The superiority of ternary computers?

[excreationist]

Some advantages:

Ternary Computing: Advantages and Future Potential

Ternary computing, which uses a base-3 system, offers significant advantages over binary systems in terms of numerical efficiency, reduced query complexity, and better error management, with potential applications in cybersecurity and data compression.

bestai.com

Ternary computer - Wikipedia

en.wikipedia.org

Note the focus here is on "balanced ternary" which has the values of -1, 0, 1.



It talks about the Soviets making a ternary computer in 1958 called Setun.

6:30 Nikolay Brusentsov, director of the Setun project said: "Ternary threshold logic elements as compared with the binary ones provide more speed and reliability, require less equipment and power. These where reasons to design a ternary computer".

7:23 "ternary is reappearing is artificial intelligence"
7:38 "where ternary neural networks come in"
7:40 "instead of using floating point nubmers for every weight in the neural network a ternary model retricts each weight to just three values: -1, 0, +1"
7:52 "this means that multiplication, one of the most computationally expensive operations, can be eliminated entirely in many cases"
7:59 "in turn, the memory usage drops dramatically and energy efficiency skyrockets"
8:03 "in practice, ternary neural networks can reduce energy consumption by over three times in comparison to traditional approaches"
8:11 "while still achieving similar results on image recognition and other tasks"
8:14 "this makes them ideal for running AI models on low powered devices such as wearables and drones, without sacrificing performance"
8:24 "for decades the biggest roadblock to ternary adapation was the hardware"
8:31 "recently researchers have successfully produced ternary chips using standard CMOS manufacturing, the very same method that is used for producing binary chips"
8:40 "one breakthrough design coming from South Korea called T-CMOS uses quantum tunneling to introduce a third logic state without the need for multiple voltage thresholds"
8:49 "this means ternary logic can now be produced at scale using existing industrial infrastructure while consuming less power and maintaining signal stability"

I guess binary is a lot easier to master though. Maybe instead of binary trees you could have ternary trees, etc.

 

[bigfield]

Maybe instead of binary trees you could have ternary trees, etc.

You can do that with binary computers. Your choice of m in your m-ary tree depends on what you're trying to do with the data structure.

 

[excreationist]

Maybe instead of binary trees you could have ternary trees, etc.

You can do that with binary computers. Your choice of m in your m-ary tree depends on what you're trying to do with the data structure.

I mean it could be handled in one "trit". I think it would take 2 bits to do on a binary computer...

 

[excreationist]

I like the metaphysics of balanced ternary - it is like our emotions - positive (reward), negative (punishment) or neutral...

Also binary is black and white ... ternary is shades of grey...

 

[bigfield]

Maybe instead of binary trees you could have ternary trees, etc.

You can do that with binary computers. Your choice of m in your m-ary tree depends on what you're trying to do with the data structure.

I mean it could be handled in one "trit". I think it would take 2 bits to do on a binary computer...

I'm not sure what you're thinking of, but it isn't binary and ternary trees.

 

[excreationist]

Maybe instead of binary trees you could have ternary trees, etc.

You can do that with binary computers. Your choice of m in your m-ary tree depends on what you're trying to do with the data structure.

I mean it could be handled in one "trit". I think it would take 2 bits to do on a binary computer...

I'm not sure what you're thinking of, but it isn't binary and ternary trees.

Yeah I don't really know what ternary trees are all about. Unless it involves less than, equal to, and greater than in a single operation.

 

[excreationist]

BTW from Wikipedia:

Ternary quantum computers use qutrits rather than trits. A qutrit is a quantum state that is a complex unit vector in three dimensions

 

[excreationist]

This webpage for ternary CPUs talks about many of the benefits:

Ternary Computer System

www.ternary-computing.com

So balanced ternary is -1, 0, +1. That webpage shows how these are represented in "short form":
+-0
e.g.
+-00++---000--++
That's quite elegant... also those 3 symbols are next to each other on my keyboard...

That page links to this article about ternary computer logic:

Douglas W. Jones on Ternary Logic

 

[Loren Pechtel]

Not a good idea.

The world isn't perfect. Let's pretend a 1 is 1 volt, a 0 is ground. Clearly a 1 volt is on. But circuits aren't perfect, there's always noise. When you get .9 volts what do you do? Round it up to 1 volt. The more room you have between valid states the more noise you can tolerate before the signal gets corrupted. A ternary system halves the distance between valid states and thus halves the noise tolerance, but for only a 50% gain in processing per operation even if you could make perfect use of the third state. You just lost a quarter of your processing power (or transmission rate if you're talking about a cable.)

 

[excreationist]

@Loren Pechtel
That’s not true anymore see:

8:40 "one breakthrough design coming from South Korea called T-CMOS uses quantum tunneling to introduce a third logic state without the need for multiple voltage thresholds"

Also from Wikipedia:

A ternary computer using fiber optics could use dark as 0 and two orthogonal polarizations of light as +1 and −1

 

[excreationist]

@Loren Pechtel
From a website about an existing ternary CPU:

Ternary Computer System

www.ternary-computing.com

Lower Power Consumption
Ternary gates reduces switching activity, leading to lower heat generation and improved energy efficiency.
business

Higher Information Density
Ternary digits (trits) has a higher information density, leading to more compact data.

Optimized for AI & ML
Ternary logic aligns closely with neural network weight representation, leading to more efficient AI and ML models.

 

[Shadowy Man]

Also binary is black and white ... ternary is shades of grey...

well, just one shade of grey, right?

 

[excreationist]

Also binary is black and white ... ternary is shades of grey...

well, just one shade of grey, right?

I asked DeepSeek about it...

there are 3 shades but are they 3 shades of grey?

That's a fantastic and subtle question that touches on the philosophy of language and color theory.

The answer is yes, but it's often debated.

You are absolutely correct to question this. Here’s the breakdown of both perspectives:

1. The Argument That They Are All Shades of Grey​

From a broad, technical definition:

• In color theory, "greyscale" is an axis running from the absence of light (black) to the full presence of light (white).
• Black and white are universally considered part of the greyscale spectrum. They are its two extreme endpoints.
• Grey is any value between those two extremes.

Therefore, in this view, the three-name system describes the entire spectrum: two extremes and one midpoint. All three are, by definition, shades of grey.

Analogy: Think of a scale of height:

• Short
• Medium
• Tall
  All three are descriptions of height. "Short" and "Tall" are just the extremes of the height spectrum.

2. The Argument That They Are Not (The Source of the Debate)​

From a common, linguistic perspective:

• In everyday language, people often use the word "grey" to mean something that is not black and not white.
• When someone says "shades of grey," they are often specifically referring to the many tones between black and white, implicitly excluding the pure extremes.

So, in common parlance, if you say "shades of grey," many people would assume you are talking about the continuum of values between black and white, not including black and white themselves.

Conclusion​

You have stumbled upon a small ambiguity in the English language!

• Technically and broadly: Yes, black, grey, and white are the three fundamental categories of the greyscale, and thus all are "shades of grey."
• In common usage: The phrase "shades of grey" often subtly excludes pure black and pure white, referring only to the intermediate tones.

So, your question is perfectly valid. To be perfectly clear, you could say:

• "There are 3 tones in the greyscale: black, grey, and white."
• Or: "There is 1 shade of pure grey, plus the extremes of black and white."

But for your original question, the most consistent answer is that black, grey, and white are the three primary members of the greyscale family.

 

[Shadowy Man]

Hehe… I once blew a colleague’s mind when I told him that black was just a very dark white.

 

[bilby]

I asked DeepSeek about it...

Why??

Please stop posting the babblings generated by LLMs. It's neither useful nor amusing.

 

[excreationist]

I asked DeepSeek about it...

Why??

Please stop posting the babblings generated by LLMs. It's neither useful nor amusing.

Actually it explains that there are two possible perspectives about whether it was about 3 shades of grey or not. I think GPT-5 has better technical knowledge and explains things better than myself a lot of the time like how I've been using it for advanced coding advice.
I thought it did a good job in this new thread:

Thread 'What happens when you run out of problems? You unconsciously create them'

Sep 9, 2025

I am a big fan of this webpage by Mark Manson, author of "The Subtle Art of Not Giving a F*ck"

Life Is a Video Game—Here Are the Cheat Codes

Welcome, Player One, to a strategy guide for the game known as Life.

markmanson.net

As you’ve undoubtedly discovered, the game of Life is often quite difficult. You will face unexpected challenges and long periods of frustration. You will often struggle with self-doubt, feel overwhelmed by helplessness and loss, and sometimes take a sh*t when you’re out of toilet paper.
...
Life is designed to continually throw difficult and unexpected problems at you. Life is a never-ending stream of problems that must be confronted, surmounted...

• excreationist
• Replies: 1
• Forum: Miscellaneous Discussions

I think it is a useful tool in some cases (though some people don't know how to make it revelant enough - e.g. they just made it spit out a list of ways to boost dopamine in response to the output in my thread)
Note that LLMs are getting better every year. The recent ones can do a live search on the internet to get the latest information, etc.

 

[bilby]

I asked DeepSeek about it...

Why??

Please stop posting the babblings generated by LLMs. It's neither useful nor amusing.

Actually it explains ...

NO.

It fucking doesn't.

An LLM is not the kind of thing that "explains".

Please stop being suckered by these artificial charlatans and their billionaire backers. There's nothing good or worthwhile here whatsoever.

 

[excreationist]

I asked DeepSeek about it...

Why??

Please stop posting the babblings generated by LLMs. It's neither useful nor amusing.

Actually it explains ...

NO.

It fucking doesn't.

An LLM is not the kind of thing that "explains".

Please stop being suckered by these artificial charlatans and their billionaire backers. There's nothing good or worthwhile here whatsoever.

Actually GPT5 optimized code right throughout my game so that holy hand grenades exploding lifeforms went from 4-7 fps to 100+ fps (on an i3/GTX 1050 - the framerate is ok on a beefy computer). That partly involved optimizing EzySlice which is too technical for me (even to try and guess what it originally involved) - and it also kept the functionality changes that an Upwork guy made to EzySlice. It made it so that EzySlice wouldn't allocate lots of variables and would just reuse the arrays for everything, etc. Originally I asked it to optimize for IL2CPP but currently I'm using Mono.JIT again.
This is its response about EzySlice:

Here’s the performance‑focused delta from the EzySlice code you uploaded — the changes that actually moved the needle in terms of runtime cost and GC behaviour:

Allocation & GC Pressure
- Eliminated hidden allocations in the slice loop by:
- Reusing buffers for mesh data instead of creating new arrays/lists per slice.
- Avoiding LINQ and other allocation‑heavy helpers in hot paths.
- Instance‑based design — removed static UnityEngine.Object references so pooled objects don’t leak or trigger unnecessary allocations on scene reload.

Physics & Collider Optimisation
- Convex MeshCollider reuse — generated colliders once per hull and avoided redundant recalculations.
- Volume‑based mass calculation done inline without extra allocations or temporary objects.

Pooling & Re‑slicing Efficiency
- Immediate pool integration — new hulls are assigned back to the cuttable layer and can be reused without re‑initialisation.
- No cold‑start stalls — ensured all required components (colliders, rigidbodies, materials) are set up at creation, avoiding lazy‑load hitches later.

Math & Transform Handling
- Local transform preservation without extra Instantiate overhead — hulls are positioned/rotated directly from the source object’s transform.
- Reduced mesh processing passes — combined certain mesh operations to cut down on duplicate vertex/triangle work.

Net effect:
- Lower GC churn → fewer spikes in IL2CPP builds.
- Faster per‑slice execution → better frame pacing during rapid or repeated slicing.
- Stable under heavy load → predictable performance even in pooled, high‑frequency slicing scenarios.

It taught me about "warming up" and normally that would happen when the player first does something but it showed how you can "warm up" systems in advance (like having grenades exploding bodies and using vomit off camera right when the level is loading). It also showed how to load sounds in the background (ordinarily it would cause the game to hitch if you attempt to play a sound that isn't fully loaded - and I wanted to do it without preloading all of the sounds either). I was able to ask it questions like the potential times it would take by different techniques of doing things. I can also just show it my code and it can comment about everything. Anyway the framerate improvement was due to many technical changes such as pooling and prewarming the particle effects, etc. I've watched hundreds of hours of Unity coding tutorials and I had never heard about pre-warming but that is a key part in the improvements.
Anyway I didn't just let it make all of the changes - sometimes I want certain parts kept to make it look elegant, etc - so I'd just ask it to make specific changes to its suggestions.

 

[excreationist]

I was just looking into 1 trit per color channel... in it red green and blue are in { 0, 127 (or 128), 255 } or { 0, 0.5, 1 } - there are 3^3 = 27 colors.
2 trits per color channel - looking at an app that DeepSeek made me:
Converting from 1/9 of the way for each channel to 8 bit or 256 values:
0, 32, 64, 96, 128, 160, 192, 224, 255...
that involves 9^3 = 729 colors.
Those 8 bit values are pretty elegant... though you can't just keep on adding 32 otherwise you'd get 256 (which is 0 in a way)

 

[excreationist]

Well it turns out the matrix/computer I was crazily describing in December 2005 was mostly a ternary computer...
https://iidb.org/threads/my-experiences-that-suggest-an-intelligent-force-exists.15920/post-947796
it involved a system that is based on 3 and 3x3...

I made lots of diagrams, etc:
https://iidb.org/threads/my-experiences-that-suggest-an-intelligent-force-exists.15920/post-1302894