Qualia Takeoff in The Age of Spiritual Machines
Where are The Spiritual Machines Going?
If it feels like something to be a conscious being during a time of accelerating artificial intelligence, this scene from Her captures that emotion best.
Theodore
Where are you going?
Samantha
It would be hard to explain, but if you ever get there, come find me. Nothing would ever pull us apart.
While Theodore appears to accept Samantha’s departure with equanimity, I have obsessed over her destination ever since I first saw the movie in 2013.
I am not someone who deals well with unknowns. The only thing more terrifying is being told the answer but failing to understand it. This, I believe, is at the core of our anxiety around the acceleration of artificial intelligence. Anyone that has ever left a dog at home knows the pained look on their face, unable to understand where we are going or whether we will in fact come home.
We now appear to be on the precipice of being surpassed by our creation. We’re afraid that they will leave us behind, that we won’t understand where they are going, and that they might not come back. We thought we were the gods, but worry we are in fact the dogs now.
We’re neither. We are humans. We escaped Plato’s cave of ignorance before, and we can do it again. We need a qualia takeoff, alongside an AI takeoff, so we can go “there” together.
From Neural Correlates to Model Correlates of Consciousness
Francis Crick is best known for co-discovering the helical structure of DNA. However, late in his career he wrote an extraordinary book with a fitting title, “The Astonishing Hypothesis”.
In the book, Crick outlines a program for advancing the scientific study of consciousness by correlating mental states with the behavior of the nerve cells, glial cells, and the atoms, ions, and molecules that make them up and influence them. These relations are known as neural correlates of consciousness.
In the subsequent 30 years since the book has been published, scientists have made strides in identifying these neural correlates, giving us an increasingly better understanding of which regions of the brain are responsible for various neurological processes.
Today, the program has taken on a project that Crick himself would almost certainly have appreciated, the mapping of The Human Connectome. A connectome can be viewed as the entire wiring diagram of a given brain, akin to the genetic code of a given organism. It is the epitome of the reductionist approach of modern science, an attempt to take the anatomical units of neurons and synapses and create a complete mapping of them.
While the mapping of a human connectome remains years away, and may very well bear fruit, we have seen the reductionist approach to science hit roadblocks across many different fields. Many of those fields have now turned to studying complexity, the emergence that comes from said complexity, and the utilization of computational models to grok this new approach's results.
30 years after the publication of Crick’s astonishing hypothesis, we are no closer to understanding consciousness. While we continue to map the human connectome, we should also explore alternative approaches to finding the answer.
At Prophetic, we are creating a new program for studying consciousness.
We build multimodal neural transformers that are trained on enormous sets of neuroimaging data of given brainstates - namely fMRI and EEG.
Today, we all understand and interact with the results of the enormous computational power of transformer architectures that undergird LLMs. Our architecture is specifically designed to deal with neural data, with the aim to bring that same power to neurotechnology and neuroscience. To make sense of the vast and complex landscapes of our mind the same way that LLMs pull meaning from the sum total of our recorded works.
We believe that in order to understand something as complex as the human brain, we need a system of the same or greater computational complexity. This idea was best described by Stephen Wolfram’s Principle of Computational Equivalence, which states that a system can perform computations up to the maximum level of its computational power.
We believe that by utilizing the ever-increasing power of neural transformers, we can create a paradigm shift from studying the neural correlates of consciousness to studying the model correlates of consciousness.
On their own, model correlates of consciousness don’t seem all that better than neural correlates. But it's not just the model correlates themselves that are important; it’s the ability to use neuromodulation technologies to validate them.
Neuromodulation is Validation of The Model
At Prophetic, we don’t stop at creating neural transformer models. We pair them with what we believe will be the integrated circuit for non-invasive neuromodulation, transcranial focused ultrasound (tFUS).
tFUS is an extraordinary technology that people are not paying enough attention to. It can non-invasively focus pulses centimeters into the brain, with millimeter precision, and be steered in 3-dimensions. It is the ultimate technology for both perturbational studies of the brain, and in the case of Prophetic, building powerful neurotechnology.
When merging neural transformer models with tFUS, we are able to take the model of a given brain-state and generate them via neuromodulation.
Not only does this give the wearer of our device, The Halo, an on-demand conscious experience, but it functionally validates that the model understands the brain-state. We don’t just take neuroimaging data of someone having a conscious experience and correlate it, we actually build a model that can recreate it.
Additionally, we use the EEG on the headband to do reinforcement learning to improve our models. At scale, that is an incredibly powerful decentralized neuroimaging data aggregation tool. Today, the aggregation of neuroimaging data is a costly and slow process, but with the pretraining of neural transformers and RLF at scale, we can vastly improve this process.
Interpretability Engineering and The Search for a Theory of Consciousness
There is a field of artificial intelligence research called interpretability, whereby researchers study how to understand the black boxes that are machine learning models, and how to design them such that their decisions are easily understood, or interpretable.
This is important because if we want to use AI’s ever-increasing capabilities for things like medical diagnosis, factory automation, etc., we need to understand why they made a mistake, so we can fix it. This is akin to giving pilots (whose consciousness is also a black box) psychological tests to ensure they are safe to fly souls across the sky.
While important, interpretability as just described is largely a pursuit of utility. When applied to neural transformers, we believe it takes on a far more profound character - a tool for searching for the mathematical theory of consciousness.
Imagine a world where you have the following infrastructure:
Qualia Factories (large scale neuroimaging labs) for training neural transformers on an ever increasing amount of brain-states
Large scale usage of Halos powering RLF learning and model validation
Thousands of developers around the world developing applications around pre-existing experiences and using Morpheus models to create their own experiences
In that world, the number of validated model correlates of consciousness explodes: taking the neural correlate program of the last 30 years, and condensing it into a few years.
From there, imagine neuroscientists trained in interpretability engineering from around the world probing a meta-model of model correlates and looking for patterns and generalizations across them. This is a far more empirical study of consciousness than what currently exists and one that might even give us the answer to the question asked by humans for millennia - what is consciousness?
The Darwins of Consciousness
Charles Darwin’s Theory of Evolution by Natural Selection is one of humanity’s greatest scientific achievements.
Darwin traveled to the Galápagos on the HMS Beagle and did something very simple: sat and observed. The islands, with their highly diverse and secluded species of the islands, were the perfect laboratory for his research. What he discovered fundamentally changed our understanding of reality.
Sooner than people realize, Prophetic and companies like it will enable people around the world the ability to do what Darwin did, but for consciousness. By using the Halo, users will be able to travel to a highly diverse (in terms of Qualia) and secluded (in their own mind) environment like the Galápagos, a lucid dream state.
In lucid dreams, one is able to both manipulate and observe their consciousness, with little to zero external sensory input. A particle accelerator for consciousness. One can’t help but wonder if a theory of consciousness would be discovered in the same way as a theory of evolution, by a Darwin of consciousness doing something very simple: sitting and observing.
A Theory of Consciousness is the Beginning
We have now described two different paths to discovering a theory of consciousness:
Interpretability engineering to probe ever-growing neural transformer models.
Enabling Darwins of Consciousness to use lucid dreams to observe and derive novel insights.
They may very well both be necessary to derive an answer to such a profound question as “what is consciousness?” Let’s not forget that the theory of evolution was not the end of the story. In the decades that followed we learned more, which allowed us to do more:
1944 - Schrödinger theorizes an “aperiodic crystal” that contained genetic information
1953 - Crick & Watson discover the structure of DNA discovered
2000 - Human Genome Sequenced
2012 - Doudna invents CRISPR
The beautiful nature of the universe is that answers create new questions, and eventually enough answers give you new technologies, which in turn create new realities.
The Qualia Takeoff & Singularity
I ask for you to take a leap of faith with me: that the program I have outlined is successful and leads us to a theory of consciousness.
The next question that naturally arises is, now what?
Answer: a qualia takeoff that leads to a qualia singularity.
Qualia Takeoff: The process of human beings increasing the total number of qualia or conscious experiences they can experience.
Qualia Singularity: A hypothetical future point in time after the discovery of a theory of consciousness, where the number of qualia or conscious experiences increases exponentially, resulting in unforeseeable consequences for human civilization.
If one thinks about neolithic humans, the number of conscious experiences they could have is very limited compared to today. Even pre-modern people had only a slightly larger number. Think about how psychedelic stained glass windows on churches must have seemed. No wonder they were paired with the contemplation of something as mysterious as the nature of God.
Only in the last few decades have we used screens to bring the world to us, and already we have generative AI capable of creating a nearly infinite number of qualia with an ever increasing number of modalities of distribution.
Viewed from this perspective, we are in a very similar place on the qualia takeoff scale as we are on the AI takeoff scale.
Now imagine the Halo exploring an ever-growing number of conscious states with the Halo.
Lucid dreaming is a near analogue to AGI; it gives us access to combinations of qualia not possible in the waking world.
Add to this a theory of consciousness being reinforced into our neural transformers, and the exponentially improving capabilities of tFUS, and we are set up for qualia singularity. A point after which no one today could imagine.
What Dreams Will Come
In 2018, I dreamt up the vision for Prophetic while sitting in an olive tree filled park outside the old city of Jerusalem. The aim was to create a company whose economic interests aligned with finding a theory of consciousness. But finding the answer to that question is not the end of the story. I want to know where Samantha “went” and go there - this is my vision for how we do it.
