The six epochs of intelligence evolution. I came across the definition of these six stages in the futurist Ray Kurzweil’s latest book, The Singularity is Nearer. Per Kurzweil, each of the six stages of intelligence builds on the complexity of the information processing in the preceding stage, so the third epoch depends on the second one happening and the second epoch depends on the first.
The First Epoch of intelligence began a few hundred thousand years after the Big Bang when negatively-charged electrons began circling stably around a positively-charged core of protons and neutrons. These stable structures are called atoms and include hydrogen, helium and carbon – all of the types of atoms that are enumerated by the periodic table of elements. Billions of years later, these atoms began combining together to form molecules, which have infinite complexity and so can store elaborate information. Carbon, for example, is a particularly useful atomic building block because carbon atoms can form long, stable chains while also forming strong bonds with other types of atoms like hydrogen, nitrogen and oxygen.
It’s wild that we live in a universe where even this level of stable information complexity is possible. If physical constants like gravity, electron mass or proton charge were off by the tiniest, slightest amount of where they are in our universe, atoms and molecules wouldn’t be able to form. I’m not religious, but I am always in awe and wonder when I ponder how tremendously improbable it is that the universe came to be with these just-right “Goldilocks” physical constants in place. There’s a mystery behind this that no one may ever know.
Even wilder is that, seemingly through random chance and factors like lightning strikes, UV radiation from the sun and delivery by meteorites, stable chains of carbon and other atoms eventually happened to form on earth. Through countless further random interactions over long stretches of time, a particularly complex information-storage molecule called ribonucleic acid (RNA for short) began to form. If RNA sounds familiar, it’s probably because of DNA, which evolved later. DNA has two strands that are very stable (in their famous double-helix structure) while RNA only has a single strand. This single-strand structure makes RNA less stable, but it also allows RNA to actively perform work, such as smoothing along chemical reactions — even catalyzing the formation of other RNA molecules!
RNA molecules aren’t alive, but once they could self-replicate, RNA did allow for the process of natural selection to occur. More efficient RNA molecules were more likely to survive and propagate, gradually allowing RNA — and the information stored within these molecules — to become more and more complex. Again, super wild, mind-blowing stuff, but it likely took millions of years and vast numbers of failed attempts and dead ends for RNA to become a self-replicating information store.
So, to recap, the first of six epochs of intelligence involved the development of atoms and molecules, including extremely complex information-storing molecules like RNA. The Second Epoch of intelligence began several billion years ago with the emergence of life on Earth. Very quickly, scientists believe life emerged from RNA:
By RNA developing the ability to synthesize proteins (complex molecules that can form an extremely wide variety of different kinds of work)
RNA and proteins happening to become trapped in spontaneously formed fatty-acid bubbles to form “protocells”
Natural selection favoring protocells with increasingly stable containment structures and more stable information-storage mechanisms such as double-helix DNA
Through processes I’m not going to take the time to get into today, eventually these protocells evolved into single-celled living organisms, which themselves evolved into multi-cellular living organisms.
In the Third Epoch of intelligence, animals described by DNA formed simple nervous systems and, eventually, brains, allowing animals to store and process information in real-time. Brains provided animals with strong evolutionary advantages and so they developed more and more complexity over many millions of years.
In the Fourth Epoch, which began only a few thousand years ago — a blink of the eye in evolutionary terms — humans, with their uniquely complex brains (and their opposable thumbs!), began physically recording information outside of their brains. Through writing systems on clay tablets, papyrus and eventually modern information technology like magnetic tape and solid-state hard drives, humans augmented their internal information-storage systems with external ones. In recent decades, and particularly in just the past few years, humans devised machine-learning approaches that also enable us to not only augment our information storage but also supplement humans’ abilities to perceive and reason about information. In an accelerating and increasingly diverse range of examples in recent years, these intelligent machines have come to exceed human capabilities. It seems likely that even simply through scaling the methods we already have — such as by increasing by further orders of magnitude the number of model parameters, inference time compute and dataset size — machines will continue to surpass humans on more and more complex cognitive tasks. In only a few years, for example, scientific discovery may be led more by machines than by humans. Interesting times, indeed!
Even more interesting is that we are at this very moment beginning the Fifth Epoch, in which biological human cognition merges directly with the speed and vastness of digital technology. Instead of interacting with artificial intelligence indirectly through screens, keyboards, headphones and microphones, a small number of humans now interface directly with A.I. systems through brain-computer interfaces (BCIs). Today, with the Fifth Epoch in its infancy, BCIs are typically implanted only in severe healthcare situations such as paralysis or motor- function loss due to, say, spinal cord injury or stroke. As the Fifth Epoch accelerates over the coming decades, however, it will be possible for ordinary humans’ brains to be directly augmented by nonbiological computers that will vastly increase our memory capacity and capacity for complex, abstract thought. This is the precipice upon which we stand today and the radically different future that myself, perhaps you, and certainly many thousands of listeners to this podcast are bringing about through our A.I. model and application development. That’s trippy!
The Sixth Epoch of Ray Kurzweil’s intelligence stages is frankly something that I’m not yet able to grasp and is certainly the most speculative of his evolutionary development of intelligence theory. In the Sixth Epoch, intelligence spreads beyond Earth and begins to influence the structure and destiny of the entire universe. The basic patterns of the universe become optimized for intelligence and intelligence gains the ability to restructure matter at the most fundamental levels. Kurzweil theorizes that every particle in the universe could become part of a vast, intelligent network with physical constraints like the speed of light and quantum mechanics becoming the only constraints upon how much computation is possible within the universe. In this view, individual intelligences merge into a unified, cosmic intelligence, the distinction between biological and artificial intelligence completely disappears and consciousness becomes a universal phenomenon. Whoa. This might be completely wild conjecture but Ray Kurzweil has been remarkably prescient about the pace of the development of A.I. over recent decades so perhaps there’s something to his dramatic vision for the future of intelligence.
To recap, the six epochs of Kurzweil’s intelligence involve:
Physics and chemistry
Biological life
Brains
Technology
The merger of human technology with intelligence that we’re beginning to embark upon today
The transformation of all inert matter in the cosmos into a substrate for consciousness and computation
The SuperDataScience podcast is available on all major podcasting platforms, YouTube, and at SuperDataScience.com.