Thread: Artificial intelligence: Electron energy states

I'm going to paraphrase what my chemistry teacher briefly explained in my class, and want to know what you think...

Could a machine possibly read the energy level or number of electrons of an atom to derive solutions in "gray areas" rather than the Black-and-White processing of current machines, thus giving machines the ability to say "maybe"?

Every element on the periodic table has properties, based primarily on electrons. Could we create a "chemical" computer that is based on the readings of these properties, coming from a core, a cocktail of elements rapidly changing, manipulated as required, and being read? Instead of basing our computers on the simple on/off of transistors, what about basing them off of the readings coming from a mix of atoms?

Current computers are based on transistors that turn "on" or "off", rapidly, to create the basis of binary. However, more complex computers could be created using, as a basis, the properties of various atoms/elements, like how many electrons the atom has, and at what energy level they are at, to create more complex thinking. Of course, we would create a "key" from which the output of the computer would translate into language or other complex calculations.

Is he on crack, or is there some validity to this far-fetched, theoretical hypothesis?

Is he crazy? Probably not.

Is he really reaching for things? Oh yeah.

Could it happen? Maybe.

He is missing a major part of electronics though. There is a yes-no-maybe architecture to all logic devices. A lot of newer devices are tri-state or more. The High-Z state could be considered a Maybe state. Or the state in between what humans have defined as hard on or hard off. On most systems it's from about .9 V to 1.65 V that is a grey area. We simply don't know what the chip will do in that area because it isn't always the same.

At a much higher level, experimenters are starting to really understand non-linear math. These types of operations could be used to come up with "unpredictable" results that are never exactly the same. This could be the key to machine AI, or just a tool for it to use.

Nature's living organisms are DNA code based complex computers. They work with codes and uses chemical elements to make functional and robust systems. This system can even multiply itself and adapt to its environment through mutation. So, humans have three possible ways:

1. Continue on existing un-natural yet logically human (on and off, 1 and 0) digital systems and develop new types of un-natural systems which should only based on human logic distinctively.
2. Forget about our separate ways: follow and copy what nature does, and adapt them for human purposes.
3. Use both methods together and alternatively.

The unimaginable fourth way is to give up our computers. I don't think so...

Originally Posted by corsican

I'm going to paraphrase what my chemistry teacher briefly explained in my class, and want to know what you think...

Could a machine possibly read the energy level or number of electrons of an atom to derive solutions in "gray areas" rather than the Black-and-White processing of current machines, thus giving machines the ability to say "maybe"?

Every element on the periodic table has properties, based primarily on electrons. Could we create a "chemical" computer that is based on the readings of these properties, coming from a core, a cocktail of elements rapidly changing, manipulated as required, and being read? Instead of basing our computers on the simple on/off of transistors, what about basing them off of the readings coming from a mix of atoms?

Current computers are based on transistors that turn "on" or "off", rapidly, to create the basis of binary. However, more complex computers could be created using, as a basis, the properties of various atoms/elements, like how many electrons the atom has, and at what energy level they are at, to create more complex thinking. Of course, we would create a "key" from which the output of the computer would translate into language or other complex calculations.

Is he on crack, or is there some validity to this far-fetched, theoretical hypothesis?

Quantom Computer's, Ditigital Computers, DNA computers yea the question Isn't could a machine it should can we stop the machine........

tri-state has been around for a while, it is more so in physic engines within video games, and the AI of mob's encountered with in the physics engine. your chem teach should look in to the possibilities of quantom computers.....







and if any one has a QC out there they would be willing to let a gamer test let me know

I'm going to go ahead and say that this was actually my idea, and I made up the part about my chemistry teacher. Well, this idea came to me during an intro chemistry class, as we were learning about electron energy levels and such. I acted like my teacher made it up, because I wanted to avoid possible embarassment on what seems to be a forum for geniuses.

It was at a time when I was really curious and fascinated by artificial intelligence. At the time, I had also read about how our brain does not only think in black/white, but with gray area in the middle. I realized that current transistors went against this, and that current attempts at AI were still based on 1/0 basics. So, during that lecture, I just "put 2 and 2 together". We could avoid binary in much more natural ways, by changing the way computers think, from the bottom up. And electron energy levels are much more complex than 1/0, there are tons of possibilities. Instead of having the computer "read" 1s and 0s based on simple electricity, it would read the electricity itself, in a sense, by differentiating the status of the electrons.

I'm 17 and will be applying to colleges... some of their entrance essay prompts ask about things like how we took school material and thought about it in our own ways. Would touching on my experience in the chemistry class be laughed at by admissions officers, or is it a viable, respectable idea, that shows my ability to think originally, especially in the field of science and technology? Oh and I'm really sorry for being initially misleading...!

I'm asking it on these forums, hopefully to listen to the advice of something more knowledgeable about physics, chemistry, the atom, and computers, to avoid embarrassment in the future. My idea doesn't have to be completely correct, just interesting enough and possible enough to impress college admissions officers. As in, like, MIT people. :-D

Originally Posted by thedrunk

tri-state has been around for a while, it is more so in physic engines within video games, and the AI of mob's encountered with in the physics engine. your chem teach should look in to the possibilities of quantom computers.....

You can use the binary mathematics of a computer to model as many states as you want. If you want consider 256 different states, just assign each state an 8 digit binary number. If you want to consider 1000 states, use a 10 digit binary number.

He seems to be wanting a computer that doesn't work in binary. People built some base-three computers back in the 1950s, but gave up on them because in general binary seemed better (easier to program, easier to make).

Do any of you think that binary might not be the best way to run a computer?

Our computers use binary because it was easier to make, from scratch, by the hands of humans. It was also sufficient for math and logic, when only one answer was needed.

Now that we have adequately powerful computers, which happen to use binary, could we use their processing power to develop a computer more complex and more difficult to create, a computer that would use, say, base 8 or 12 or more? While it would be difficult to create, wouldn't the computer be capable of much more compact and speedy calculations? Wouldn't such an architecture be more inviting to the "gray area" our brains work with?

Instead of a 1 or a 0, it could end up with a 3, 5, or 9 etc, each of which would evoke another set of possibilities, which would invite a more complex tree of possible solutions, especially useful when our computers would need to deal with possibilities and/or assumptions. I think these characteristics need to exist if the computer is built to learn and adapt. Learning is hard when you can only deal with either 1s or 0s. While it is possible, it's just more lengthy, and not tuned towards the goal.

I personally think that a more complex and less rigid foundation for a computer's thinking process, such as what I tried to explain, would be more convenient for programming a more and independent adaptive computer. So far, we're just using binary to emulate this process.

And instead of using voltage differences to define the extra numbers, mayve we could manipulate electron energy states, and use those states as a means of reading higher base numbers. 1 electron, 5 electrons, etc, would correspond to the 1s and 5s or 8s etc.

Originally Posted by corsican

It was also sufficient for math and logic, when only one answer was needed.

Computers perform arithmetic calculations, and arithmetic calculations always have only one answer. Doing the arithmetic in a different base won't change that. If I ask "What is five plus five?" it doesn't matter if I work it out in base ten as 5+5=10 or in binary as 101+101=1010. I still get the same answer.

Now that we have adequately powerful computers, which happen to use binary, could we use their processing power to develop a computer more complex and more difficult to create, a computer that would use, say, base 8 or 12 or more? While it would be difficult to create, wouldn't the computer be capable of much more compact and speedy calculations? Wouldn't such an architecture be more inviting to the "gray area" our brains work with?

As I said, doing the arithmetic in a different base won't change the fact that the arithmetic only has one answer. There is no "gray area" in the answer to an addition or subtraction problem.

Don't get me wrong, there might be good reasons to make the sort of computer you describe; maybe it could calculate faster or use less power or something. I just don't think that doing your math in a base other than base 2 would necessarily make the computer better able to see "grey areas."

Well I agree with everything you said.

However, I think that the next generation's computers will do more than solve problems with a single answer – they will need to be designed to make assumptions, or derive the "best solution". And while we could achieve the same ends through binary than through any other base, binary would be the least efficient.

Binary was useful when we only needed computers for simple problems with one answer, it might not be the most efficient way of working problems with no definite solution.

Base 20 (or any other base) wouldn't be more capable of seeing "gray areas" than binary, but it would use less energy and would work faster at doing problems where "grey area" is required or frequent. A computer working at a higher base is just more fine-tuned to the more subjective environment of problems to deal with.

Originally Posted by Scifor Refugee

Originally Posted by thedrunk

tri-state has been around for a while, it is more so in physic engines within video games, and the AI of mob's encountered with in the physics engine. your chem teach should look in to the possibilities of quantom computers.....

You can use the binary mathematics of a computer to model as many states as you want. If you want consider 256 different states, just assign each state an 8 digit binary number. If you want to consider 1000 states, use a 10 digit binary number.

He seems to be wanting a computer that doesn't work in binary. People built some base-three computers back in the 1950s, but gave up on them because in general binary seemed better (easier to program, easier to make).

well a quantom takes binarry to a new level nto only can it function on 1 and 0 it can do a 1, 0, and 10, 01, but only in a tri state of function which you would either have a 1, 0, 10 or 0, 1, 01 ( the simutaimies 10 or 01 as a third to add to 1 or 0 process)

which I feel is more to what he is trying to say it would sorta be liek hwo a DNA computer will reproduce till it solves a given problem and store the answer in its data base for if that process comes forth agian. I am thinking he is referring towards a AI because he wants the base function to learn which besides ai processes I cant think of any thing .....

Excuse my ramblings...

The notion of 'trinary' or other models is interesting but but I think the main need in AI development is processing speed and programming logic. You need to have hardware working together with clever programming.

Computers solve problems. There's no real thought, just the length of time it takes to crunch numbers. I have thought about AI but my thoughts often slam into a wall of reality. It might be faster but "smarter"? I think is a whole separate issue. Greater speed, imo, simply facilitates more processing that could be done to simulate a more natural response. But the problem with AI is that it has to be programmed to simulate life. What makes us intelligent is not our hardware. Our hardware facilitates human intelligence, providing a network of computers passing data between them. I do think that the only way to approach our level of intelligence is by looking to nature's model. I personally view it as an alien feat of bio-mechanical engineering.

The first thing we have to do is understand how our bodies communicate. I think energy radiates most naturally in waves. Sending bursts of information may be slower than sending a "wavelength" of information (similar to the curvature of a DNA) and reading the entire wavelength as opposed to reading whether or not it exists. Data storage needs to improve as well as data transmission. We transmit data in waves already simply because it is the natural state of directed energy (radio waves for instance). Instead of receiving the wave like a catcher, we could take a picture of it as it is received from the side or construct a chamber in which the walls are sensitive and "feel" the wave. We then have data being read as subtle variations of "feelings".

In the human body there are actions that take place without thought and actions which somehow start as the thought is forming in your mind. How is it that my typing can just about keep up with what I'm thinking when typing is far more complex a thing to do? In reality, this is what we're doing to computers. Computers have to take our instructions and break them down into something that makes mathematical sense. I think what you'd need to do is go beyond mathematical sense and allow compilers to break down instructions in new ways. A computer is constantly serving as an interpreter. It has to go from English to programming language of choice to machine language to binary. It's native language is binary. I think the original poster was getting at the complexity issue which I guess can also be a speed issue. Sure, you could represent the same information both ways but the real question is which would be faster?

I think in order to fake intelligence the AI has to be given experiences that it can learn from. It has to be able to record its own logical path through problem solving and then instead of breaking down a and calculating 10 + 10 = 20, it could just remember certain calculations so that it never has to make them again. Imagine if a computer only had to make NEW calculations? What if a computer could ask its own RAM what 100 + 100 is? If we give a computer certain formulas it should be able to detect when these formulas are applicable in other situations and use other information, other than numbers, in order to solve problems.

We would also have to create morality with emotional, moral, and spiritual formulas for understanding life. Often physical attraction can be simulated by some formula of parental attributes and experiences with the opposite sex. If you can take these experiences, turn them into solvable problems and average the results together, the AI should be able to make a decision based on more than just logic. Because in my opinion its the combination of a host of variables that make us unique and construct the human decision making process. For something else to act human it would have to be given (virtually if nothing else) all the things that humans have; including bodies. Otherwise, how would it know how fast to walk with an old lady without any concept of the tiredness and strain put on her body? And it has to understand that all humans are different based on their culture and experiences so that it can take all these factors into consideration the same way we do on a daily basis. Otherwise, it wouldn't know how to handle us and may use what it does know (fear and anger) to cause us harm. And if their body is synthetic their calculations through life would need to be adjusted to accounting for this difference.

Sadly, it would even have to account for human ignorance, injustice, and prejudice.

And got off the original subject but I think we need to go beyond both 2 and 3. Most polls use a scale of 1 to 5 or 1 to 10 in order to express a wider array of answers between yes or no, agree or disagree, love it hate it, etc. An AI needs to feel strongly about some things and not strongly about other things so that it can be firm and resolute but also be open-minded and humble (and no there's no sense in simulating humble responses if in its own mind the AI always thinks its right anyway). The idea of being wrong is somewhat foreign to a computer. But even after calculating billions of possibilities and assumptions it can still be wrong in action or in communication or in understanding a human (because humans aren't perfect and bug up everything we touch). "Data" from Star Trek just came to mind. I think his character reflects the gist of what I'm saying.