Will computers become so advanced that they can solve any problem?

Will there come a time in the not so distant future, when we as a people can ask the computer what the cure for cancer is, and the computer can then formulate a cure?

What is love?

What is love?

Please baby don't hurt me

TIME Magazine actually kinda talked about this issue. The rate at which data can be processed is doubling every 18 months. This means that data growth is exponentially increasing. At a certain point, there is going to be an explosion where processing speeds are increasing blindingly fast.

Estimated at around 2025, one computer will have the processing power of the human brain. By ~2050 one will have the processing power of all the human brains on Earth. At this point, we wouldn't even be having to think anymore. It would be computers being their own computer scientists, figuring out how to make advancements in the field even faster, and faster.

Just because a computer has the processing power of a human brain, doesn't mean it is one.

TIME Magazine actually kinda talked about this issue. The rate at which data can be processed is doubling every 18 months. This means that data growth is exponentially increasing. At a certain point, there is going to be an explosion where processing speeds are increasing blindingly fast.

Estimated at around 2025, one computer will have the processing power of the human brain. By ~2050 one will have the processing power of all the human brains on Earth. At this point, we wouldn't even be having to think anymore. It would be computers being their own computer scientists, figuring out how to make advancements in the field even faster, and faster.

Processing power does not equate problem solving though.

Sounds like the Multivac to me.

Time to post Asimov's 'The Last Question' again

http://www.multivax.com/last_question.html

Quote from: EnglshGentleman on March 21, 2011, 11:19:36 PM

TIME Magazine actually kinda talked about this issue. The rate at which data can be processed is doubling every 18 months. This means that data growth is exponentially increasing. At a certain point, there is going to be an explosion where processing speeds are increasing blindingly fast.

Estimated at around 2025, one computer will have the processing power of the human brain. By ~2050 one will have the processing power of all the human brains on Earth. At this point, we wouldn't even be having to think anymore. It would be computers being their own computer scientists, figuring out how to make advancements in the field even faster, and faster.

Processing power does not equate problem solving though.

How doesn't it? Modern AI use algorithms for problem solving and reasoning. The more complex the puzzle, the more time and memory a computer requires to solve it. Right now that is a problem because complex things take monumental amounts of both of them, but if in the future processing speeds are blindingly fast, then this would not be an issue.

What is love?

A word that has been defined surrounding certain chemical reactions in the brain.

TIME Magazine actually kinda talked about this issue. The rate at which data can be processed is doubling every 18 months. This means that data growth is exponentially increasing. At a certain point, there is going to be an explosion where processing speeds are increasing blindingly fast.

Estimated at around 2025, one computer will have the processing power of the human brain. By ~2050 one will have the processing power of all the human brains on Earth. At this point, we wouldn't even be having to think anymore. It would be computers being their own computer scientists, figuring out how to make advancements in the field even faster, and faster.

Assuming that technology doesn't plateau.

No.

1. You still have to tell the computer HOW to solve the problem, which in turn may be beyond our ability.
2. There are problems that are not logic based such as "Who should I vote for?".  Or "Which religion is correct?"  Remember, to solve a problem, the solution must be acceptable to the person who has the problem.

This conversation reminds me of the Matrix:

No.

1. You still have to tell the computer HOW to solve the problem, which in turn may be beyond our ability.

How about machine learning though? I mean, computers won't go making more computers because they don't have the means for it, likewise they won't go killing all humans (or use them as energy source) but they should be able to solve problems that we haven't told them how to do.

No.

1. You still have to tell the computer HOW to solve the problem, which in turn may be beyond our ability.
2. There are problems that are not logic based such as "Who should I vote for?".  Or "Which religion is correct?"  Remember, to solve a problem, the solution must be acceptable to the person who has the problem.

This conversation reminds me of the Matrix:

At first you do, but if you make their problem solving complex enough, and they are fast enough, is it not beyond reason that they could deduce further algorithms? Sure, they can't make emotion based choices, but they could certain make for science and technology based ones.

the thing is your brain is technically functioning the same way as a computer, with electrical signals and storage.  its just our brains are more effecient.  our neurons either fire or not, like binary.  a future computer that is as powerful as all the brains in the world is probably able to solve anything our brain solves.  its just that we should make fail safes first

My CPU is a neural net processor; a learning computer.

Any problem, as in, any problems humans have demonstrated can be solved?

1. You still have to tell the computer HOW to solve the problem, which in turn may be beyond our ability.

I immediately thought of this as well.

The only way a computer could solve any 'solvable' problem, is if it had a comprehensive understanding of concepts and their relations. Right now a computer must be taught how to add 1+1 bitwise. 00000001 and 00000001 simply amended to 00000010 via predefined logical gates. Yet this is still not an understanding; it is just following predefined operational rules. It is being spoon fed the steps.

...It is similar to the difference between memorizing the math formulas, and understanding why they work. And understanding will carry you further, why memorization is still dependent on others for providing the formulas.

However, I do believe it is technically possible. Especially since I view humans as biological computers.

What is love?

Highly subjective. Words are arbitrarily invented to describe (usually) objective concepts. A word without objective meaning is logically nonsensical. Same goes for the soul.

I could ask a computer "what is grbjskja,lo?" and it would be equally perplexed by my insane question.

I immediately thought of this as well.

The only way a computer could solve any 'solvable' problem, is if it had a comprehensive understanding of concepts and their relations. Right now a computer must be taught how to add 1+1 bitwise. 00000001 and 00000001 simply amended to 00000010 via predefined logical gates. Yet this is still not an understanding; it is just following predefined operational rules. It is being spoon fed the steps.

...It is similar to the difference between memorizing the math formulas, and understanding why they work. And understanding will carry you further, why memorization is still dependent on others for providing the formulas.

However, I do believe it is technically possible. Especially since I view humans as biological computers.

You cannot program a computer to be capable of learning?

You cannot program a computer to be capable of learning?

I fully believe it is possible, but not by an extension of our current technological trend alone.
We'd have to reinvent the wheel to an extent.

Perhaps we could add a new sector of memory that allows the computer to organize, connect and assign concepts. Computer Driven Memory (CDM)? The static operating system would provide a foundation to allow it to build a dynamic one more efficiently. No current memory allows for neuroplasticity. All the low-level rules of memory, are held constant in current technology. It makes it a billion times easier for humans to work with, but also creates constraints.

Of course it would have to be constantly fed experience data in order to extrapolate patterns, relations and ultimately concepts.

We have actually already made ground on reasoning robots. It isn't exactly a different kind off memory though.

http://en.wikipedia.org/wiki/Artificial_intelligence#Tools

We have actually already made ground on reasoning robots. It isn't exactly a different kind off memory though.

http://en.wikipedia.org/wiki/Artificial_intelligence#Tools

The mention of new memory was intended as one approach to increase efficiency. After all, emulating a virtual dependency construct sounds very resource intensive.

However, if I'm not mistaken, the existing 'AI's' are not deducing anything. It is analyzing guess-and-check scenarios in order to form models that approach an optimal solution to a predefined problem.

Imagine if it had to solve basic math problems by looping through a guess-and-check approximation methodology? To be seamless and unbounded, it has to understand addition and then correctly deduce how to apply it. I want an AI with no failed attempts.

The way I understand the human thinking process is that when you have something, let's say an addition sign, it connects to a bunch of other things much like a database.

So it connects to the process of addition, it connects to the word plus, addition, sign, it connects to the letter "t" though loosely, and it connects to the christian cross (loosely). 
Each connection branches out to other things that the brain eventually relates to it.  So when I see "1 + 1" i could branch out to the number 1, then the process of addition, then the number 1 again OR I could branch to my memory that has the answer already since the equation "1+1" is already in my memory.

The way I understand the human thinking process is that when you have something, let's say an addition sign, it connects to a bunch of other things much like a database.

So it connects to the process of addition, it connects to the word plus, addition, sign, it connects to the letter "t" though loosely, and it connects to the christian cross (loosely). 
Each connection branches out to other things that the brain eventually relates to it.  So when I see "1 + 1" i could branch out to the number 1, then the process of addition, then the number 1 again OR I could branch to my memory that has the answer already since the equation "1+1" is already in my memory.

Or maybe you already do the first process, but it just happens so quickly you do not even realize it.

how i think the thinking process is is:
first you are prompted (like entering a variable)
     so the word "addition" can give you a+b, a*3=a+a+a+ ...
you take the prompt, and break it up to understandable parts
then you do a search through your head of valid responses
through those responses, you then cycle through them and make a best fit
then you respond.

so 1+1
you think of your math, or whatever
you break it into 1 object-addition-1object
math becomes best response
then you have 1 + 1 = 2
and you say 2

i think learning is this way
you are given the information
you sort the information into subcategories withing your total knowledge
you then make connections with that information to other places

But a better computer could recognize you might need to use 1+1=1

Will there come a time in the not so distant future, when we as a people can ask the computer what the cure for cancer is, and the computer can then formulate a cure?

This problem is basically http://en.wikipedia.org/wiki/P_versus_NP_problem

Quote from: Wakka Wakka on March 22, 2011, 06:06:31 AM

Quote from: EnglshGentleman on March 21, 2011, 11:19:36 PM

TIME Magazine actually kinda talked about this issue. The rate at which data can be processed is doubling every 18 months. This means that data growth is exponentially increasing. At a certain point, there is going to be an explosion where processing speeds are increasing blindingly fast.

Estimated at around 2025, one computer will have the processing power of the human brain. By ~2050 one will have the processing power of all the human brains on Earth. At this point, we wouldn't even be having to think anymore. It would be computers being their own computer scientists, figuring out how to make advancements in the field even faster, and faster.

Processing power does not equate problem solving though.

How doesn't it? Modern AI use algorithms for problem solving and reasoning. The more complex the puzzle, the more time and memory a computer requires to solve it. Right now that is a problem because complex things take monumental amounts of both of them, but if in the future processing speeds are blindingly fast, then this would not be an issue.

The problem with AI are not resources, its the theory imho.  AI still needs to find its "Einstein".

However, if I'm not mistaken, the existing 'AI's' are not deducing anything. It is analyzing guess-and-check scenarios in order to form models that approach an optimal solution to a predefined problem.

This is how the conscious part of the human mind works, too. We make mistakes because we have each created our own model of how mathematical operations are to be applied, rather than being programmed to execute those operations flawlessly.

Being able to pick up new skills quickly and being able to use those skills perfectly are necessarily contrary goals, because one requires volatile memory and the other reliable memory.

Here is the actual article in TIME I was referring to.

http://www.time.com/time/health/article/0,8599,2048138-1,00.html

Quote from: ﮎingulaЯiτy on March 23, 2011, 10:30:15 AM

However, if I'm not mistaken, the existing 'AI's' are not deducing anything. It is analyzing guess-and-check scenarios in order to form models that approach an optimal solution to a predefined problem.

This is how the conscious part of the human mind works, too. We make mistakes because we have each created our own model of how mathematical operations are to be applied, rather than being programmed to execute those operations flawlessly.

Being able to pick up new skills quickly and being able to use those skills perfectly are necessarily contrary goals, because one requires volatile memory and the other reliable memory.

Good point. Though, many things that are abstractions of reality are taught as absolute rules. For instance, I would that anything heavily based in logic and observation, say the rules of economics, could be reinvented, but instead they are taught to each generation as static rules. Of course there is an element of practicality when humans are learning. They are slow, forgetful, and lose the information when they die. In those respects, computers should have much more potential.

I suppose when I mentioned 'an AI with no failed attempts', I should have said an AI that can derive possible answers, but tests those answers before giving credence and/or expanding upon the results. Humans seem to confuse beliefs with facts all the time, or weight evidence improperly, while a computer should not.

Here is the actual article in TIME I was referring to.

http://www.time.com/time/health/article/0,8599,2048138-1,00.html

"This transformation has a name: the Singularity. "

That's hawt.

But a better computer could recognize you might need to use 1+1=1

1+1=2