World’s Not Yet Possible | Primitives, Emergence, and Interrogation

“Universes of worlds as well as worlds themselves may be built in many ways.” - Nelson Goodman, “Ways of Worldmaking”

“...the definition of a science must necessarily be progressive and provisional. Any extension of knowledge or alteration in the current opinions respecting the subject matter, may lead to a change more or less extensive in the particulars included in the science…” - John Stuart Mill, “A System of Logic”

Two concepts intersect and help us think about how we create worlds. The first is primitives. The most ubiquitous are the standard alphabet and Arabic number systems. Twenty-six letters and ten digits are used and manipulated to create all sorts of worlds. We can create an infinite number of worlds using these. The holiest of books, the songs we sing, the way we tell time and measure our years on earth; all of these artifacts were worlds created by the use of primitives. 

The sophistication of the worlds we can build rests upon our capabilities and creativity in manipulating these primitives. A poet summons the power of words, built by letters, to construct imaginative worlds colored by tones of emotion. A mathematician juggles numbers using logic to define new rules that help us engineer matter into instruments and structures - bridges, temples, and skyscrapers. Primitives exist beyond letters and numbers, the most commonly overlooked are atoms. The periodic table, for instance, includes the matter primitives for the known elements that construct our universe, currently 118.

The second concept is emergence and hierarchical closure of systems. Related to the first (primitives), but distinct, this phenomenon creates a scaffold through which knowledge and worlds can be built. Emergence happens when groups of primitives take on behavior that is not known to exist within the individual primitives themselves. To return to our letter primitives, letters themselves can never be ordered to create the narrative of a Shakespearean play. It is only under the emergence of words, imbued with meaning by a collective of people, can a captivating story and theatrical performance come to life. A world is built on this scaffold - letters used to build words, to create sentences, to construct a narrative. The emergent narrative framework now stands on words. 

Words become the building blocks that help generate new worlds. The emergence of a new system - a group of primitives, creates something we call closure. This means that one need not understand the minutiae of each individual letter, just the whole (word) and context to capture an understanding of the new system. In other words, looking at and being able to identify each letter in a paragraph does not mean I read; but, by learning to read and understanding what those words mean I level up to comprehension of a new world, which was always there, just limited by my knowledgebase. Thus, I can neglect letters individually and capture words & context exclusively to play in the new world - I “close” the layer of serial & independent identification of letters and turn on the holistic pattern capture of words and contextual meaning.

“Let us imagine we encountered a mysterious building on a newly discovered alien planet, and we wished to investigate its peculiar architecture. Even before looking for the name of its architect, the first thing that we would ask is: what kind of materials and forces would be necessary to sustain this type of structure.” - Federico Campagna, “Technic and Magic, The Reconstruction of Reality”

What is a world? The consensus reality that we operate in becomes a world. The adoption of a way of thinking or doing is a world in the broadest sense. It defines the limits of our existence - what is probable and possible. To extend the range or capabilities of our existence, we must discover new worlds, or build them. We are endowed with an imagination that can generate new worlds at will. The requirement for manifesting these worlds into our collective reality is trust that your world is worthwhile and possible. When I first read the quote by Einstein “imagination is more important than intelligence,” it didn’t make sense to me. How can intelligence not be greater than imagination I wondered? It took me many years to understand why this is so, and the punchline is that imagination is how you create new worlds - something Einstein thrived at.

Emergence in the digital world

All this leads to artificial intelligence. The current foundational models built on primitives, have created a new world to build on. We have a new world and new tools. The large language models are conceptually the product of primitives that, through scale, take on emergent behaviors. By training these models to predict the next best letter in huge data sets, they reach a point where they can infer responses to text inputs. In a recent MIT lecture Hyung Won Chung, articulates the inflection point where the scaling of data + compute leads to these emergent generative AI characteristics. While this type of phenomena might be new or seem magical for many, there is a well established precedent in science and specifically in what are termed non-equilibrium systems.

Ilya Prigogine won the Nobel Prize in Chemistry (1977) for his pioneering work in non-equilibrium thermodynamics. The simple version of what his discovery is given by a demonstration of a pot of water over a hot stove top. As water begins to boil steam rises in a random chaotic way. As the temperature keeps rising and reaches a critical threshold the steam suddenly forms a distinct shape, i.e. a pattern emerges. In this example, the system (water), is placed in an environment that puts it far from equilibrium (room temp) through heat, the eventual push of the system away from equilibrium eventually leads to the crossing of a threshold that creates a new order (emergence). This demonstrates that order can arise spontaneously in far from equilibrium systems. Nothing in the analysis of the previous states of the system would lead one to predict that a new pattern would emerge. 

The concept of the emergence in systems is fascinating, a description of this phenomenon in Prigogine’s book, “Order out of Chaos,” describes this emergent pattern, even in biological systems:

“the new constituents, introduced in small quantities, lead to a new set of reactions among the system’s components. This new set of reactions then enters into competition with the system’s previous mode of functioning. If the system is “structurally stable” as far as this intrusion is concerned, the new mode of functioning will be unable to establish itself and the “innovators” will not survive. If, however, the structural fluctuation successfully imposes itself-if, for example, the kinetics whereby the “innovators” multiply fast enough for the latter to invade the system instead of being destroyed-the whole system will adopt a new mode of functioning: its activity will be governed by a new “syntax.”

Notice how Prigogine uses the interesting terms “innovators” and “syntax”, and this was first published in 1984. Critical points within systems that bifurcate into emergent patterns is more widespread than most of us intuit. The fundamental heuristic can be distilled into another one of Prigogine’s profound musings - "...how may purely quantitative growth lead to qualitatively new choices?"

World’s not yet possible is the right frame for the zeitgeist. The key word is “yet”. Progress suffers from lock-in, as does science, when we assume that we stand on an absolute foundation of truths. The ever-changing nature of science, due to its perpetual falsification mechanism, demands that we speak with a bit of trepidation. To use words like yet, currently, perhaps, and other ambiguous terms, while frustrating, is more accurate. It also lends itself to developing an open mind. As we now discover in artificial intelligence research and development with foundational models, language is a strange and wonderful symbolic system with emergent capabilities. This isn’t the first time that we’ve discovered or realized the influence that language has on our reality.

A strong lineage of philosophers, inventors, and scientists have made contributions to conceptual models of language and reality. The shortlist of language demystifiers includes:

• Ludwig Von Humboldt’s, On Language - the idea that language and thought are deeply connected.

• Ludwig Wittgenstein’s early philosophy - the structure of language reflects the structure of reality

• Ernest Cassirer’s, Philosophy of Symbolic Forms - symbolic systems mediate our experience

• Alfred Korzybski’s, General Semantics - language influences our experience, physiology, and understanding; the map is not the territory.

The future belongs to philosophers and English majors.

If words are the gateway to new worlds, then it follows that our mastery over words becomes a super power. When was the last time you tried reading a philosophy book? If it’s been a while, give it a try. Notice how the sentences are quite lengthy? What else do you notice? They build structures of logic with dependencies. They included premises, challenge statements, and create self contained systems. I suppose J.D.’s will also do well. 

In the short time since AI foundation models have been available, one of the biggest insights has been the rediscovery that knowledge interrogated wisely yields insights. Thinking in systems and imagining novel outcomes will also be useful. We will slowly start peeling ourselves away from incrementalist thinking. Beyond thinking bigger, my other request is for us to build captivating things with AI; let us demand aesthetically pleasing tools and systems. One day we will look back at legacy software and be mortified by what we had to put up with for so many years. 

The secret to new worlds is still not globally known. If we connect the dots and fast-forward on this show, I’m afraid or elated (not sure which one) that the answer is the question. Like some weird Jeopardy show, or hostage situation, interrogating AI to understand the new worlds possible comes down to our skills in crafting questions. 

The other secret is to think about the end state, not intermediate states. Often, we fail to attempt a project or entertain an idea because the path to the end state is not clear or it’s riddled with unknowns. It turns out that our AI friends can fill in this “unknown” space quite well. This will take a while to become a ubiquitous insight as humans tend to over index on control and micro-manage projects.