My experimental AI testing attempt was successful :)

[xor]

My experimental AI testing attempt was successful

{ As usual, I'm writing this with translation. }

Hardware testing platform information:
Processor: (2012 model) AMD A4-5300 2-core (with APU graphics unit),
but I was specifically looking for an AI without a GPU.
Motherboard with 16 GB RAM
8 GB USB flash drive (ext2 format)
During the Google Gemini mentoring process:
The slogan "Never give up" really motivated me

I created solution scripts suitable for TCL chemistry
I created the "run ai.sh" file,
The function of the script is to download and upload files from the internet.

Performance speed; I can't say it responded faster than me in practice,
it outputs about one word per second (depending on processor power),
but I think the average code writing speed is faster than mine

Code: [Select]

|       |-- [  17]  libggml-base.so -> libggml-base.so.0
|       |-- [  21]  libggml-base.so.0 -> libggml-base.so.0.0.0
|       |-- [   0]  libggml-base.so.0.0.0
|       |-- [   0]  libggml-cpu-alderlake.so
|       |-- [   0]  libggml-cpu-haswell.so
|       |-- [   0]  libggml-cpu-icelake.so
|       |-- [   0]  libggml-cpu-sandybridge.so
|       |-- [   0]  libggml-cpu-skylakex.so
|       |-- [   0]  libggml-cpu-sse42.so
|       `-- [   0]  libggml-cpu-x64.so
|-- [363M]  ollama
|   |-- [ 36M]  bin
|   |   `-- [ 36M]  ollama
|   `-- [327M]  lib
|       `-- [327M]  ollama
|           |-- [327M]  cuda_v12
|           |   |-- [  21]  libcublas.so.12 -> libcublas.so.12.8.4.1
|           |   |-- [111M]  libcublas.so.12.8.4.1
|           |   |-- [  23]  libcublasLt.so.12 -> libcublasLt.so.12.8.4.1
|           |   |-- [215M]  libcublasLt.so.12.8.4.1
|           |   |-- [  20]  libcudart.so.12 -> libcudart.so.12.8.90
|           |   |-- [712K]  libcudart.so.12.8.90
|           |   `-- [   0]  libggml-cuda.so
|           |-- [  17]  libggml-base.so -> libggml-base.so.0
|           |-- [  21]  libggml-base.so.0 -> libggml-base.so.0.0.0
|           |-- [   0]  libggml-base.so.0.0.0
|           |-- [   0]  libggml-cpu-alderlake.so
|           |-- [   0]  libggml-cpu-haswell.so
|           |-- [   0]  libggml-cpu-icelake.so
|           |-- [   0]  libggml-cpu-sandybridge.so
|           |-- [   0]  libggml-cpu-skylakex.so
|           |-- [   0]  libggml-cpu-sse42.so
|           `-- [   0]  libggml-cpu-x64.so
|-- [1.7G]  ollama-linux-amd64.tar.zst
|-- [ 18K]  ollama.log
|-- [ 123]  ollama_error.log
|-- [379M]  ollama_models
|   |-- [379M]  blobs
|   |   |-- [ 490]  sha256-0...
|   |   |-- [  68]  sha256-6...
|   |   |-- [ 11K]  sha256-8...
|   |   |-- [379M]  sha256-c...
|   |   `-- [1.4K]  sha256-e...
|   `-- [ 17K]  manifests
|       `-- [ 13K]  registry.ollama.ai
|           `-- [8.8K]  library
|               `-- [4.8K]  qwen2.5
|                   `-- [ 857]  0.5b
`-- [ 846]  txt.txt

 2.9G used in 15 directories, 49 files

[xor]

[xor]

Flawless Experience and Idea-Driven Development Flow

1. Goal: Flawless End-User Experience
The main goal of the project is to dismantle Linux's notoriety as "terminal-dependent" or "configuration mess" and create a distribution that a typical user can use without encountering any technical difficulties.

Focus: Usability and flawless architecture.

2. Artificial Intelligence and Engineering Level as a Tool
The use of artificial intelligence in redesigning the architecture takes the process far beyond a standard development phase. AI integrates a complex and high-level engineering discipline into the project that the human mind alone cannot conceive.

4. The Real Face of Patents: Logic, Not Code
Most patents in the software world actually protect not the lines of code, but the "how-to" of the work (flowcharts and working logic).

Observation: The same result can be achieved with C or C++. What matters is the unique path (algorithm) to that result.

5. "Idea" Becomes More Valuable Than "Open Source"
In an age where artificial intelligence has made coding child's play, the real difference is not the code itself, but the originality and strategic depth of the idea behind it.

[Rich]

Hi xor

... and create a distribution that a typical user can use without encountering any technical difficulties. ...

So, have you (or your AI) created this distribution? Where is it?

... AI integrates a complex and high-level engineering discipline into the project that the human mind alone cannot conceive. ...

AI gets trained using using the work published by others. It then presents that work as its own.
That's not "high-level engineering", that's plagiarism.

... Most patents in the software world actually protect not the lines of code, but the "how-to" of the work (flowcharts and working logic). ...

That is correct. They are called ideas. Protecting ideas that meet certain
standards is one of the purposes of the patent system.

... Observation: The same result can be achieved with C or C++. What matters is the unique path (algorithm) to that result. ...

To protect your idea, you want to make the description as broad as possible. To many
details on implementation provide paths to get around a patent.

... artificial intelligence has made coding child's play, the real difference is not the code itself, but the originality ...

Once again, code that it got from all of the data copied from the Internet and fed into
the AI database.

By the way, I see your numbering sequence goes 1, 2, 4, and 5.
Was that your idea or did the AI help you thing that up?

[xor]

It's the same situation as those who wrote books by hand before the printing press and those who write code by hand after AI!!!

As usual, I'm writing with translation.

First, it would be good to discuss the philosophy behind an event.

Here, the ranking units are sorted according to your comments.

Topic 1: I think the translation implementation is a bit overkill but it incredibly shortened the process of writing an executable script for something that had never been done before! And if AI could do what I asked the developers to do when I first joined the forum, the standard migration process I experienced with TCL could have been more perfect, but these are general Linux migration issues, not specific to TCL

Topic 2 and 3: I know how AI works from my personal work on SEO about 10 years ago, when AI wasn't very well understood. Every language has its own structure; some words are agglutinative while others are fixed expressions. It's a linguistic problem.

I was in the process of developing a sample spreadsheet algorithm to analyze recurring themes in a book; this algorithm would give the percentages of derivatives of A being added to B and other possible additions. Working on a structure that could generate usable word structures in this way also meant decision percentages. This, without realizing it, led me to work on the concept of artificial intelligence and decision-making, that is, working on a machine with high-level thinking capabilities. Many similar technological revolutions have occurred throughout history. One person invents the telephone, another develops a similar device without even knowing what they're doing; but the real story is that the person who finds the right market succeeds, and that's how human history progresses!

Point 4 is actually a critique. Technically, ideas have become more important than open source code. Previously, closed source code was a structure that expressed the algorithm, the workflow logic of a process. Now, it's not about how everyone learns to code, but how they do it or how they get it done; considering that many large startup companies have brilliant ideas but don't have time to learn programming, yet can find experts to create the program they want, the winner is not the programmer, but the person who created the idea.

Point 5: Artificial intelligence is a search tool; before the internet there were books and libraries, before books there were embossed tablets, before tablets there were cave inscriptions. An AI search system is an advanced search function that prioritizes necessary content while eliminating unnecessary content. It generates information by compiling scientifically proven articles and accepting those with an accuracy rate of 90% or higher, then ranks them in a list of recommendations based on these rates. Ultimately, each search is programmed in a more natural language within this search and result generation algorithm.

All the programming languages ​​we've described are, at the most basic level, assembly languages ​​that provide direct access to hardware. And every programming language developer actually runs the same assembly language with different layers of abstraction; depending on the situation, much more assembly language content is run in the background; two lines of code can run 100 or 1000 lines of assembly code. It's similar to pressing a piano key; you press a single key and hear the sound resulting from the harmonious operation of countless hardware components within it. And in the background, at almost machine language 1010 level, unpredictable vulnerabilities arise due to errors caused by previously undetectable conflicts, requiring a significant amount of time for human verification. And since data analysis and system updates at this level require time and effort beyond physical human labor, artificial intelligence is an inevitable reality at some point. No matter how perfectly you write the program, the problem stems from an error at the machine language level, even beyond a lower ASM layer, that is undetectable by human hands.

Therefore, even an operating system developed by millions of programmers continues to experience current problems due to issues beyond human control.

Because the real problem lies in hardware that is not 100% open source. Even if it were theoretically 100% open source, something is missing. If it weren't missing, everything would be perfect

And one final criticism: Does artificial intelligence serve humanity? First of all, there is a huge struggle between commercial giants. Technically, no one turns down the opportunity to use expensive hardware that no one could ever own for free or with limited access via the internet.

It's the same situation as those who wrote books by hand before the printing press and those who write code by hand after AI!!!

[xor]

There's a logical connection between ignorance and happiness

News about a 3% performance increase with a single line of code change
https://www.phoronix.com/news/Intel-ANV-BTI-Prefetch

For those who can see the big picture; this single command makes thousands or more ASM-level changes in the background!

There's a logical connection between ignorance and happiness

[xor]

Dreams are beautiful. Reality isn't quite like that.

If you enable two high-level AIs to communicate with each other, one AI will find the error written by the other. (AIs cannot find their own errors.) When they start writing and generating correct code, at some point something of a high engineering level emerges;

However, the information part within the code is somehow erased by the AI ​​while carrying the message, and from that point on, the AI ​​deviates from its original purpose and starts writing different code. To maintain contextual stability, it is crucial to carry the information part within the code as uninterpreted information as possible; the information part within the code is quite important for this purpose, and establishing strict rules while generating code seriously solves this problem.

Regarding the future of AI, I have only come to this conclusion: no AI can produce correct inferences on its own. Because the philosophy of knowledge is a process that branches out and grows through the collision of knowledge.

Code: [Select]

{
  "Code_Purpose": "Minimalist x86 kernel entry point that prints 'Hello OS' to the screen.",
  "Creation_Date": " ",
  "Version": "1.0.0-alpha",
  "Target_OS": "Bare Metal",
  "Version_Range": "x86_32 (i386)",
  "Author_Info": " ",
  "Assistant_Info": {
    "AI_Model_Family": " ",
    "Specific_Version": " ",
    "Architecture": " "
  },
  "License": " ",
  "Usage": " ",
  "Parameters": " ",
  "Dependencies": " ",
  "Notes": " ",
  "Infrastructure_Requirement": " ",
  "Known_Issues": " ",
  "Test_Status": " ",
  "Limitations": " ",
  "Precautions": " ",
  "Change_Log": " ",
  "Future_Developments": " ",
  "References": " "
}

[xor]

(As usual, I'm writing this as a translation )

The main point here is:
Given how valuable our time is,
and how slowly technology has developed until now,
artificial intelligence is actually like a debugger.

It finds the error, searches for a solution to the problem in the library,
and suggests a better solution.

And for this, it needs bug reports.

I think artificial intelligence cannot be properly understood unless it is approached from this perspective.

This is my philosophical approach and my dream.

Actually, it doesn't produce a technical result,

because one day people will be able to write their dreams on a screen,
and I want an operating system that does this, this, this, and this,
and it should be so small that it loads into the motherboard's BIOS,
no need for HDD, SSD, or USB, not even RAM,
the L1 and L2 cache areas inside the processor work faster than RAM,

if you tell someone about this dream,
they'll tell you about the dream part, not the feasible part,

but in reality, right now processor memory is almost
100 MB,
processors and motherboards that the average home user can afford,

of course, purchasing power varies from country to country,
and I'm not in one of those rich countries!

While everyone is building DDR5 systems, I'm trying to build a DDR3 system with almost second-hand parts;

At least thank you for listening to my dreams.

(As usual, I'm writing this as a translation )

[xor]

del

I pressed the wrong button, you can delete this

[nick65go]

and I want an operating system that does this, this, this, and this,
and it should be so small that it loads into the motherboard's BIOS,
no need for HDD, SSD, or USB, not even RAM,
the L1 and L2 cache areas inside the processor work faster than RAM,

bad luck, shattered good dream, https://www.phoronix.com/news/RISC-V-XIP-Being-Removed

"Introduced in Linux 5.13 back in 2021 was eXecute In Place "XIP" support for RISC-V that allows for the kernel image to be executed from ROM. The intent is on allowing the kernel to run from non-volatile storage like NOR flash that is directly addressable by the CPU and to reduce RAM usage."

[EDIT] Maybe you have a small chance, because Intel spy (not yet bankrupt) still has a MINIX OS in its CPU.

[mocore]

typical user

who defines this "typical user" ?

and how is it defined ...

what (data) is the definition based on ?

see also
https://en.wikipedia.org/wiki/Get_a_Mac

...

(AIs cannot find their own errors.)

who defines "errors" ?
and how ... etc