Forgive me if I am using this term AI as excessively as everyone else. I have said this before and I am going to repeat it here: most of the AI stuff that people and the industry are talking about is nothing more than a machine learning model, nothing more than a machine, a computer, a software, a system which learns an algorithm and then tries to generalize it, to widely apply an algorithm. All these models’ noise is nothing more than a bunch of numbers representing an algorithm, and believe it or not, many of them have been around for decades.
It is all about the mathematics, the computation, the very original idea of computers as a means of calculating things which are beyond human reach in terms of speed and scale. It is a mathematical system highly relying on linear algebra and calculus which you probably learned in school, well, perhaps you scored really well on it, but not necessarily learned it or were able to ever apply it in real life and solve real-life problems, two different stories. The reality is that without linear algebra, we can’t represent data or model computations, and without calculus, we can’t efficiently optimize most models. The point is, what you deal with typically is nothing more than machines learning patterns; from the LLM with its magical next token generation to the hurricane’s next move prediction. Not that I am downplaying the importance of the machine, not at all, it is very fascinating, but that is not necessarily artificial intelligence. Defining intelligence would be controversial and I am pretty sure we would eventually agree to disagree about it. But I only want to simplify the noise of AI for you and I am going to say see it as another software, an operating system.
I like the OS analogy because an OS is giving a broad range of applications to the end-user, it is an environment to then establish other utilities. Imagine machines with this new generation of software, called AI, are capable of providing the ground for you to do a bunch of things and even have your own very customized utilities. The only difference, just like any software, is the user interface. Windows normalized the ability of a user to interact via mouse instead of boring monochrome terminals, as much as Python changed the boring world of assembly. Software has just gone much more intuitive in terms of interacting with the user, and I see AI as a new software with the most comfortable way of interaction: chatting. To push this further, think of it like an automatic transmission in a car versus a manual stick shift; the engine, the fuel, and the physics of moving forward are identical, but the interface changed so anyone could drive without stalling. Or think of it like a translation dictionary; the words were always sitting in the book, but someone finally built an index that flips to the right page the second you look at it.
Of course you see I am intentionally overly simplifying AI. You know AI is not just about chatbots, of course, but the point is to get out of this and look at it from a distance. We discovered how mathematics can change our lives, we always knew it, we always pursued it during the entire human history, we always recklessly used math to solve our problems. The only difference is that then we invented computers, and today we popularized the accessibility of the most complicated algorithms which are only possible to use just because the computational capacity increased drastically. This literally puts us in a true time machine because we no longer need to spend two years to research on an anti-body protein or months on analyzing cosmos photos. So yeah, we are saving a lot of time, so then that is pretty much a time machine only capable of going forward, and if you are sitting still, now you are the past.
We realized how fascinating it is to exploit the math, to shape it as a trained algorithm, to establish logics and procedures simply beyond human capacity. Just a simple machine, and human has learned how to make machines which are making things easier, pure physics law, right? That is why if I am forced to use the word intelligent, I would like to describe what we typically call AI as a data-driven intelligence system, or my favorite short version, a smart algorithm. So next time you think about AI, this model or that model, think about an organized way of utilizing mathematical functions to run a procedure to do something, to decide, to predict. I mean, you already heard a million times, computers do everything with numbers, right? Simply imagine AI is a universe where each element has a distinct representative number. We are talking about multi-dimensional numberings so you can define any object within this universe, more precise than a human can ever do. And again, what brings this to reality and useful for everyone? That is computers, and computers operate with software, with instructions. AI is nothing more than a new intuitive software.
Artificial intelligence is a huge and highly abstract and generalized term. It is like thinking if you know a lot about physiology, let us say all of it ever written, then you decide to say you know how the human body works, and AI in this analogy is the planet Earth. So basically what we describe as AI today is nothing more than a perfect physiologist, I hope that makes sense. I am saying that AI is a big word to use when talking about face recognition, coding, and chatting with a chatbot or even letting it book the next trip for your honeymoon. I believe we are 50% on the way to reaching real intelligent systems where I can label them as artificially intelligent, but for now we are just enjoying one of the biggest accomplishments of the computer science, and that is the close to full utilization of mathematical algorithms and how to teach them to machines.