The computer uses transitors which close and open allowing for only two
positions to use a binary code of 1's and 0's. The semiconductors that the
small transistors in the chips are made of will either conduct current or
not conduct current according to temperature. So, the millions of
transistors act like swithes which are either open or closed, to make a 1 or
0 for the binary code.
Let's consider these switches to be saying either yes or no. Comparing this
to human thought which has imagination and free will, it seems that human
thought also has maybe, and not just yes or no.
So, suppose a computer could be designed with transistor switches that say
yes, no or maybe. Most of the transistors would be the standard variety that
just say yes or no so the binary code can work in most areas of the computer
the way that it does now.
However, to give the computer free will and imagination, in the calculating
part of the computer certain key positions in the circuits could have these
yes, no or maybe swithces. The computer would excercise free will and
imagination at those points.
The computer would be no longer just a calculator but something beyond that.
How would you design a yes, no or maybe switch?
Well, possibilities could be:
If the switch is on, and you send it the signal to go off, then it either
goes off or stays on. Likewise if the switch is off and you send it the
signal to go on, it either goes on or stays off.
Another possibility is that the switch goes into a mode where it switches
back and forth from on to off, repeatedly.
These special maybe switches could also be set up so that they act like
regular on and off switches sometimes, and then sometimes they say maybe.
Perhaps a random signal generator could be used to help with the maybe
switch.
Experimenting could be done to see if this is feasible.
Now, a computer that thinks for itself, just like humans that think for
themselves, would probably sometimes make mistakes. However, the computer
that thinks for itself, having an imagination, might also see things
correctly in a totally new way, and this could be very beneficial because of
the large amounts of data that computers can handle in a short time.
It may also lead to terrible trouble with a rebellious computer that wants
to take over like Hal on the movie 2001 and collosus on the movie, Collosus
the forbin project.
Any comments:

Not possible.

First of all... that wouldnt give computers free will.

Second of all, there are ways of making a conventional transistor with 3 states, or a biological transistor with upwards of 40 (47 IIRC) different states.

You might be interested in looking up "Fuzzy Logic", which is a sort of a 'maybe' for computers.

I think this would just result in an unreliable peice of hardware. But I could be wrong.

Some fundamental errors. Transistors do not switch by heat, they switch by either current or voltage, they are either current mode switches or field effect switches.
Transistors used in computers are switched; however, this is but one use of xsistors. The other use is in linear operating regions, to wit: amplifiers and such.
Before going into hardware to develop a 'thinking' puter, study some artificial intelligence and some control systems. In ai, we can approximate some types of thought or deduction, but if the inputs are not controlled properly, the thing will go insane. The best way of dealing with this is with 'fuzzy logic', which is capable of wider ranges of decision making.
A bigger problem is that the human mind receives and processes far more information than is possible to do with computers of any known type. We operate in a linear range rather than a digital range. In order for a digital puter to operate in the linear, there must be converters to make the analog into a digital number for the puter to crunch. A-D converters consume time, even with DSP puters, and insert certain inherent inaccuracies, which must be dealt with, consuming even more time or increasing hardware, which will also increase time.
All we can do is simulate. Let's take a for instance: how about the fly that is hovering out of the sun under your back porch roof. That fly, to hover, must compensate for breezes, thermals, visual objects moving around it, and so on. Further, since he is likely hunting for his next meal, he's got to keep his eyes on those things that might be, as well as the smells and humidity, directional location, and so on. He also needs to keep his senses alert for danger in his vicinity. All of these things are handled concurrently, and we have no computers available to us that would even approach such activity.
Yes, we have concurrent execution; however, that multitasking that is so highly regarded is still, in fact, multitasking in sequence. A standard computer cannot do two programs at the same time. Such things as DSPs can do some concurrent execution, A-D conversion, and so on at the same time (which makes them fast), and usually have an architecture such that instruction execution occurs in one or two clock cycles, possibly more, depending on the instruction. As an example, a good DSP will be able to construct a passable 60 Hz sine wave (as in puter backup power), but only if there are hardware filters available. If the DSP is given series of faults, it will not cope, some will get by, or it won't correct all the faults. A fault might be such as a switched load (turning on a light attached, for instance). But the compensation will be limited, however acceptable. Just an example of what a fast dedicated computer will and will not do. Making a fly is very difficult.
Imagination is simply not possible for a machine whose basis for calculation is simple binary/trinary operation. Also is virtually impossible for a linear machine, since the basic input must come from somewhere.
Interesting idea, though. Don't stop thinking.

Try looking up TTL(transistor-transitor logic), fuzzy logic, "float", etc. "float" is the third state, like what you're describing, and it's not exactly an ideal state.

Neural networks allow for a certain level of "maybe" to "learn", but from what I know about brain neurons, "dumb" neural nodes cant and wont compare to "smart" brain neurons. Also its takes an array or network of "dumb" nodes to have its "learning" ability, more than just "maybe" transistors.

Also, youre still talking about using silicon to build your "brain", which has serious limitations when trying to fit a brains worth of transistors into one package. Moores law wont allow that for at least another decade, and then theres the software... many experts say that for every 1 hour of hardware engineering it creates 24 hours worth of software engineering.

Look up DNA and "biologically inspired" computing. It's still in its infancy, but its abit like what youre talking about, but far beyond in # of potential states.

I personally think they'll just skip all of that and grow live brains from live "smart" brain neurons. Well they already did that, but it was a small one, and it only flew a flight simulator. I think it's possible to grow them in enourmous sizes, and hook them up with massive arrays of electrodes ala BrainGate technology. Once they prime that project it'll kick the DNA and NBIC programs into overdrive. Check out my Beyond AI thread in this section to blow your mind.

If for example you set a transistor switch to be random, that means probability of 50% of the time it will switch when signalled to do so and 50% of the time it won't switch, then you set it to perform randomly. Suppose the computer, which has patterns of electric and magnetic fields in it, will cause the switch to switch when it is signalled 70% of the time. Or to switch 70% on and 30% off. That would mean that the computer has made a decision as to how it is going to switch. It would not be random after the computer used it. You could call that a primitive form of free will or imagination.

Randomness in a computer program would just be randomness, and could not be free will. Randomness in a transistor, that is set to either switch or not switch, could be affected by something as small as a magnetic or electric field variation hitting it, to determine how it will switch or not switch. The make up of the computer [or the running of a different program] could cause the small silicon transistor to switch 70% on and 30% off, in which case it is no longer random and the computer has decided for itself which way it is going to switch. - primitive form of free will
Real randomness must maintain 50% one way and 50% the other.
A random computer program, no matter what language you use in the software will always be random, and could never be influenced by the computer. A random transistor could be, because it is so small and easily influenced by small magnetic or electric fields, if it is set to go one way or the other. [A functioning computer should be full of very small, stray electric and magnetic fields.]

a computer with free will just makes me think of an ATM that wont give me my money lol

Everything you said in this post may be true for an ordinary computer, with yes or no transistors. An ordinary computer is just a very complicated calculator.
Giving a computer transistors that are set to go one way or the other, which can say maybe, and not just yes and no, gives the computer circuitry itself, the decision as to what way the switch will change. [Any functioning computer should have many stray, small magnetic or electric fields that would determine what way the switch goes, if the switch is set to be balanced to either go one way or the other.] That would mean the computer is deciding how the switch will change, to either yes or no; -- this could be a primitive form of imagination or free will on the computer's part.
There may be many things possible with a computer having yes, no or maybe transistor switches, which are not possible with ordinary computers having only yes or no switches.
I agree with you that imagination is not possible for a calculator.
Giving the computer the ability to determine which way switches will go by its own functioning, means it has become something else (in certain circuits) different from just a calculator.

My pc has got free will its chosen to crash 3 times this morning hmm and sometimes it just sais no no no