Artificial intelligence is a hot topic right now, a real game changer for sure. For understandable reasons. But how does it work? Usually, intelligence as we knew it required a wet computer such as our brain to work properly. But somehow, we managed to reverse engineer it. To be fair, mother nature did it much better than we have. An average brain needs just ~ 20 watts of energy to function. Which is much less than 700 to 1000 watts used by a single Nvidia B200 chip. And these chips are often combined in clusters with thousands of other such chips. To better understand how researchers reverse engineered intelligence, one should first turn to their local college and get a PhD in biology. I’m just kidding, a PhD in computer science is just fine. “Stop it! Just tell us!”, I hear you say. Well, let’s get down to business then. We don’t need to know all the details for this one anyway. As you will soon learn, neither did scientists.
The Zoo of Minds
There are all kinds of intelligences in nature. Some forms of smart can be found where you would least expect them. Wood Wide Web refers to a complex mycelium network that connects tree roots under the forest floor. Scientists discovered that this decentralized web of mycorrhizal fungi exhibits intelligent behaviors such as problem solving and sharing of information. Just few feet above, you can find another example — ants. An ant colony can build complex structures and even wage wars against other ant colonies. Similar to the ‘wood wide web’, the whole colony acts as a single, collective intelligence. Did you know that octopus has 9 brains? One central brain located in its head, and an additional brain in each of its eight tentacles. Octopuses, just like dolphins, crows, elephants, chimpanzees, and a few other animal species are remarkably intelligent. Some of these animals can use tools, while others, such as dolphins and elephants, can even display deep empathy.
Few of these examples are quite different from the human brain — the golden standard for intelligence: different hardware, different architecture, but the final result is the same — intelligent behavior. The important point to consider here is the fact that each of these organisms lives in a completely different sensory universe. They do not perceive their world the way we do. How organisms experience their environment depends on their senses. In 1934, biologist Jakob von Uexküll coined the term Umwelt — the idea that every organism has a completely different, self-centered subjective universe that is based on its sensory organs. A bat probably experiences its world as a 3D topological map of bouncing sound waves. A tick can only smell sweat and feel heat, which is pretty much its whole world. Even as human beings, we can only experience a small fraction of reality. For example, our eyes can detect less than 1% of electromagnetic spectrum — we call that small slice, ‘a visible spectrum’. From the evolutionary standpoint, this is all we ever needed to survive.
The High-Level Heist
So what exactly is intelligence? We don’t exactly know. There are many definitions: the capacity for reasoning, learning, logic, abstraction, problem solving, planning, critical thinking, adaptation, the list goes on and on. The one type of intelligence we are most intimately familiar with is that of Homo Sapiens. Encyclopedia Britannica defines human intelligence as a “mental quality that consists of the abilities to learn from experience, adapt to new situations, understand and handle abstract concepts, and use knowledge to manipulate one’s environment.” It’s hard to compress intelligence in a single short definition. But it is easy to spot intelligent behavior.
So how did scientists manage to reverse-engineer intelligence? That’s a great question. One would assume that they copied all the individual parts of the human brain, or at least some of the more important parts. But as it stands out, that would be too complicated. For one, the human brain is the most complex system we know of, and we still don’t even know what most of the parts even do. So instead of literally stealing the ‘hardware’, researchers simply stole the ‘software’. This is called the functionalist approach — instead of copying the “how it works”, they copied the “what it does”.
We did something similar with airplanes. And when I say we, I mean them, because, if I’m being honest, I didn’t help at all. But back to the subject. Imagine wanting to build an artificial bird. If you look at the real birds, you’d think that your artificial bird would need mechanical feathers in order to fly. But feathers aren’t easy to make. In fact, feathers are small masterpieces of nature, intricately complex and difficult to produce. But what if you don’t even need feathers? What if all you need is a bit of math and physics? Once you learn that the air moving over the wings generates lift, you’ll have the main ingredient for your artificial bird, and you’ll realize that feathers are entirely optional.
That’s more or less how researchers created artificial intelligence — they skipped biology, the whole squishy business, and just copied the way thinking appears to work on the surface. They copied things such as attention and memory instead of literally copying how neurons and neurotransmitters work (they did that too, though, to a point).
The Emergence Insurgence
As it turns out, intelligence is a type of an emergent property —
something that arises from the way many parts of a complex system work together to bring about something that’s not just greater, but also substantially different than the sum of its parts.
Let me illustrate the concept. Consider an ant colony. Each ant by itself isn’t very bright — all it can do is follow chemical trails left by other ants. But the whole colony of ants can achieve some pretty remarkable things. Another example. You might have seen a large flock of birds flying around in the countryside. Looking from afar, the flock comes together and forms a shape that may appear as one singular organism. Each bird is doing its own thing, but together they produce something that no individual bird intended. A large school of fish is another, very similar example. Wetness is another easy to understand instance of an emergent property. A single molecule of water isn’t wet, but billions of water molecules together create wetness. Traffic jams, economy, consciousness, and even life itself are all examples of properties emerging from complex systems.
As you can see, complexity isn’t just a concept, it’s a phenomenon. It’s worth noting that people often confuse the term ‘complexity’ with the word ‘complicated’. Complexity refers to structure; complicated refers to high difficulty. Complex systems are unpredictable, whereas complicated systems are predictable if you know the rules. Complex systems are self-organizing. Complicated systems are generally designed and controlled. Brains are complex; jet engines are complicated.
Emergence is a property that a complex system has, but which the individual parts of that system do not have — no single neuron in your brain is conscious, but all of them together somehow are.
There are few other things about emergent systems that are worth mentioning here — these are sort of general emergent properties found in most complex systems.
Non-Linearity and Unpredictability: In complex systems, small changes can lead to massive, disproportionate effects (the “butterfly effect”), making long-term prediction nearly impossible despite knowing the underlying rules.
Self-Organization: Systems like ecosystems, neural networks, and economies organize themselves into higher order without a central controller, driven by feedback loops.
Weak vs. Strong Emergence: Researchers distinguish between Weak Emergence (behaviors that can be deduced via simulation) and Strong Emergence (behaviors that are nearly impossible to predict or reason about in retrospect).
The Gap
There are still some major gaps between human and artificial intelligence. What this high-level approach of copying the “software” still hasn’t achieved is strong reasoning capabilities as well as genuine creativity.
Psychologist and a Nobel Prize winner, Daniel Kahneman popularized the idea that people have 2 modes of thinking: System 1 (fast, automatic, associative, and intuitive) and System 2 (slow, analytical, deliberate, and logical).
If I ask you, “What’s 2+2?”, you instantly know it’s 4 — that’s the System 1 thinking. But if I ask,”What’s 323 x 737?”, you’ll probably need to stop and think about it for a bit before giving me the answer — that’s the System 2 mode of thinking. Current AI systems operate almost exclusively in a ‘System 1’ mode. While they can recall any information they’ve learned very fast, they lack the ability to pause for deeper reflection or critical thought. This speed is ideal for chatting about the things that are already known, but it fundamentally limits their capacity to come up with, or even invent, new things.
Researchers are currently trying to build System 2 capabilities into AI (using techniques like “Chain of Thought” or search-based reasoning architectures) so the model can pause, plan multiple steps ahead, and verify its own logic. There’s also the Grounding Problem, aka Embodied Cognition which refers to the fact that AI knows more or less everything about our world, but lacks the actual experience. This is the difference between theory and practice. I can explain every detail of riding a bike, down to the tiniest movement, but once you actually try it for yourself, you won’t be able to do it right away. AI knows that the word “apple” is mathematically closest to the word “red” and “fruit”, but you know how it actually tastes, you’re familiar with the sound of its crunch, and its weight in your hand. This lack of actual experience severely limits how deeply AI can actually understand the world.
Creativity
So, what else can we steal? If just copying how attention and memory work gave us artificial intelligence, what other high-level principles can we copy? Mother nature did a wonderful job. She gave us masterpiece after masterpiece. But she never gave us the schematics. Thankfully, she gave us mathematics. Maybe Plato was onto something with his “Forms”. He believed that forms are abstract, perfect ideas that exist independently of human minds and physical reality. He thought that the realm of forms is above reality, and even more real than the physical world. But maybe his forms are hidden inside the reality, hiding behind the veil of complexity, too intricate for our simple minds to discern. Maybe the forms he was proposing are what emerges from complexity. Maybe intelligence is one such emergent form. We’re entering the era where we don’t need to understand every atom in order to harness the power of nature’s ingenuity. We just need to recognize the emerging pattern and work from there.
We are only now realizing that there is something more powerful (or at least more complex) than just raw intelligence. And that something is creativity. Most of us aren’t that different from AI when you think about it. Just like AI, most of us just parrot what we’ve heard from others. Maybe the System 2 mode of thinking isn’t what makes us special, maybe it just demonstrates how computationally slow we are. If that’s the case, maybe “chain of thought” and similar techniques we are trying to equip AI with won’t make it any smarter. Or at least won’t make it VERY smart and creative, which is what we’re after. Maybe creativity is friends we make along the way. Probably not, though. Maybe the emergent pattern of creativity is just more complex than that of intelligence. Luckily, we now have everything we need to successfully navigate said complexity, and I think it’s only a matter of time before we solve that puzzle, too.
Originally posted on Medium by Tom Nikola
Leave a Reply