The Evolution of Information

We’re made by information.

Technically, we’re mostly water, protein, fat, and other minerals and molecules. However, information dictates your exact physical structure and, as a result, and, to some extent, your psychological structure. That information exists in the form of DNA. That DNA contains the set of instructions that produced you. To wax ever-so-slightly poetic, you, dear reader, are information reading itself.

It’s common to think of information as the result of life, but I would argue that it’s quite the opposite. We are not the creators of information but rather the result of it. A very thin line separates the concept of life from the concept of information. Viewing information in this way, one can wonder if biological life is just one substrate among many for information to manifest itself. Through this lens, the term artificial intelligence seems to be precisely one word too long, and our species is but one among many manifestations of this phenomenon. Let’s explore the evolution not of life but of information.

From the Universe to Earth

In the beginning, the universe was mainly just a lot of hot, dense matter. There was no organization or structure to this matter; it was just a lot of particles careening around. However, one thing that can be said is that it was much more uniform than it is today. Today there are vast reaches of space with nothing in it, massive nebulae, stars, cosmic rubble, dark matter, and various other bits of disorder strewn about. Another way of putting this is the second law of thermodynamics; the universe’s entropy increases as it expands.

This is where the universe’s first information systems emerged. These were not the traditional computer-based systems we’re used to thinking about but rather the emergent patterns of particles and energy that began to form in the early universe. Over time, these patterns became more complex, eventually giving rise to stars, galaxies, and other structures. Entropy has an odd property, hinted at by Norbert Weiner, that when information (or life) is involved, it tends to decrease locally even while, in accordance with the laws of thermodynamics, it increases on the whole.

The evolution of information continued on one of these localized pockets of deceased entropy. Our little rock, third from its star, became host to simple pre-organic molecules capable of self-replication. These molecules evolved into single-celled organisms and eventually into the complex ecosystems of information our planet supports today, both organic and otherwise.

Humans: Hosts, Guardians, and Creators

One member among millions of species, a hairless ape, played a significant role not just in shaping and transmitting information but in its evolution. That’s us. We’ve developed verbal communication, allowing information to pass from one host to another, and written communication, allowing information to be transmitted across long distances to multiple hosts and even persist through the death of a host across multiple generations. It was in us that information leaped out of the physical and biological into the cognitive. No longer was information limited to the physical world, but it could now exist in the realm of thought.

Information needed vectors to gain energy more efficiently, so agricultural implements were created. It needed a way to propagate faster, so the printing press was developed, allowing for unprecedented dissemination of information to a much broader range of hosts. Information then invented the telegraph, allowing it to propagate across vast distances, though only to a single receiver. Television and radio allowed the broadcast of information to a massive number of recipients. The internet allowed for a multi-directional flow of information between hosts and efficient and persistent storage and access on a scale never before seen.

The Conversational Era and Beyond

It was this point which prompted me to write about this topic. Until the recent development of artificial intelligence capable of surfacing information conversationally (LLMs such as ChatGPT), the primary challenge and question in the internet was not one of access but of heuristics, i.e., How can I find what I’m looking for?” The question of access had nearly been solved by the printing press, broadcast media, and other forms of information transfer; the real challenge was in connecting information to the context in which it was useful or, to phrase it as Google does, in their mission statement: “to organize the world’s information and make it universally accessible and useful.”

Search made the internet, the internet as we know it — the ability to surface relevant information quickly. But if we examine the pain points currently, it’s more often that search leads to a webpage, in which some information exists that may or may not be contextually relevant for the searcher. This process is hit-and-miss, and it’s time-consuming. LLMs and other machine learning models are stripping away the intermediary steps between the searcher of information and the information itself.

As information’s evolution unfolds, we stand at a new milestone. This new phase introduces a transformation, where interactions with information resemble human communication through language. For example, the development of AI chatbots, such as ChatGPT, allows users to engage with information more naturally and conversationally, simulating human-like dialogues.

A conversational era has just begun, facilitating seamless dialogue with the vast collective knowledge of the human species embodied by language models. This advancement allows for extremely rapid access to relevant and valuable information, but it still lacks what even human toddlers possess: the ability to reason. After today’s language models, the promise of models which operate not just on pattern matching but on abstract thinking, symbolic manipulation, and logical inference. These neuro-symbolic models will eclipse today’s LLMs by orders of magnitude in scalability, accuracy, and efficiency and will represent the “second half” of the conversational era.

Following this stage, information will become embedded and contextual. An interspace will arise, where pertinent information surfaces according to context — the room one inhabits, the conversation taking place, or even the luminosity atop a mountain. All intractable space will become another vector for information to propagate.

The penultimate stage of this evolution, and incidentally the last one we’ll be much involved in, is one where artificial agents provide contextually relevant information and actively and accurately predict human behavior and thoughts. These agents would be so accurate that the required information is readily available before individuals become aware of their own needs. Furthermore, much of the information would have already been acted upon by a set of ‘user preferences’ that determine the level of control one wishes to delegate to an intelligent agent working on their behalf.

The final stage in the evolution of information? I think it’s the singularity, as John von Neumann described it. One in which information is able to remove any barriers to its proliferation in the universe through recursive improvements to its means of proliferation.

A Humbling Notion

The evolution of information has taken a fascinating path, from the early emergence of particle patterns in the universe to the complex ecosystems on our planet today. Common knowledge dictates that information is a byproduct of life. However, I submit that the reverse is true — we are not the creators of information but rather its outcome, and the role of life and our small species will play in that tale has yet to be determined. Today’s technological advancements should give us pause. I think they can reshape how we access, store, and interact with information and our understanding of that thin line separating it from life.