This is not another post about AI, but its opening is related.
Among my friends who are skeptical of AI, there’s a common refrain about large language models like ChatGPT: Such models could never actually have real understanding of anything; they’re merely doing pattern recognition, and generating possible ways to continue the pattern they’ve been fed.
In reply, I point out that human infants begin to build their understanding of the world by doing exactly the same thing: They start with a disconnected series of images taken in by their eyes, and they slowly find patterns in that data. Eventually, those patterns allow them to predict which actions they take are most likely to lead to the satisfaction of their desires. Such patterns build in layers, until they learn causes and effects that may be many steps removed from each other: They become proficient at manipulating a complicated world. So is that real understanding, or just pattern recognition? I believe there is no sharp boundary between the two.
In fact, readers of this blog may notice that my usual starting point, my first principle, is that the world contains patterns of matter and energy. I claim that any truth we may hope to find can only come from observing those patterns: When we find a consistent pattern, we call it truth. Pattern recognition must therefore be the foundation of all knowledge.
Its higher development is also the main advantage our species has over the competition, and the basis for both the social and intellectual levels of complexity. Faced with this importance, we should be able to agree that pattern recognition is at the very core of what it means to be human.
There are of course alternative ways for thinking about the core of humanity. Recent posts on Staggering Implications have focused on beauty, and suggest that our subjective impressions of beauty can never be fully explained by merely mechanical and evolutionary accounts of the world. In that case, an alternative core of humanity must be proposed, giving us a type of dualism based on the separation of objective and subjective realms. But such dualism goes against all of the patterns I recognize.
In fact, I’d like to suggest that beauty, rather than being a separate entity, is intimately related to pattern recognition itself.
Let’s step back for a moment and consider three famous transcendentals; truth, goodness, and beauty. I’ve already noted above that truth is identical with pattern recognition. And in my October post, I proposed that goodness, by definition, is that which fits our fundamental pattern of energized perseverance. Will the third transcendental also conform to a pattern-related definition so easily? Maybe not quite.
I think beauty is a little bit resistant to definition, because in some ways it’s a catch-all category: We tend to call things beautiful when they give us pleasure but we don’t know why. In such cases, I believe thermodynamics can add useful structure to our language (because thermodynamics is what gives structure to our world).
First, in order to analyze thermodynamic goodness, we can map it onto Cartesian coordinates consisting of complexity and order. The complexity axis should be familiar by now, as it proceeds from biological to social to intellectual levels. The order axis, on the other hand, was discussed in my paired principles post, and progresses from chaos to order. Using these two axes, we can look for goodness at each level of complexity, and always find it near the middle of the order axis (at a balance between that level’s chaotic principle and that level’s ordering principle). For example, at the genetic level, good biology makes use of sexual recombination to add a bit of chaos to otherwise faithful replication. And at the tribal level, charisma in leadership can add a useful measure of chaos to normally rigid tradition. (If you want more examples, just ask!)
With such thermodynamic zones of goodness in mind, I think we’re in a better position to explain why we find certain things beautiful.
A woman, for example, may seem beautiful not just because all of her biological functions appear healthy, but also because her features contain the right balance between familiar patterns and less common elements (which may suggest some distance from ourselves).
On the social level, groups of people can achieve beauty through music and dance, in both cases by combining measured unpredictability or abstraction with more conventional patterns that we may recognize from everyday life. Similarly, architecture may achieve beauty through a balance between clean, efficient functioning on the one hand, and a perhaps bold departure from normal patterns (especially if that adds some functional possibility) on the other.
I meant these as social examples, but the arts have unavoidable intellectual components as well.
On the intellectual level, we must balance the ordering principles of logic and reason with the chaotic principles of creativity and emotion: We may arrive at truth through pure pattern recognition, but we are also responsible for creating and extending the highest patterns ourselves. We put our creativity to use in that manner because of the instinctual, evolved desires we refer to as curiosity and ambition. Fundamentally, we get pleasure from finding harmony in a pattern, we get curiosity when confronted with new patterns, and we feel ambition to understand, extend, and manipulate patterns. These are survival skills for us as social and intellectual animals.
So in that case, what can we say about intellectual beauty? We can say it consists in a balance of reason and emotion; in patterns that give us pleasure for the harmony and possibility they hint at, and for the curiosity and ambition they may arouse.
Now, with that survey of beauty at hand, let’s return to the biological level of complexity for a moment, and consider our first needs of water, food, and shelter. It should be obvious why certain landscape and agricultural features hint at goodness for us, and therefore appear beautiful. But that’s not the whole story: We can also find beauty in landscapes that offer unusual patterns, higher complexity, or the hint of possibility for new ways that we can live or things that we can build.
The higher complexity, especially, harbors special appeal. We may call a sunset spectacular, not because it hints at any immediate survival advantage or triggers any ambition in us, but simply because it appeals to our innate curiosity, and reassures us that we always have more to learn about the world.
But there’s also more direct thermodynamic aspect of sunset to consider. As the planet spins us towards darkness, the last rays of light may appear precious in contrast to the lack of energy elsewhere. Our goal is energized perseverance, and on some level, we’ll always feel respect for the source of that energy. Its beauty may be highlighted by the struggle to make its tortuous way through the evening atmosphere to reach our eyes.