conclusion from that... and store the question and answer. just like how human learn... well there nothing more perfect than human in this planet ^^

AI is very different from neuroscience. It focuses on how we think, not the specific mechanisms. Numenta's focus on neurons rather than AI has been very successful (we kind of know how the cortex works now, kind of.)
Sequence learning is essential to intelligent machines. We learn sequences, which we then combine to form higher level sequences.

Originally Posted by John Galt

Right. Thank you for that. I was intrigued by this comment: "most humans only use 3-5% of there cognitant(sic) function". This sounds dangerously like the old wife's tale that we only use 10% of our brain. Please assure me you meant something else and explain what that something else is.

Humans lose connections between neurons as they grow older. This may be because we start with many random connections, or things to test. We weed out any connections which don't correspond to reality.
We can make an intelligent machine very easily, and we already have (Grok, created by Numenta.) However, it takes millionths of the computing power it would take to simulate the cortex.

is it even ethical to make one? it would take ppls jobs

yes and yes and yes and yes and yes...

i made a robot years ago something simple really it would hoover floors for me the main problem was allowing for easy movement around the enivroment.

I think we're already beginning to see the emergent traits associated with high level of AI in things like Google searches.

It certainly won't be like a man thinks, nor would that necessarily be the best thing--humans carry a lot of evolutionary behavior baggage with us.

not really depends on your viewpoint on it. we are still here. we are pretty resilient. whereas computers are anything but. one large solar flare would cause serious electrical damage. a large enough planetary magnetic flip could wipe the hard drives of a lot of computers maybe delete the internet

Originally Posted by fiveworlds

not really depends on your viewpoint on it. we are still here. we are pretty resilient. whereas computers are anything but. one large solar flare would cause serious electrical damage. a large enough planetary magnetic flip could wipe the hard drives of a lot of computers maybe delete the internet

Delete the internet eh. I hate it when that happens. The worst is when you shift+delete the internet, then you can't get it from the internet trash can. :-)

is it even ethical to make one? it would take ppls jobs

There are some jobs it would be a good idea to not have people do. In science alone there's volcanology and glaciology and cryology among others which could all do with a few non-human operatives taking measurements and making observations. In other occupations like mining and explosives detection and disposal there are several processes and procedures that are pretty unsafe for human biology.

Though it takes human imagination to come up with Intermission: Hauntingly Beautiful, Wildly Low-cost Solution to Clearing Afghan Landmines | Design on GOOD - something both beautiful and practical to deal with mine disposal. Do we think we could come up with a computer that would create that all by itself? My suspicion is that it takes someone who's played lots of simple childhood games or done lots of art + mechanical experimentation.

Originally Posted by fiveworlds

not really depends on your viewpoint on it. we are still here. we are pretty resilient. whereas computers are anything but. one large solar flare would cause serious electrical damage. a large enough planetary magnetic flip could wipe the hard drives of a lot of computers maybe delete the internet

Solar flares won't be a problem sometime in the future.

Originally Posted by fiveworlds

is it even ethical to make one? it would take ppls jobs

Computers already have. That's a big issue with information technology. People are greedy, so they'll use info tech for whatever is profitable, even if it takes people's jobs. Without government intervention or something, AI will be used to take people's jobs.

Originally Posted by circuitgram

IS IT POSSIBLE TO BUILD A MACHINE WHO CAN THINK LIKE A MAN ?

The intelligence part of the brain is much simpler than the basic parts. It is basically one thing repeated many times, whereas the simpler parts are complex wiring designed to give a specific output for a given input (I think.) We've already kinda built the intelligence part, but the rest will take a while to create, if we ever do.