"I have a dream" – Martin Luther King Jr, from his famous speech in 1963.

Sure, we all have a dream. But not everyone with a dream executes what he or she wants. When it comes to execution, we are often paralyzed. I remember when I was 16. I was truly confused and fascinated by AI. Seeing the progress, I immediately knew that AI should be my field in the future. But the problem is that the path of AI is not known yet.

We have been told that a certain level of AI, like AGI, will come in a few years with full expectation, a claim I'm going to pull apart in the next piece, because the mechanics of how these models actually learn tell a very different story. Undoubtedly, we also see the negative impacts that AI brings to society. Massive job destruction and ethical dilemmas that seemed impossible to solve were everyday, and I've spent time trying to name and classify exactly how AI fails when it does, which is its own piece entirely. Those problems still exist today.

I believe there is a path that humanity has not walked on yet. I feel like it's a mystery that can't be solved no matter what we do, something like a cipher, a mysterious code, sometimes tied to how civilizations hid meaning in history.

Saying this, I also can't deny the challenges that we are facing due to AI. As I mentioned above, we can see the struggles people are facing every day due to AI, and I don't want to sugarcoat that. I believe in a balanced approach, and this is the lens every piece here runs through, not just this one.

It is like standing up on a stage. If you are on the edge of the stage where only one of your legs is touching the stage, the odds of falling go up fast. But if you stand in the middle, the probability of you falling will reduce significantly.

We will look at the positive and negative aspects AI brings to society clearly in these articles. Some of the articles include my perceptions and theories, while the others might include scientific truth. I'll flag which parts are established fact and which are my own theory as we go.

As we said earlier, a cipher is a code. To this date, computers only use 0s and 1s to process things. Programming languages such as Python have become more usable every day due to AI model training. I always wonder, what if there is a new method that we have ever imagined to train AI models? Don't get me wrong, the framework provided by Python and pipelines from other companies like Hugging Face is truly crucial; I can't deny their impact.

Models work by converting every little piece of data from the dataset into tokens. These tokens are again converted into vectors by the process called vectorization. That is why people need to identify the words "understand" and "predict" in the AI training process. We are not at a level of making AI understand, but we are training them to be very intelligent predictors. Before you say anything, yes, you are right that there are reasoning models. In fact, they are called LRM (Large Reasoning Models), so reasoning algorithms are being built.

But personally, I don't believe that this is enough. We need a new set of programming that could simplify training or fine-tuning models. Maybe a new language? I don't know that yet. When this actually arrives, and what happens if it goes wrong, are the two questions the rest of this space is built to chase.

But till we figure it out, it will be a cipher. Till we figure it out, it will be under the Cipher.