A new kind of way to look at ai

Research
21

Also sandboxing is you friend for mutations here’s a clip from my updater

import hashlib
from secure_container import Sandbox

class HotPatchSystem:
def init(self):
self.sandbox = Sandbox(mem_limit=“100MB”)

def apply_patch(self, code: str) -> bool:
    if not self._validate(code):
        return False
    
    result = self.sandbox.execute(code)
    return result.exit_code == 0

def _validate(self, code: str) -> bool:
    banned = ["exec(", "subprocess", "rm -rf"]
    return not any(cmd in code for cmd in banned)
22

Musings on Discrete Neurons, Cycles, and a Pulley I Can’t Pull Yet

Arggggh!
I have to order tools to pull that rear pulley off the mower. So… it’s going to be a couple of days.

A Turn Toward Python

About Python: oh my. When I took a class on C, I told myself I’d never use an interpreted language again. I had come from the land of BASIC, so C felt like a revelation. But here I am, standing at the edge of a new paradigm—Python.

I can see now that I’ll have to learn it. And apparently, there’s a thing called object-oriented Python? Never heard of it before this journey.

Neural Nets and the Discrete Brain

I haven’t worked with neural networks directly, but I’ve read about them and had some long, wandering chats on the subject.

So here’s a question, asked a bit blindly:

Is a neuron typically just an object with a float value that encodes connectivity?

If that’s true, then let me pose an alternative vision:

Suppose we go discrete.

Imagine a neuron is a binary segment (bseg)—length ranging from 2 bits to n bits. Length 1 is a bit dull: 1 → 1, 0 → 0—simple identities. But now imagine the neuron is 4 bits long. That gives us:

  • 8 elements per cycle
  • 2 distinct 8-element cycles for a given parity reference (e.g., b0)
  • A complete rotation space of binary behavior

And here they are:

Binary Representation of Two Cycles

Cycle 1: ( 7 12 5 15 8 3 10 0 )

Decimal Binary (4-bit)
7 0111
12 1100
5 0101
15 1111
8 1000
3 0011
10 1010
0 0000

Combined Binary String:
01111100010111111000001110100000

Cycle 2: ( 9 13 11 14 6 2 4 1 )

Decimal Binary (4-bit)
9 1001
13 1101
11 1011
14 1110
6 0110
2 0010
4 0100
1 0001

Combined Binary String:
10011101101111100110001001000001

Discrete Neurons: A Thought, A Cycle

Let’s imagine each of these 4-bit structures as a neuron, cycling through its 8 internal states. Each one has directionality—clockwise or counterclockwise. We can assign meaning to these rotations, state transitions, or interactions.

So what could we build with such a system?

What if the position of the neuron in its cycle is its “state,” and interactions are mathematical—deterministic, symbolic, and fully discrete?

A neuron that spins in place, but in motion across a symbolic landscape.

I welcome feedback on this concept of a discrete neuron. Not a float, but a fixed-length bit segment with a structured cycle—like an information gyroscope. Multiply it by a billion, entangle them, simulate interaction…

Now that’s the stuff of my daydreams.
It helps me fall asleep with hope.

A final thought:
It takes a village to build real AGI. Maybe a garden of binary neurons is a good place to start.

23

I’m not sure about what you mentioned there the neuroevolution emulator takes the ai string alters it slightly different on each layer then if it’s an improvement on the current string a sandboxed instance opens to test the string before application so it’s not a real neural object it’s more just a thought process improver I have uploaded a project code that shows the interactions of it about line 340
The-basics-every-ai-should-have-/The basics of ai at main · madmoo-Pi/The-basics-every-ai-should-have- · GitHub

24

Okay more ideas. So a neural network is 4 layers? If there are 4 layers and each layer is a bit then a cycle of the whole 4-bit of the 4 layers might be interesting. Does that make sense? Maybe I just don’t understand yet.

25

I chose 4 layers as I’m working towards edge deployment you can have any number of layers (hardware dependant) normally on a cloud is actual hardware (too rich for me ) so emulation is the way forwards , this emulates natural growth and development to optimise its own thoughts (speed , accuracy and efficiency)

26

Alright, my bad, I’ll have this stuff moved or deleted.
I think I am train wrecking your thread.
What do I do flag it?

27

Don’t delete bud we’re all here to share and learn bud chat away I just uploaded a base for development a secure ethical ai base if any researchers or builders wanna trial and build on

29

Hope this helps:

1 Like
30

This topic was automatically closed 12 hours after the last reply. New replies are no longer allowed.