Computer Science was always about trying to replicate the brain. The internet enabled it.

Origins

For most normal human beings, the idea of a weekend usually comprises of resting, going out to have a great time, or simply drinking that bottle of whisky that was lying unconsumed for a long time. And yet for some, a good weekend is one, spent wondering about the similarities between AI (and by that definition computers in general) and the human brain. Well, I am not normal, and today happens to be a weekend, and hence a great day to rant about the supposed similarities.

But where do we start? This weekend is no exception. Many years ago, in and around 2017, in a hostel room in Central University of Rajasthan, we (Sauvik Sen, Vinay Kumar, Rohit Thakur, Sudipta Dash, Abhishek Agarwal and yours truly) would often partake in "deep thought" sessions (over genorous helpings of bhaang). Those were the days when we were getting introduced to the depths of our respective subjects (physics, chemistry and computer science). And thus, as it is perhaps a rite of passage for wannabe intellectuals, we would often talk about the deepest philosophical questions, without having much scholarly depth to the discussions, except naive foraging along the precipices of our individual disciplines, trying to find a common ground, and a resolution to the questions. One of the questions that we tried to tackle in that time, was the nature and purpose of consciousness. Sauvik was working with someone in Delhi University back then, working on some theory of gravity that was completely unintelligible to me. And if my stoned memory serves me right, in our youthful scientific arrogance, we were trying to establish a fanstastic proposition. We reasoned, if there is a theory of everything, then consciousness (which we could not define) and intelligence (which we clearly lacked) would be derivable from that mythical equation. To that mix of weird propositions, Vinay Kumar chimed in, with a quote that has since managed to haunt me: "Consciousness is Complexity". We were too stoned to reason through that statement. Years passed by, and my pursuit of science led me to another set of "deep thought" sessions, this time in a different country.

Over a beer along the banks of the river Vltava, in Prague (around 2022), there was another intellectual deep dive into the question of the nature of human consciousness. The group, comprising of colleagues from the Faculty of Mathematics and Physics at Charles University and some alumni of the same, were discussing metaphysics. In the midst of a spirited discussion on how the existing theories of Physics explained how the universe came into being, a physicist remarked that the entire story of the universe, was a story of increasing complexity. From the fundamental particles to more complex states of matter, to stars, planets and then life. Life! That is another story of increasing complexity of organic chemistry. There it was again, the mention of complexity. Two different settings, two different crowd of "scientists", the same underlying thought. The story of everything seemed to be a journey of increasing complexity. Now, I had some idea about what Vinay was yapping about all those years ago. Took me a while to see it!

Emergence of Computational Complexity

I know, I know. Computational Complexity is a dangerous term to throw around, especially in the vicinity of theoretical computer scientists. I use the term here in a specific contextual sense, and not necessarily in the way that we use the term in computer science.

It is kind of established that the first general purpose digital computer (ENIAC) was built to get the US army to calculate complex artillery firing tables. It was strictly a machine that was to be used to win the war against the Axis powers. If you think about it, in spirit, ENIAC was similar to Collosus (being worked on by folks like Turing on the other side of the pond). Both were incredible machines, that were designed and being used to win the bloody war. That was it! No wonder, the most powerful technology are often built in war time.

It was only when the war ended, that people thought about the limits of these machines. Church and Turing worked on that, while others built more powerful computers to make them useful for other "hard" problems. With time, complexity grew; of both the design of computers and that of the problems they were being designed to solve. And hence we say that the computational complexity grew. Not the right phrase to use? Well, my blog, my rules!

Slowly, computers started talking to each other, the internet emerged, a superhighway of information, regulated by protocols. And with the internet, grew humanity's shared trove of information. Information about anything became accessible and the cost of dissemination of knowledge became close to zero. If one was curious, there was a way to satiate that curiosity. Talking computers also promised parallelization of effort. The focus shifted from centralization and gatekeeping of information/problems to a much distributed nature of collaboration and information sharing. There were multiple fresh set of eyes looking at a problem, collaborating, solving it, one perspective at a time.

And thus as problems grew more complex, and people figured out how to apply the human method of collaboration to computers. They figured that instead of building one huge machine to solve the problems, they could use a cluster of computing machines to have the same results. Each machine would solve a part of the problem and then all parts would be taken together to compile the final results. Computers were now made to collaborate and hence ditributed computing grew. Computer scientists realized that, if complex problems could be broken down into parts, and if those parts were solved in parallel, then a complicated problem can be solved faster; complexity to tackle greater complexity!

This idea was implemented across different domains. One of these domains was gaming. A small company called NVIDIA started making hardware that let folks do parallal computing at scale. And that seemed to help run the graphics in games faster and more efficiently. And then in 2012, Alex Krizhevsky along with Geoff Hinton and Ilya Sutskever figured out that the math and engineering that allowed games to be run, could also be used to train artificial neural networks. And 10 years later, we had ChatGPT and subsequent consumer facing LLMs solving the most complex problems.

The history of computing, can be seen as a history of increasing complexity.

Creatio Ex Nihlio

And God said, “Let there be light,” and there was light.

Well, there is some truth to it. As far as I understand the Big Bang and everything that happened after it, first it was all energy. And then some particle physics lessons later, we had matter/antimatter and whatnot. Now, if you are not aware, I am great fan of the Theory of Everything. No, not the film, the real stuff. I am also a fan of "spheres" and "crabs" (which my Bengali ancestary can be used an explanation for). Yes, I would be confused too!

You see, back in my High School (in Bokaro), my physics teacher at school, taught us that, a sphere represents the lowest energy configuration for various physical systems. I kid you not, I was fascinated. I mean, the math was elegant. It seemed that the universe preferred a particular configuration, for logical reasons ofcourse. But the notion that the universe seemed to have a choice was just fascinating. And then, in college, an ex-girlfriend told me about "carcinisation". It is an observation made by biologists where it is seen that non-crab species tend to mutate into crab-like bodies. It fascinated me because, it was just like that observation about spheres; the universe preferring certain things. Why the fuck does this matter?

Well, here comes the stoner logic. Let's assume that everything in this universe can be explained by a Theory of Everything. And as remarked before, everything is derivable from it. For that to work, the theory must have certain constants. And those constants are why the theory can explain the sheer variety phenomenon in this universe. This is of course, very theoretical! And thus an effect of that theory is, it will have a certain "bias" (because of all of those constants). Taking a leap of imagination, those biases would sort of prefer certain configurations and "manifestations". In other words, even though a certain phenomenon can be explained by different mathematical models, different perspectives, all those models/perspectives HAVE to be derivable from the theory of everything. In other words, the core algorithm will be the same. And that algorithm can be implemented in different ways, with different optimizations with respect to the medium where that algorithm is being implemented.

Consider the problem of addition of three five-digit numbers. A kid doing mental math would do it differently than how you would have a silicon circuit do it. At the end of the day, the phenomenon (addition) is the same. So, in my naive understanding, this logic is what drives the universe's preference for spheres/crabs. So far so good?

Now, I am not a scholar of biology, so discussing matters of evolution and how it works, would be a foolish endeavor. But what I do understand and believe is that, with increasing complexity, life forms found more and more complex forms of "intelligent behavior". We believe that one of the factors that enabled humans to become the apex predators was the skill of language and tools. And mind you, it took us millions of years of evolution to have a brain that is really great and recognizing patterns, making tools and talk. And ever since, we have been trying to develop more and more powerful tools.

From nothing, we created the first tools, then made the first machines (wheels), crawled our way to the industrial revolution and then managed to build the first computers. Creatio ex nihlio.

Two strands of evolution; one natural and one artificial. The first, with the objective to be better and survival, the second, with the objective to make more "human-brain-like" tools.

Internet for evolution

Now, one thing that we have to understand is that humans grew better at surviving because of how good we got at storing, using and disseminating knowledge. From the primitive traditions of transmitting knowledge as folk songs/tales/chants to the modern systems of schooling, we realized that knowledge, if codified properly, preserved and recalled, can actually be transmitted to the younglings easily. Yes, this is ofcourse a subtle callback to Dawkins' "meme theory".

The internet with all its crap, porn and racist shit is the biggest example of Dawkins' meme theory taken to the extremes. The internet, with all of it's crappiness enables the easiest way of transmission of ideas, behaviors, or styles of every kind of human demography. The internet is like the biggest collection of human memories, experiences and thoughts. The internet, for all matters and purpose, is a living, breathing form of collective consciousness.

Now imagine using this data, to train artificial neural networks with a very well defined objective function; learn the distribution of the data. We already saw that the history of computer science was a journey through evolution. And we also established that the current form of the internet is something that the world has perhaps never seen. If we believe the theory that our genes are carriers of evolutionary memory, the memory that helps us survive, the internet too is evolutionary memory. A complex structure of information. Now, ofcourse when a machine learning algorithm learns from the data of the internet, it is understandable why it feels so "human".

Machine Learning systems are getting proficient at tasks that took humans thousands of years to master.

Bostrom was right

Historically, computers got more complex, to solve more complex problems. Currently, we are at a stage where we have got systems like ChatGPT or Gemini or Claude or the gazillion others are slowly getting better at humanity's last exam. A lot of people believe that AGI is almost here.

Over the years, we developed the hardware and the complex engineering that can allow a program to use the hardware to learn from data. We made (and continue to make) the systems better. The data of internet (and thus the collective knowledge of humanity) is what is being modelled by the latest AI systems. Humans and AI, two very different implementations and yet the same underlying phenomenon. In the last 10 years, we have managed to build machines that seem to be as "smart" as a normal human being. And to be that "smart", it needs just months of going through the entirety of human knowledge.

All those years ago, Vinay said, "consciousness is complexity". Our analysis of the history of the universe is that, it keeps getting more complex; as if trying to be conscious. Well, if the AI systems are so good already, there is no point ruling out that Nick Bostrom's theory of super-intelligence might be right after all. The point AGI comes, it will be a new step of complexity for the universe. We have never imagined the consequences of such complexity, complex machines.

Maybe, the thing is, the universe is just trying to figure itself out, with all the components just incrementally giving it more and more knowledge about itself. But the point is, we started building computers, because we wanted machines as smart as humans, as powerful as our brains. Pioneers like Vinton Cerf, Bob Kahn and Tim Berners-Lee built the final tool that actually enabled that dream.

And now, we are just a tick away from an intelligence explosion. The question is, what reality would the universe experience after that? Will we discover truths that will finally wipe us? Or will AGI be the next step in our evolutionary story. As scientists, did we play too much as God?

Now the earth was formless and empty, darkness was over the surface of the deep, and the Spirit of God was hovering over the waters.

Rajasthan to Prague to this thoughful evening in Hyderabad. Maybe God does exist. Maybe I believe in a mathematical God, something that is incomprehensible by our puny, simple brains!