LLMs and Programming in the first days of 2024

The year 2023 marked a significant milestone in the realm of artificial intelligence, particularly with the advent of ChatGPT and the subsequent utilization of Large Language Models (LLMs) that operate locally. As an experienced programmer, I found these tools to be a game-changer, accelerating my ability to write code and saving me from wasting mental energy on mundane aspects of programming.

LLMs have proven to be an invaluable resource, particularly in writing high-level code in Python. They have helped me understand when to use them and when their use would only slow me down. However, it's important to note that LLMs are like Wikipedia or video courses on YouTube: they assist those with the will, ability, and discipline, but they offer marginal benefits to those who have fallen behind.

In the field of programming, LLMs are akin to "stupid savants" who possess vast knowledge. They may not take us beyond the paths of knowledge, but they can often lift us from our absolute ignorance to the point where we know enough to move forward on our own.

Despite their limitations, LLMs represent the greatest achievement of AI to date. They are capable of writing programs that they have not seen in that exact form, showing a certain ability to blend different ideas that appeared in the training set with a certain frequency.

However, when it comes to system programming, LLMs often fall short. They are not yet capable of complex reasoning, and their output in such scenarios is often poor. This is particularly evident when dealing with more complex models or algorithmic programming.

Despite these shortcomings, stating that LLMs are not helpful for programmers is extremely rash. They can significantly speed up the ability to produce code by taking over mundane tasks, allowing programmers to focus on things that matter.

In conclusion, the advent of LLMs has undeniably revolutionized the field of programming. They have become an indispensable tool that has the potential to save programmers from the drudgery of mundane tasks and accelerate their ability to write code. However, they are not without their limitations, particularly in the realm of system programming. As the technology continues to evolve, it is hoped that these limitations will be overcome, further enhancing the utility of LLMs in the field of programming.

The original article: http://antirez.com/news/140