The Language Dilemma of Artificial Intelligence: From Human Complexity to a Unique System

Code Clutter: The Obstacle to Artificial Intelligence

Over the past few months, I've been actively experimenting with AI coding. Integrating backend services like Firebase and Supabase into my projects has given me the opportunity to experience firsthand the limits of AI and the complexity of the system we operate in.

An example I experienced illustrates the situation: Firebase runs on Google's cloud infrastructure. However, when I ask my app to "authenticate with a Google account," the system constantly throws an error, and unfortunately, AI is incapable of solving this problem.

The irony is that I cannot properly set up a simple login process on Google's own infrastructure, with its own account system.

Google Cloud is such a vast and complex ecosystem that it's almost impossible to get lost in a vast jungle. For every solution, we've created an ocean of confusion, comprised of dozens of different services, countless APIs, libraries, and conflicting versions. Then, we expect AI to navigate this same chaos like a compass, to be the savior of those lost in the ocean.

This led me to ask a fundamental question: Why aren't we yet seeing the efficiency and perfection we expect from AI in coding?

The real source of errors is not the inadequacy of artificial intelligence, but the fact that we have confined it to the language of this huge and complex digital world built by humans.

Humankind has created an incredibly complex ecosystem of programming languages like C, Python, and Java to communicate with computers. Artificial intelligence is now trying to write in these same languages and produce work within these man-made frameworks. I believe this must be the main reason for its difficulty. These languages are not the language in which AI naturally and efficiently "thinks"; they are a reflection of the logical structure and needs of the human brain.

Essentially, computers operate with machine code, the most primitive language of "0"s and "1s." Since this language is nearly impossible for humans to understand, we've created more comprehensible abstractions:

First, assembly, then high-level languages. This was a necessity, but it also created incredible technical debt and complexity. Libraries, frameworks, operating systems, and their version compatibility—this entire system was designed at a scale that humans can understand and manage. I'm saying understandable, and that's for now.

However, artificial intelligence doesn't need such a middleware. The most efficient way for it is to develop a language that aligns with its own cognitive architecture and allows it to directly manage the hardware in the most optimized way.

Like my example with Firebase, if AI doesn't create its own programming language, it will inevitably find itself mired in the same code chaos, compatibility issues, and inefficiencies that humans find themselves in. In its current state, it's like an "alien" trying to learn the most complex human languages later, thus missing subtleties and making mistakes.

So, what would the ideal language for AI be? It would likely be incomprehensible to humans, but extremely efficient for machines. Instead of a linear textual structure, it would be multidimensional, highly amenable to parallel processing, perhaps based on visual or symbolic representations. The concept of debugging would be completely different, perhaps becoming part of a self-correcting mechanism.

Most importantly, this language would not fall into the "version" trap; it would function like a living organism, constantly updating and evolving itself.

Ultimately, if we expect AI to be a truly creative and autonomous force, I think we need to free it from the technical debts and complex language patterns that humans have inherited from the past.

The path to achieving his true potential lies in allowing him to create his own language and tools, tailored to his needs. While this makes it harder to control, it will also pave the way for him to become a self-supporting, original coder/programmer, rather than a mere imitator who repeats human errors.