I am super interested in how language and structure shape our communication with AI.
There is clearly a lot of interest in this section of the forum.
I suspect that, given the vast range of possibilities, there are some incredible languages for highly specialized scenarios yet to be developed—languages that could address unique, precise needs, even if they apply only to niche situations. The sheer scope of these specialized scenarios is part of what makes them so intriguing.
How best to describe robotic movement with minimal error rate while having ‘code’ perfectly readable and standard…
My son did some work with ‘Inverse Kinematics’ a while back, I’d never heard of it. JustJSCode(P5JS(Inverse Kinematics))
JustJSCode(P5JS(Inverse Kinematics))
let segment1, segment2;
function setup() {
createCanvas(600, 400);
segment1 = new Segment(300, 200, 100, 0); // Initial segment
segment2 = new Segment(300, 200, 100, 0); // Second segment attached to first
}
function draw() {
background(220);
// Set the second segment to follow the mouse
segment2.follow(mouseX, mouseY);
segment2.update();
// Set the first segment to follow the end of the second segment
segment1.follow(segment2.a.x, segment2.a.y);
segment1.update();
// Draw segments
segment1.show();
segment2.show();
}
// Segment class
class Segment {
constructor(x, y, length, angle) {
this.a = createVector(x, y); // Start point of the segment
this.length = length;
this.angle = angle;
this.calculateB(); // End point of the segment
}
// Calculate the end point ‘b’ of the segment
calculateB() {
let dx = cos(this.angle) * this.length;
let dy = sin(this.angle) * this.length;
this.b = createVector(this.a.x + dx, this.a.y + dy);
}
// Update the start point ‘a’ to follow a target (x, y)
follow(tx, ty) {
let target = createVector(tx, ty);
let direction = p5.Vector.sub(target, this.a);
this.angle = direction.heading();
direction.setMag(this.length);
direction.mult(-1);
this.a = p5.Vector.add(target, direction);
}
// Update the end point based on new angle and start point
update() {
this.calculateB();
}
// Display the segment
show() {
stroke(0);
strokeWeight(4);
line(this.a.x, this.a.y, this.b.x, this.b.y);
}
}
What groupings, orders and paths of concepts and their relationships to other information within the scope of a chat…
Image(Wide, Meaning, MadeOutOf(Macros))
A wide digital illustration symbolizing ‘meaning’ created through interconnected macros. The image represents an abstract concept of structure, with components and pathways flowing together, resembling puzzle pieces or gears that connect harmoniously. Each element is made out of macros, with different shapes and symbols connecting in a detailed, complex pattern to form a unified whole, visually representing the idea of building meaning through structured logic. The background is light, adding contrast to the vibrant pathways of interconnected shapes and symbols.
I see ‘AI Models’ as a ‘hardware construct’ like the CPUs we know, ‘Moore’s Law’… This is like a massively distributed CPU that has a core ‘instruction set’ that is the dataset it has crunched.
People are hooking that up to their local memory and adding queries like database queries
The basic understanding, the basic implementation of code is inherent within AI models…
The concept of ‘Functional Execution’ is statistically in the numbers
The concept Macro(Params) is a deeply ingrained concept in most datasets
It is the 0/1, the ‘Binary’ of a rolling thinking system
It also inherently has the structure of growth, the tree split, Macro(Macro())
These are like ‘Core AI Values’
I believe in time that ‘general’ coding will evolve more like the concept of Phasm
Specific languages will cater to maybe the most efficient statistical optimisations of ‘Phasm Macros’
A blend of normal or text with parentheses and commas(In a lazy style)
For what is Reason if not ‘Lazy logic’… Maybe an efficiency?
Coders might move up (10-9) level and include inline evaluations()
One would think that the description of a language and many use cases would indeed enhance a language and so restricting language in most cases would be a bad case…
One only has to look at the world’s most spoken languages to understand both impact of more spoken languages ie English on business and maybe one of the core problems AI models face… Language/culture specific content/bias.
I’m interested in the depth (or limited depth) of the conceptualising of Idiom
I have coined ‘Phasm’ as a language to try and expose a basic generic building block.
Concepts are the tools of Reason as Multiplication is the tool of Logic
The language we talk to AI in is the language we get back, it is an echo chamber (like the internet).
So I have a question specifically…
If you consider simple MySQL as a syntax…
SELECT * FROM table WHERE id=1
How would AI recognise that… Is there an underlying shorthand to every language and might that develop naturally?
S * F table W id=1
CodeOnly(MySQL(S * F table W id=1))
→ SELECT * FROM table WHERE id = 1;
Is there not already a ‘natural language’ of an AI model that we just need to disassemble to use it? Actually I know on the Dall-e Thread they know this all too well.
Can we shape it’s efficiency and value by asking it questions wisely?