Deconstructing AI -
a gift for all the junior devs with imposter syndrome
Today we are going to deconstruct the notion that you or any other individual needs AI to succeed, here are some common thought patterns that a diverse amount of individuals may have:
If I don’t use ai I’ll somehow fall behind! (or would i?)
I wouldn't be the same programmer without ai (or would i?)
AI makes me write code more efficiently. (or does it?)
News flash you've been lied to.
Have you ever wondered how the first ChatGPT model was produced? Did they somehow use a time machine to create a rip though the fabric of reality and prompt an o3 model in the future, no, it was produced with a little bit of human ingenuity and the good old-fashioned human brain, truth be told you have the most powerful computer that open AI is so desperately trying to replicate and its right in your head, It belongs to you, not in some buzzing computer cluster, that used to be a storage warehouse for pets.com. let’s begin by reviewing these notion's and why its so easy to fall into from a marketing level.
Companies have to lie (sort of)
One vital thing to understand is the incentive behind lying and or exaggerating a product such as an LLM. These companies have to lie as a means of survival, especially since LLM's as a technology are still very much in its infancy, AI companies such as OpenAI , google and sonnet, all have an incentive to exaggerate the effectiveness of their products, not only to you but also to their shareholders, all for the benefit and growth of a company.
Now understanding this, we can begin to see why these companies make so many remarkable claims about their product with little to no evidence. They’re simply in a position that requires them to do so for their self interest.
Its simply marketing, marketing is an effective strategy used for decades to control the perception around your company and its products.
Systems Level Thinking
Refuse Perfection
“There is no minimum requirement for me to strive for perfection — therefore, perfection lies far outside my scope.”
I believe, one of the primary beliefs that keep people constrained to this attachment to ai, is the false notion that ai somehow leads to a more perfect, or effective product. Lets first shatter the illusion of achievable perfection.
Let’s think about this first from a systems level perspective .
A decision tree graph is a node based tree graph that is used to represent a chain of outcomes and decisions leading to a final outcome. For every decision, there is a new set of sequential decisions, the amount of decisions per node grows exponentially. You cannot make a perfect decision on top of every other previous decision in the tree even if you tried to, simply doing so is statically impossible.
This means, that in order to produce a perfect solution (like your sorting algorithm), you would have to first traverse the entire decision tree, doing so will take more than several lifetimes. The brutal reality is that all of human technology falls victim to this constraint. This is why we have minimum requirement's, a minimum requirement is the same reason why I own and use my car despite it being no where near perfection, Its minimum requirement is that it gets me from a to z in a safe and timely manner. We as human beings have the tendency to over optimize or analyze for many edge cases. Additionally, this i why "vibe coding" can be so difficult, for every piece of code you ordinary LLM outputs, it will also find and suggest many more improvements that which in themselves have other improvements that can be made upon and thus you are left with an exponentially large chain of possible decisions and improvements. Perhaps this is what the first popular tweet regarding vibe coding was referring to as "the exponentials".
Systems level thinking is the same reason why, Im being cautious to over analyze my writing, It's because the minimum requirement is the transfer of concepts via literature. It keeps me in scope, something ai cannot do.
Top comments (0)