Toadfish Thought of the Day

[Griffin NoName]

Computers mesh so well with the human brain precisely because they are designed to think in a different way than we are.

I don;t think they do think in a different way to us. When we are thinking certain things, different bits of our brain light up. ie. brain cell off or on. Computers think in terms of off or on. 1 or 0. Yes or No.

[Bob in a quantum-state-of-faith]

Computers mesh so well with the human brain precisely because they are designed to think in a different way than we are.

I don;t think they do think in a different way to us. When we are thinking certain things, different bits of our brain light up. ie. brain cell off or on. Computers think in terms of off or on. 1 or 0. Yes or No.

Perhaps so.  But, it is also quite clear that our neurons, or more accurately, the interconnection between various neurons are analog, not digital.  That is to say, each connection varies in intensity--sometimes a great deal of difference between the weakest and the strongest signal.  This permits near-infinite variation at every given neuronic connection point, which adds up to an amazing capacity.

Contrast with digital computers, who only experience a single intensity on or off at each transistor connection point.

In an interesting side note, Iaasic Asimov predicted that analog computers would be the choice made by engineers, rather than digital.

If engineers ever do decide to create analog computers, then perhaps they can be similar to human neurons.

[Bluenose]

Actually, analogue computers have been made.  The static (no motion) Grumman Tracker simulator I was trained in when I was learning anti-submarine warfare was an analogue computer.  It was a wondrous Heath-Robinson affair with all manner of weird looking devices, push-rods and bell-cranks,  variable speed motors, screw jacks, carefully calibrated linear amplifiers, and so on.  It actually worked remarkable well, but the digital six-axis full motion simulator for the Sea King helos was actually far more capable (even apart from the motion) and the controllers were able to do a great deal more to simulate real world situations.  Of course this was in the early 1980s and the computers used would have been both incredibly expensive and not very high performance by today's standard, but nevertheless it worked very well.

[Swatopluk]

For some specialised tasks hybrid computers are still in use. But the Achilles heel is always the analogue digital converter.

[Bob in a quantum-state-of-faith]

Indeed.  I've seen even earlier dabbling with analog computers. Probably the most famous, was Babbington's mechanical one, which is often used in steampunk fiction.   I also remember reading about a cold war era Russian computer that was built around base 10.  The details escape me, but it was essentially a digital machine, but instead of binary/hexadecimal it was base 10.  I seem to recall it was entirely BCD, which would've been a waste of capacity.   

I've also read of experimental trinary computers, or ones based on a base-3 system.   There really is no reason why engineers couldn't create base 3 or base 4 machines, apart from the memory issue-- transistors can easily be constructed to be analog (and in fact, they often are used in analog circuitry, such as audio amplifiers).  But memory is an issue-- currently, there are two basic memory types in common use, magnetic and capacitive.  Magnetic is those spinning metal disks commonly referred to as "hard drives".  Capacitive is everything else memory-electronic.   For a 1, the single memory cell (a single microscopic capacitor) has a measurable charge, a positive voltage.  For a zero, the capacitor is discharged.  This is how electronic memory works.  To make it other than 0 or 1, you'd have to develop a 3rd state that is both stable and easy to measure (to read it back out).  A varying level of charge won't do-- as all capacitors leak a bit of charge over time, so it would be very difficult (if not impossible) to know for sure, if the remaining charge was supposed to be a 1, or a 2, or whatever-- as over time, 2's leak down to 1's and so on.  This limitation is also why dynamic ram has to have a refresh rate-- to re-up the charges of the 1's.  And reading D-ram is destructive:  you have to immediately "re-write" a 1 after "reading" it (seeing if it's a 1 and not a 0).

So, if engineers can come up with something other than capacitors for memory cells?   It appears we are stuck with binary computers.  There does appear to be some promise using organic molecules, by copying what evolution has created over the millinea.   

[Swatopluk]

Some things cannot be calculated efficiently with digital means but are open to swift analogue solutions. A hybrid computer will delegate such stuff to the analogue part and do all the other stuff digitally. One area where a very simple form of this is still in use (to my knowledge) is weapon guidance systems. Sensory data is sent first through a 'mixer' and thus pre-processed before it gets digitized an analyzed. At least in older models the 'output' too goes through an analogue unit before it goes to the actuators.

A classic case is the fingerprint database search system. Until a few years ago it was impossible to find matching fingerprints in large libraries by algorithmic picture comparision within an acceptable timeframe. The trick to process the data in reasonable time was to use an analogue optical interference unit. This allowed to throw out obvious mismatches more or less instantly, so the algorithmic (digital) system could concentrate on the small set of remaining cases.
The fingerprints were not stored as data files but as physical images* (I guess on some kind of microfiche or film that could be quickly moved through the optical analyzer).

*iirc optically preprocessed, so they did not necssarily look like images of fingerprints

[Bob in a quantum-state-of-faith]

That is an excellent observation, Swato.   Using weighted sensor data can seriously improve the final results.  I do remember a number of years ago, when the buzzword was "fuzzy logic", or as I like to think of it, weighted decisions.

I remember all they hype that "fuzzy logic" had in the media, but once I explored just what it was, it boiled down to weighting your inputs commiserate to how valid (or which ones, or how important, etc)  in order to make a decision.

Or to put it in statistical terms, some inputs have only 0.50 rating, whereas others could have 0.75 or 0.10.  This weights the various inputs for the final decision.   

Some sorts of information does not digitize well, and so analog analysis would be better. 

Ain't science fun? 

[Sibling Zono (anon1mat0)]

currently, there are two basic memory types in common use, magnetic and capacitive.

I imagine that from a very macro perspective, flash memory is capacitive, but we are talking electrons trapped in a semiconductor for years, no need to inject energy to keep it alive (like RAM) and not necessarily requiring to delete the cell in order to read its state.

[Bob in a quantum-state-of-faith]

currently, there are two basic memory types in common use, magnetic and capacitive.

I imagine that from a very macro perspective, flash memory is capacitive, but we are talking electrons trapped in a semiconductor for years, no need to inject energy to keep it alive (like RAM) and not necessarily requiring to delete the cell in order to read its state.

Actually?  As far as I can find out, flash memory is electrically alive, and has in each module some sort of stored power.  I do not know if that is true, but that's what I read from years ago about the subject.   I suppose I should have been more comprehensive, as I did not cover ROM memory nor PROM memory.  These two memory types use a "burn in" method to permanently modify the electronics (the exact details I have forgotten) using over-voltages as needed.  The zeros are represented by intact links, and the ones are represented by links who have been burnt out by over-voltages.  This is necessarily a one-off thing, but once written, do not require any sort of electricity to maintain.

A very primitive model of this could be built, using simple automotive fuses-- with a fuse holder representing each bit of information.  Applying over-current (in this case) would "blow" a particular fuse, changing it's value.  Then, so long as the voltages used to read the memory were low enough, you would have a permanent information representation.

In the case of ROM memory, however, the information is encoded during the actual creation of the electronics in question, a bit different from the "programmable" sort.  As far as I know, however, there is no longer any pure ROM being made, as PROM is much more general, and simple enough to make and program. 

Finally, there is the EPROM, or for Erasable Programmable Read-Only Memory  The "spiritual" ancestor to EEPROM for Electronically EPROM.  These were the first of the so-called FLASH ROMs. 

Incidentally, as far as I can find out, FLASH goes back to how you could erase the original EPROM modules:  you would expose it to UV light (usually after removing a little black-out cover).  This process was commonly called "FLASHING" the memory.   The UV light reset the links which were broken during the programming stage, which still used over-voltages to write.

Eventually engineers figured out how to create memory cells that could be reset by using even higher voltages (in some cases) or reverse voltages (in other cases), and the so-called EEPROM was born. 

For the modern flash ram, such as the ever-ubiquitous USB sticks, I do know that higher voltages are used for the write cycle, but lower voltages are used for reading.  In this way, the flash memory has a limited life:  there is a strict limit to the number of times you can write to these things, but reading them is unlimited.   So, if you want to use one of these as a storage media?  Flip it's little "write enable" switch, if it has one-- it will preserve it far longer, if you don't let Windoze write it's usual crap every time you insert it.  Re-enable writing only if you need to add or change it's contents.

But even then, or so I've been told (I'd love to be proven wrong, here, actually) after about 10 years or so, it will become essentially dead anyway, as it's internal "keep alive" power cell runs dry.   It'd be nice if I was wrong about that, as I do know that CD-ROM, and CD-RW discs have an average shelf life of about 5 years.  Less if they are exposed to even mild temperature fluctuations, or harsh lighting (such as what CFLs emit-- it's probably the UV that does it).



Who'd a thunk it?  A thread about Thoughts For The Day drifts into electronic memory models ....

(and yeah, I shoulda also mentioned how CD's, CD-ROM and CD-RW's work as a memory model, to be complete-- the don't use electronic anything to record 1's and 0's.   In the case of original CDs and DVDs (and original BluRay too) they use aluminum foil... yes, I said aluminum foil.  This is in the form of aluminum vapor vacuum deposited onto plastic, to make it reflective to the reading laser.  For a zero, the little pit is smooth like a mirror.  For a one, the little pit is rough, such that the laser beam does not bounce back to the sensor, but is scattered. (at least I think I got the order correct here--I may have it backwards)  Anyway, original manufactured CDs & DVDs can theoretically last forever, so long as the physical thing is in good condition.  But not so for the ones you can make yourself:  they use a dye instead of aluminum, to create those bumps.  Either the write-once versions or the erasable ones, both use a dye suspended in a viscous liquid (think stiff grease or non-hardening glue) as a minute layer between the aluminum layer which acts as a back-plane.   To "record" a zero, the laser does nothing.  To record a 1, the laser's energy (usually voltage) is increased, and it heats the dye--melting it, rendering it not-smooth-- a 1.  In the case of RW discs, an even higher energy pulse re-melts it yet again, to smooth it out back to a zero.  The fault is in this dye layer, which is extremely sensitive to temperature changes, which can smooth out the 1's or even wrinkle the 0's, scrambling the data over time.  UV light has a similar randomizing effect.   So, the lesson is:  do not depend on CDRs or DVDRs for long-term storage of your valuable data-- don't do it!.  Use actual hard drives, which seem to have a nice 10-15 year shelf life, if stored in a cool, dry environment-- dry being essential to avoid gumming up the bearings, and cool to avoid heat-scrambling of the magnetic bits inside.  The only real worry about old hard disks is having the supporting electronics around long enough to get at the data it contains.  Flash memory suffers from this last problem too-- obsolescence of the reading-electronic capability.  ... meh... interesting times, no?)

[Sibling Zono (anon1mat0)]

Flash memory is either NAND or NOR where the distinction is how fast you can write a block (NAND) and how fast you can read a bit (NOR), so that the former is used for USB Drives/SD cards/etc, and the latter for computer BIOS. Life depends on write cycles, and the electrical state (0/1) is determined at a practical quantum level (electrons trapped in the semiconductor). I don't know if they actually die after X years even if not used, but I'm inclined to think that they may last more than a single decade. The data stored may get corrupted (leakage) but that doesn't mean that the device itself becomes unusable.
---
If you want a more long term form of storage, why not use a laser on a hard glass (like the one used for those tridimensional portraits that you can find on malls lately). I'm sure data stored that way could last for centuries.

[Bob in a quantum-state-of-faith]

Flash memory is either NAND or NOR where the distinction is how fast you can write a block (NAND) and how fast you can read a bit (NOR), so that the former is used for USB Drives/SD cards/etc, and the latter for computer BIOS. Life depends on write cycles, and the electrical state (0/1) is determined at a practical quantum level (electrons trapped in the semiconductor). I don't know if they actually die after X years even if not used, but I'm inclined to think that they may last more than a single decade. The data stored may get corrupted (leakage) but that doesn't mean that the device itself becomes unusable.
---
If you want a more long term form of storage, why not use a laser on a hard glass (like the one used for those tridimensional portraits that you can find on malls lately). I'm sure data stored that way could last for centuries.

Cool!  Thanks.  It's been too long since I looked into these things, so I suppose I'm rather behind the curve here.   If I felt better, I'd go dig around and learn... meh.

I agree with you about the laser-etched glass, though.  Glass can conceivably last millions of years unchanged.  The only problem I could see, would be a decoding method once encoded.  I suppose you could use a set of macro instructions written in several languages... on how the microscopic data was encoded (big endian or little endian, 8 bits, 16bits, 32 and so on)

You could even etch in multiple "layers" within the glass, if you used the right sort of laser beam.  This could theoretically let you have thousands of layers in a given section of glass.  You'd read them back, by varying the focal point...

[Griffin NoName]

Just write in pictograms for easy reading. My garndaughter can already read simple Japanese, way behind progress in English, although in that respect she is fine on phonetics.

[Bob in a quantum-state-of-faith]

The individual has always had to struggle to keep from being overwhelmed by the tribe. If you try it, you will be lonely often, and sometimes frightened. But no price is too high to pay for the privilege of owning yourself.  ~ Nietzsche

[Opsa]

I like that one. I feel like that sometimes.

[Sibling DavidH]

It may sometimes be, that - given the prevailing Zeitgeist of non-Nietzschian apocoptalism - the philosopher may stumble within the ambit of his intra-personalistic ratiocinations or proto-logical formulatory deliberations and put the custard on the baby and the cat in the freezer.

Prof. Dr. Erwin Lüttschwanz