The Mental Meanderings of Jas1199_

The Mental Meanderings of Jas1199_

Note: Any mistakes in spelling is My fault along with using a very simple text editor with no spell checker.


First, I am a avid Science Fiction fan.
I began reading SiFi during the 1960's. At 72, I listen to SiFi short stories on YouTube regularly during the day.

Second, I love open source projects. Be they be in software, hardware or some combination thereof.

Third. The idea of RISC-V cpu's and Linux (or other open source software not even conceived of today) along with a 512 bit RISC-V CPU presents an interesting idea for the future of computing.

Meanderings:
001. UNI-CODE at 512 bits per character. The NEW BIT.

002. The 1k byte (1024) 2ea of the New 512bit being equal to the NEW BIT, creating the NEW BYTE, 2k byte, 4k byte, 8k byte, etc.

003. This is to facilitate computing for the next century plus.

004. Consider this, we have today (2026) UNI-CODE for a large portion of the languages on this planet. If we can codify a new UNI-CODE based upon the 512 bit character (The NEW BIT) we could have the beginnings of a new computer base that could conceivably carry us into the next few centuries.

005. These few ideas could create a new base of computing for the next several 100 years. With the NEW BIT (??the V-BIT??!!!) we could go forward to new level of computing. HOWEVER, we must start TODAY, as this shall require a massive move from the current computing society. NO backward compatability! The closest thing you could do is to take parts of the current UNI-CODE and re-map it into the NEW BIT. Do not have even one existing UNI-CODE directly transfer to the NEW UNI-CODE. Include as many coded languages as possible.

006. We need to start this new philosophy of computing. Begin to write the NEW standards. [I offer my apologies for using "NEW" excessively. Someone else will have to create a better term. I do not have any idea for future wording.] Have someone who knows a little computing but not so engrossed as to be constrained by current computing models.

007. Currently we can manufacture silicone chips at 4 microns and are striving to build 2 micron vertical layers. Manufacturing will change. What it becomes is beyond my thinking.
I grew up with transistors, knew vacuum tubes. TV's with crt's. My family did not have a color TV until about 1975. And that was used and I repaired it using Heath Kit test equipment that I built so I could repair the TV. Aviation eletronics training from Spartan School of Aeronautics in Tulsa Oklahoma.

008. And what about "optical CPU's"
Consider using multiple wavelengths of light internally. We have red, green, yellow, and blue LED's cyrrently in production.

009. If we just use the four to create a base system of:
A. Red LED on, off
B. Green LED on,off
C. Yellow LED on, off
D. Blue LED on, off
We could expand this into LED's on.
A+B, A+C, A+D, B+C, B+D, C+D and A+B+C+D aka know as "white".
I don't have the mathematics to list all the combinations this would create, but I am sure that someone who is good at mathematics can give you an answer.

010. My basic idea is to use the ideas of computing but move it beyond base two, on, off computing.

011. THINK! Move forward using a different model only remotely related to today's computing.

Any scoffing of these ideas should consider the humble orbital satellite. Originally anticipated in early Science Fiction stories. Ever hear of the Laser? I remember reading about how to build one using a red plastic rod and a camera flash bulb. Science Fiction today is tomorrow's present. How to build an analog computer it was in either Popular Electronics or Popular Science. From Sputnick through the "cold war", into computing. I have lived through it. A very interesting life.


Note:
My use of "888" at the end my documents is what I use to indicate the end of the document and/or a part of multiple modifications of said document.
888

If a collection of 512 bits is now a bit, what are the 512 individual entities now called?

Ironically a ternary system like what was explored in the 60s would accelerate AI inference in a big way.

beers said:

If a collection of 512 bits is now a bit, what are the 512 individual entities now called?

Ironically a ternary system like what was explored in the 60s would accelerate AI inference in a big way.

Click to expand...

Yes I have difficulty coming up with new terminologies.
As I said, someone much smarter than I am needs to create new terms. I remember bits, bytes, words and double words. Did they have something named larger than double words? I don't know.
The biggest problem that I see is to find someone who is not to ingrained in the current computer technology.
Someone needs to start thinking about these ideas when they are about 7 or 8 years old.
Do they start with the current RISC-V cpu's and hope that by the time of their teens someone will be building 512bit cpu's?
Memory sizes will be astronomical!
Say, 30-50 years from now computers will be very interesting.

Working towards a "Comander Data" being built.
Thank you for reading and replying to my Meanderings.

What do you think of the current optical cpu's being considered by IBM and others?
Thank you .
Jas1199


888

Kinda depends what “bits” you mean. Address width like we mean with 64-bit CPUs? Vector width like AVX-512? Memory bus width? ALU width? Instruction encoding width?

If you remember 32-bit addressing topping out at about 4 GiB, a theoretical 64-bit byte-address space is 16 EiB, which is 4,294,967,296 times larger. Current CPUs don't even implement all 64 address bits because there isn't a practical need yet.

Some of the other big bit counts refer to completely different things. DDR memory channels are commonly 64 bits wide, while GPUs can have 384- or 512-bit memory buses. AVX-512 uses 512-bit vector registers. Those are all "bits," but they're measuring very different parts of the architecture.

I'm also interested in the RISC-V space, especially as newer implementations standardize around RVA23 and performance continues to improve. I just don't see address width itself being the thing that pushes CPUs to 128 or 512 bits anytime soon. We have an enormous amount of unused address space left in 64-bit architectures already.

Optical CPUs sound interesting, much like the advancements we're seeing in networking, but there's also massive complexity in shrinking individual optical devices down toward current transistor scales.

I could see aggregate buses benefiting sooner, though. Things like CPU-to-DRAM links, chiplet interconnects, or PCIe/CXL-style fabrics could potentially use optical pipelining where bandwidth and signal integrity become limiting factors. Hollow-core optical links are interesting there because the signal can propagate substantially faster through air than through conventional silica fiber, although at microscopic scales the packaging and alignment would probably be extremely delicate.

I have a feeling we'll also keep seeing alternative architectures, including things like ternary computing, periodically resurface where they offer some useful way to scale sideways instead of simply making conventional binary CPUs wider.

Where do you feasibly see these systems going in the future, and what parts of the architecture would you actually augment first?

The "anytime soon" is what I am concerned about. The new terms for 512 wide bits and above is where I am unable to think of. New words and the definition of those words is where I fail.

If I recall correctly, the 4bit chip was developed in the early '70's.
When I was at Spartan (1973-1974) I was the 3rd hand owner of a TI calculator. I purchased it for $ 75. And all it did was add, subtract, multiply and divide. It also had 1 constant function. The red LED's were of the type that the "dome" magnified the LED to a point where it was readable. By 1979 you could buy a pen that had those same capabilities for less than $ 10. Some places were even giving away calculators free with a purchase of something.

The point that I am trying to make is that I personally am unable to think of how computing will be in 50 to 100 years. The RISC-V cpu currently has a 512bit preliminary specification written.

Coming from a background of analog radio electronics , I was amazed when cpu speeds topped 1gHz. At those frequencies stray capacitance and stray inductance do strange things.

I personally am unable to conceive of where computing will be in 100 years. What I am asking for is to make the current 512bit specification the "New Bit" the new binary starting point.
Have future specifications written with this in mind.
I am asking for a "leap" in terms and specifications WITHOUT going from "A", to "B", to "C", to "D", until you get to "L" or greater. We Must make that leap. We Must jump from "D" to "L" !
We Must write the specifications now. Not as a continuation, but a leap. Who can think that way? They probably have not even been born yet. Maybe the child or grandchild of todays freshman Computer Science college student.

What I am thinking of, is exactly how you are asking questions regarding current computer technology. You are thinking A, B, C without the leap. We MUST make that leap. How, I do not know. Maybe no one knows. All I can do is present the possibility of a new way of thinking. No A, B, C... thinking.

Remember the 1MB memory limit of 16-bit cpu's? At the time it was thought that was enough. Look where we are now. We think we will never use all the addressable space of the 64-bit chip. Thinking about the current 64-bit chip is now the same limited thinking that we had about the 16-bit chips. Technology and future manufacturing processes will be the determining factor for the new "0" starting point.

Just a F.Y.I check out the YouTube video:
Japan's new chip factory that terrifies TSMC
rubenfoto1 and polarisyn channels.

2 nanometer vertical transistors


Thank you for the conversation. I enjoy thinking of the future of computing and where it will go.

Jas1199


888

It's an interesting thought exercise, but I feel like you're ending up dragging 64x the amount of data (between UTF-8 and a 512-bit unicode implementation) across the bus/ram/storage for no real gain, for decades. That's a big penalty for the moment until infrastructure scales much more substantially.

Also old to new nuance has a lot of hurdles, if you remember recently even legacy BIOS to full native UEFI was a pain and a lot of implementations needed CSM in order to have compatibility with both standards while UEFI implementations matured more. You can kinda see the same with RISC-V being significantly newer than x86, although it also has other nuance like less significant corporate backing to accelerate development financially.

The exponential growth of each increasing bit makes the pool a lot larger than it seems. For example the entire 128 bit IPv6 space is to said to have more combinations than there are estimated atoms on the surface of the earth.

Interested to see where it goes, once we start hitting petabytes of RAM in server instances then I imagine there'd be a push for larger addressing and integer registers, assuming we even still use the Von Neumann model.

OK.

I am sure that I am not able to state my idea properly.
I am analog RF electronics.
What I would like to see, however, is to find a way to jump ahead into a newer way of thinking.
To write the new specifications before the hardware locks us down. Struggling from current hardware to future hardware AND bring a new way of thinking of software along with it. At the same time.

There must be some way of thinking differently and to jump ahead without slogging through the current computer technology. I see your point regarding a larger bit buss width .

If there is any way to create something that has a lot of information bundled into a new logic "1". How, I do not know. I am still struggling to say what my mind keeps bouncing around. To find a way.

I started on this journey to find a way to get the UNI-CODE mapped into a single 512bit unit. The current UNI-CODE uses variable number of bits to define a character. I just wanted a fixed width unit for ALL characters. This would also allow new characters to be imported. Then a language like Japanese could include all of the characters in their language to be used on computers. 1 unit equal to 1 character of the world's languages. I do not even know how many UNI-CODE characters there are. Japanese, Chinese, Hebrew, Turkish, Arabic, all have different UNI-CODE types.
All that exploded into thinking waaaay beyond my education.

I am just an old man dreaming of future computing .

Thanks for the conversation. The way you speak you must be deeply involved in the computer science field.

Thank you
Jas1199

PS. I am bouncing around an idea of using a PIC24F16KA102 micro controller along with something like a 7-10 inch CYD to create a school lunch box sized modern-ish XT. 16-bit DOS and ELKS Linux on one machine. I have much more research to do.
I have e-mailed the company requesting more information on how to do build it. They were enthusiastic about the idea. I am awaiting a reply.
Meanwhile I continue to do research as best I can. [ Chronic back pain along with hypoglycemia has me in and out of mental confusion ]



888

Jas1199 said:

I am asking for a "leap" in terms and specifications WITHOUT going from "A", to "B", to "C", to "D", until you get to "L" or greater. We Must make that leap. We Must jump from "D" to "L" !

Click to expand...

Jas1199 said:

What I would like to see, however, is to find a way to jump ahead into a newer way of thinking.
To write the new specifications before the hardware locks us down. Struggling from current hardware to future hardware AND bring a new way of thinking of software along with it. At the same time.

There must be some way of thinking differently and to jump ahead without slogging through the current computer technology. I see your point regarding a larger bit buss width .

Click to expand...

At the end of the day, it all comes down to cost. Both in the R&D and cost of production. Unless you're going to be able to find funding from government entities in perpetuity, you'll need to develop something that the market wants in an increment fashion to 1. stay in business, 2. fund your development to improve your product.

Yes, there times in history where the leaps you're describing occurs, but the demand for such a product needs to exist in the first place.

I have just watched a YouTube video by GlassTTY posted about 13 hours ago entitled: A New Old Newbury Newbear 77-68

It was discussing the early days of British computing.
About circa 1977 you bought a set of plans to build a computer.

Paper, no hardware.

Things can be done by individuals. I just want to have someone try.

As I have stated earlier, the variable length of UNI-CODE started me down this "rabbit hole".
I just wanted someone to create a new UNI-CODE so that each character would be represented by a 512-bit code. With no existing code being used. This would allow new characters to be included. You could possibly included Egyptian hieroglyphics as new characters. Any new language would be able to be placed in machine code for computers.

We need to be able to input information. The hardware will come. Manufacturing technology in ways that may not even be able to be conceived of today.

The future will be very interesting.

Thank you
Jas1199


888

Thinking more about your approach, what you're describing actually starts looking a lot like AI tokenization.

Instead of limiting each 512-bit value to a single character, you could assign values to individual characters, entire words, common phrases, or even whole sentences. The value itself would essentially be an identifier, with both systems agreeing on what each value represents.

A 512-bit value has 2^512 possible values, or:

Thirteen quinquagintillion, four hundred seven novemquadragintillion, eight hundred seven octoquadragintillion, nine hundred twenty-nine septenquadragintillion, nine hundred forty-two sexquadragintillion, five hundred ninety-seven quinquadragintillion, ninety-nine quattuorquadragintillion, five hundred seventy-four tresquadragintillion, twenty-four duoquadragintillion, nine hundred ninety-eight unquadragintillion, two hundred five quadragintillion, eight hundred forty-six novemtrigintillion, one hundred twenty-seven octotrigintillion, four hundred seventy-nine septentrigintillion, three hundred sixty-five sextrigintillion, eight hundred twenty quintrigintillion, five hundred ninety-two quattuortrigintillion, three hundred ninety-three trestrigintillion, three hundred seventy-seven duotrigintillion, seven hundred twenty-three untrigintillion, five hundred sixty-one trigintillion, four hundred forty-three novemvigintillion, seven hundred twenty-one octovigintillion, seven hundred sixty-four septenvigintillion, thirty sexvigintillion, seventy-three quinvigintillion, five hundred forty-six quattuorvigintillion, nine hundred seventy-six tresvigintillion, eight hundred one duovigintillion, eight hundred seventy-four unvigintillion, two hundred ninety-eight vigintillion, one hundred sixty-six novemdecillion, nine hundred three octodecillion, four hundred twenty-seven septendecillion, six hundred ninety sexdecillion, thirty-one quindecillion, eight hundred fifty-eight quattuordecillion, one hundred eighty-six tredecillion, four hundred eighty-six duodecillion, fifty undecillion, eight hundred fifty-three decillion, seven hundred fifty-three nonillion, eight hundred eighty-two octillion, eight hundred eleven septillion, nine hundred forty-six sextillion, five hundred sixty-nine quintillion, nine hundred forty-six quadrillion, four hundred thirty-three trillion, six hundred forty-nine billion, six million, eighty-four thousand, ninety-six.

Or about 1.34 × 10^154 possible values.

You could assign values to every Unicode character, every word in every existing dictionary and language, then start assigning common phrases and sentences, and you would still have effectively the entire namespace left. Saying "more than 99.99% would remain" actually dramatically understates how little of the available space you would have consumed.

At that point, though, it isn't really Unicode anymore. You've essentially created an absolutely gigantic tokenizer or dictionary encoding system, which is interestingly pretty close to how AI models already represent language internally with tokens.

That may actually be closer to what you were getting at with "a lot of information bundled into a new logic 1" than simply making existing Unicode characters 512 bits wide.

Hoooorrrraayyyy!!!!

You explained it much better than I ever could.

Loved your naming the super long number. I don't even know those names.

I like the idea of whole words or phrases.

Thank you.
You were able to explain something even that I had difficulty explaining.

Now all we need is to get a group together to begin thinking about how to codify all the known languages. I do have a minor concern. It usually shows up when you have different people translating the same phrase. It shows up when the adjacent phrases change the meaning on the stated phrase.

In regards to AI, have you listened to text to verbal words in an audio book?
Examples:
Bow, as in the Bow of a ship.
Bow, as in a Bow and arrow.
Also
Lead, as he lead a group of people across the street.
Lead, as a lead acid battery.

There are many, many more in AI voiced audio books.

AI has not the "knowledge" of content.

Your last paragraph states what I was thinking. I just did not know how to state it.

Jas1199



888