Non-binary computers?

[farmkid]

Here's a question for those of you who know more about computer science than me:

I've been thinking lately about the binary computers we currently use. Everything I know of works in a binary fashion. Well, what if a transistor or transistor-like device was developed that had three states instead of two. How would that change what computers could do? Or would it? Would performance change at all? How different would programming for it be?

Now, extend that idea further and imagine if instead of on/off transistors, we could build a computer using some sort infinitely variable transistors--small rheostats, perhaps. In other words, the transistor could be 1 or 0 or anything inbetween. It seems to me that would enable computers to solve problems we can only approximate now. How would such a computer be different in practical terms compared to what we have now?

 

[RobertVA]

For three state computers instead of the two state ones in common use programming and capabilities probably wouldn't be much different. Binary is so cumbersome for humans to use it's usually converted to some other base like hexadecimal, octal or even decimal anyway. With three states there would be other systems based on powers of three (maybe 9 or 21) instead of powers of two (octal is base 8 and hexadecimal is base 16):

For the uninitiated:
Hexadecimal substitutes A, B, C, D, E and F for 10, 11, 12, 13, 14 and 15. This results in each hexadecimal digit always representing the same sequence of four binary digits regardless of where the hexadecimal digit appears in the sequence. Example: h08 = b 0000 1000 and h80 = b 1000 0000.

Octal is similar, but only matches three binary digits, thus no 8 or 9 in octal ( 8 = o10)

One of the important advantages of digital is the intermediate neutral zone between the "on" voltage and the "off" voltage. A bit to much electrical resistance produces a result in that range and the error is quickly identified. With an analog computer such a situation might become a difficult to detect change in the data.

 

[USS KG5]

If you read into Quantum Computers you might find it enlightening, they are a fascinating concept.

 

[TK421]

Quantum computers, or which simple experimental setups appear to be up and running have, 1, 0 and a Qubit, which is 1 AND 0 at the same time. You should look up the work of Seth Lloyd.

 

[Deleted 2]

farmkid said:
Now, extend that idea further and imagine if instead of on/off transistors, we could build a computer using some sort infinitely variable transistors--small rheostats, perhaps. In other words, the transistor could be 1 or 0 or anything inbetween. It seems to me that would enable computers to solve problems we can only approximate now. How would such a computer be different in practical terms compared to what we have now?

Fuzzy logic involves predicates that aren't "true" or "false", but have a certain probability of being either. I'd expect your 'analog transistor cpus' to be very fast at solving them.

The problem is, we can easily and quickly simulate 'fuzzy' logic using 'hard' logic, but not the other way around (i.e. you want to avoid the chance, however small, that 'true' becomes 'false').

 

[Brent]

Sounds like Quantum Computers to me, the holy grail of computer performance we are all waiting for.

 

[Arrqh]

TK421 said:
Quantum computers, or which simple experimental setups appear to be up and running have, 1, 0 and a Qubit, which is 1 AND 0 at the same time. You should look up the work of Seth Lloyd.

Actually, a quibit is just a quantum bit, where the information itself is stored. It can be in a state of 0, 1 or the superposition 0 and 1. In other words, quibit isn't the name of a state, but is the name of the unit of information itself.

I'm just being anal about terminology

 

[SamuraiBlue]

farmkid said:
Here's a question for those of you who know more about computer science than me:

I've been thinking lately about the binary computers we currently use. Everything I know of works in a binary fashion. Well, what if a transistor or transistor-like device was developed that had three states instead of two. How would that change what computers could do? Or would it? Would performance change at all? How different would programming for it be?

Now, extend that idea further and imagine if instead of on/off transistors, we could build a computer using some sort infinitely variable transistors--small rheostats, perhaps. In other words, the transistor could be 1 or 0 or anything inbetween. It seems to me that would enable computers to solve problems we can only approximate now. How would such a computer be different in practical terms compared to what we have now?

basically what you are descibing is an Analog computer.

 

[Jadzia]

farmkid said:
Here's a question for those of you who know more about computer science than me:

I've been thinking lately about the binary computers we currently use. Everything I know of works in a binary fashion. Well, what if a transistor or transistor-like device was developed that had three states instead of two. How would that change what computers could do? Or would it? Would performance change at all? How different would programming for it be?

Now, extend that idea further and imagine if instead of on/off transistors, we could build a computer using some sort infinitely variable transistors--small rheostats, perhaps. In other words, the transistor could be 1 or 0 or anything inbetween. It seems to me that would enable computers to solve problems we can only approximate now. How would such a computer be different in practical terms compared to what we have now?

basically what you are descibing is an Analog computer.

http://en.wikipedia.org/wiki/Three-state_logic
http://en.wikipedia.org/wiki/Ternary_logic

read and digest