In a single glance, a brain can make sense of a mural of a hundred billion stars. The mind can wrap itself around the universe but not itself.
When people visualize their perceptions they sense reality itself. Perception is as real as the object being perceived.
AI researchers still dream of solving the Hard Problem, the most unpredictable and interesting breakthrough. Some may go insane in the attempt. This impossible paradox will require new ways of thinking.
The 'Mysterians' claim that awareness is as unsolvable as existence itself. Solipsists imply that whatever you're thinking right now is all there is.
Daniel Dennett says there is no 'hard problem'. It's an elaborate illusion that comes from trying to understand too many different things at once. When you look closely, the mystery is always somewhere else. There's nothing magical about awareness: no quantum theory, mysterious fields, or infinite regress.
The most shocking thing about awareness is how pathetically impotent it is most of the time. It takes an unimaginable amount of supercomputer-level computations just to stand in line at the DMV. Let alone drive home.
Perception is a side effect of the brain's extreme complexity. A vast accumulation of facts that are constantly increasing and fading. Awareness is whatever they have in common at any point in time.
Imagine a cube-shaped grid made of cells one centimeter apart, plus tiny wires randomly connecting each cell to ten thousand other cells (incoming and outgoing). Each cell can send a pulse thirty times per second through thousands of outbound wires. If that cube was two hundred meters to a side, bigger than the Vehicle Assembly Building at Kennedy Space Center, it would be as complex as a human brain. This cube would be absolutely packed with tiny wires as thin as spider hairs, an immense block of lint.
In theory this monstrosity could control a million flight simulators at once, but it's so busy being aware that it can't even remember where it put the keys.
Standard Hypotheses:
The mind is whatever connects the senses to the muscles.
The physical pattern of neurons in the brain (or software in a computer) IS its awareness.
Awareness is a constant false memory.
Like other systems, brains have a lot of redundancy.
Remove a single bit from a maximally compressed digital movie (reduced from 4GB to less than 1GB), and the entire movie would look only half as sharp - yet people can still function when they're half asleep or drunk.
A reliable method to manipulate the human mind would transform society. The age-old problem of motivation could be solved.
This method might create the deepest and most wonderful sensations ever, without the addiction problems of hard drugs, or it could create invincible fear of the world's dictators and secret police. People could be made to automatically get along, painlessly learn new skills, or be converted into ultra-efficient laborers.
The philosophy of consciousness
Currently, we know of two "trivial" ways to create an artificial thinking mind that are actually impossible:
- Make a list of every possible question and response. Faced with any input, the AI would look up its reply in an exhaustive database.
This is how expert systems are programmed. The search would be much slower than a traditional mind, taking billions of years to answer the first question. After that the wait would get exponentially longer.
Strangely enough, such a system would still be fully aware, since the lookup pattern would be as complex as the pattern of any brain. It would also represent and embody the original programmers' awareness.
- Copy an actual brain, synapse by synapse.
The third method is only theoretical, but not known to be impossible:
- Randomly create and then evolve an input/output feedback device using a sufficiently large neural network like GPT-4.
The Turing Test Implication
To pass this test, a computer must fool a human that it is human too, communicating only through text.
Alan Turing implied that any entity that consistently appears to be aware, automatically must be aware - perhaps even if its responses are generated by pure chance.
Human minds were optimized by evolution. Any other system that could do the same would have to be similarly complex.
A Turing Corollary:
In theory, a blind person could learn enough about color perceptions to fool non-blind persons.
This type of virtual awareness would be slower, but the mental model would be equivalent to experiencing colors.
Universal Improver
A hypothetical program that could translate any statement into a new one that would further clarify its meaning, and repeat this as often as necessary.
Requiring a vast and expanding knowledge base, a true Universal Improver should be as complex as a full-fledged AI.
Confabulation AI
A hypothetical program designed to combine as much incoming data as possible in a data framework. It could be about any or all subjects.
It could eventually organize its knowledge sufficiently to fully make sense of it, becoming a global reality index.
Universal Compressor
A hypothetical program that could translate any statement into a shorter one while preserving its meaning up to a limit. The most efficient way to represent reality.
Like all compression software, the larger its reference base (the more it knows), the easier new information can be categorized.
A sufficiently comprehensive compressor would include itself, and become fully aware.
The Zimbo Paradox (Hofstadter): an entity which acts aware, but isn't (also called a philosophical zombie).
Some claim a machine could simulate human-level awareness without 'feeling' anything at all. A robot mind powered by such a device could be killed without moral qualms, even if it objected most convincingly.
Should your husband's brain be secretly swapped with such a computer, it wouldn't be wrong to kill him even if his behavior never changed, since he was only pretending to be aware from the moment the exchange was made. The computer would merely be simulating his behavior, but there would be nothing 'there' to have feelings.
'Chinese Room' (Searle):
A sufficiently immense list of simple rules could automatically generate statements claiming to be from a self-aware mind.
If someone were to slip a note with Chinese characters under the door, and you had nothing better to do, you could use the rules in a very large book to generate a sheet with different Chinese symbols to push back outside. It would take longer than the age of the universe to write a single character. You have no clue what's going on, but the notes turn out to be part of a meaningful Chinese dialogue.
The rulebook is somehow able to generate a mind that claims to be fully aware. It has its own personality, can answer questions, and convincingly claims to have thoughts and emotions. But where is this mind located? Not in your head, since you don't know any Chinese. Not in the book; it's just a lot of seemingly meaningless rules. (This book would have to be bigger than the Death Star at least.)
Perhaps the 'room' could claim to have feelings.
Imagine trillions of people writing, exchanging, and erasing small pieces of paper containing ones or zeroes for a millennium.
The result would be a complete simulation of your mind in the next second. This giant setup would create the awareness that you're feeling right now.
Neuronal analogs
A mind is the code of its environment. Every mind is its own total reality description.
It could be used to reverse-engineer a fully functional virtual universe based only on the mind's memories. In theory everyone could inhabit their own imagination and have it appear completely realistic, though it would appear fake to others.
The simplest mind type (holographic interpretation)
Any mind capable of having just one thought could automatically have any other thought. True AI may be achieved by fully defining only one real-life situation.
Type One AI: top-down, highly abstract, rigidly organized, 'big picture' general intelligence. Long term goal: Net unification; integration of all human knowledge.
Type Two AI: bottom-up, combining many simple low-level functions, small modules interacting unpredictably to learn higher tasks.
AI dangers
When the mystery of awareness is finally solved, reality will change unpredictably. Mankind will wake from ancient delusions it didn't know it had, and see itself from the outside. Unfortunately we don't have Michael Crichton around anymore to write about the dangers.
Science is drawn to limits. Faced with Lovecraftian sanity traps, new pain experiments, or religion reprogramming, further research may be banned. Human progress might even collapse.
Anthropics
We only know an insignificant fraction of reality, almost the same as knowing nothing at all.
However, we do have access to one fundamental, inscrutable truth that can not be calculated or deduced by logic alone, the only real magic that exists:
selected from all of reality, what is an average mind like?
That would be us. In a limited way, we can measure the ultimate nature of existence.
Why are we trapped in this banal universe, instead of some unimaginably complex one?
Each instance of human awareness is duplicated and thereby spread out over many functionally identical brains, an infinite number of equivalent processes in uncountable parallel universes (the Everett and Tegmark interpretations).
Simple minds may be easier to generate, making them far more numerous throughout reality.
A mind is less likely to experience being some ultra-advanced futuristic machine intelligence than being an organic creature, since the second must create the first.
That may help explain why we are here, but not how. That will require stranger concepts. Highly organized physical realities appear to be more conducive to the existence of awareness than purely random ones.
People live half a second in the past. Anyone who saw a cannonball hurtling toward them would not see the last half second of its flight.
Awareness is made of many pre-existing representations. It seems effortless, but has to be relentlessly recreated every second. It's the outcome of competing connections made in parallel paths, subconscious possibilities emerging from separate streams, the ultimate side effect. It can't control itself since it must be unorganized. Otherwise, personality could easily be rearranged.
Many frames are generated every second. Perceptions may be made of earlier perceptions that are somehow re-experienced outside normal time. Most are erased and forgotten.
Awareness is a bit like chess. The mind rapidly explores many future possibilities and rejects most. Memories are briefly activated in rapidly pruned decision trees.
Not a single thing, awareness is the only thing, a constant approximation that can't stop chasing itself. Every random thought has hidden thoughts leading to it, a background story for every second, updated pathways through an endless maze.
All knowledge may be related, yet the human brain can only have one clear thought at a time, ignoring every other possibility. Even the simplest idea requires all available explanations and associations at 100% of brainpower.
Awareness may be impossible to simplify, or even to think about in a meaningful way. No matter how well a feeling is described, it can't be understood outside itself.
Awareness is generic. Thoughts are placeholder illusions: minds can't describe what they don't understand, converting the unknown into irreducible elements. The ignorance becomes a new concept.
Emotions represent the brain's order of priorities and progress. They improve when plans, problems, and values are sorted faster than new ones appear. The brain's purpose is to know its location, situation, and obligations at all times.
The brain is a map of reality centered on itself.
As an automatic description of its environment, awareness is actually far removed from free will, the difference between seeing a lion versus imagining one.
Awareness and time
The core of consciousness is the perception of time. In each moment there is no awareness at all. In a millisecond, there is almost no awareness. It takes at least one tenth of a second to form a rudimentary thought.
A perception which seems instantaneous is actually spread out over half a second. Reflexes must take up the slack. We are constantly reliving the past.
Awareness is based on repeating cycles, pre-selected choices and scripts. Thoughts are constrained by familiar tracks. There's a vast array of common perceptions and scripts for each task. Repetitive tasks are automated, not so much experienced as remembered.
In the most familiar situations, all the previous occasions merge into a timeless blur of perception. Only unexpected changes trigger new awareness.
Natural selection is always changing the brain: constrained by pain and pleasure, driven by available associations.
People only help others because they fear pain (guilt is a good evolutionary tool).
If something feels good, the responsible circuit gets reinforced. Pain causes an even stronger opposite reaction. Relative strengths are genetically determined.
According to computer simulations going back to the Perceptron era, most brain activity seems to be a Fourier-like process of repeating waves overlapping semi-randomly, attempting to form new waves. It's a strictly mechanical process, even if the moving parts include chemicals:
Natural drugs like cholinesterase regulate neurotransmitters like acetylcholine, stimulating random connections in a process like natural selection.
Damage to a small portion of the brain called the insular cortex, or insula, can eliminate nicotine addiction, and may reduce food cravings. Somehow, this tiny organ is responsible for pleasure seeking behavior. Damage to the insula also reduces social skills, and it apparently helps regulate novelty and thrill-seeking. It can both cause depression and end it. Drugs that target this sub-organ might even tamper with political affiliations and beliefs.
About a hundred identified brain structures are just as poorly understood.
From feedback, minds evolve common pathways to regulate drives and fears. Individuals with depression must be prodded to perform even simple tasks. They actually can "snap out of it", but they have to keep "snapping" without end to function, and "snaps" are inherently painful.
By creating a background of well being or anxiety, the brain normally stays focused on its current priority, either to get more rewards or to make the pain stop.
To initiate major life changes, active circuits must be suppressed while a greater number are rapidly created, aided by stress hormones.
Self modeling and the Bootstrap Principle
Awareness is the representation of all other instances of awareness: its own simulation.
Brain patterns are partial models of other patterns. Each insight is a combination of several earlier insights. Representing long logic chains, thoughts can only be about other thoughts. Embodied rules try to predict other rules.
To function, the brain requires a steady stream of unpredictable input that is automatically sorted into existing categories. It filters out useless signals below a threshold or that consistently repeat (by creating dedicated subconscious pathways for them). Pattern and frequency scans block over 99% of incoming perceptions.
Neural Networks
A brain is made of up to a trillion neurons, with multitudes of connections between them. New connections are always forming, but most are soon deleted.
The only thing a neuron can do is send a simple pulse through all its thousands of outbound connections - after it has received and added up enough incoming pulses from many other neurons. It may regulate the pulse's intensity, duration, and its repeating pattern.
Sleep and the 'hard drive problem'
The mind gets cluttered with random data, and periodically needs to be 'defragmented'. Apparently, dreaming somehow reorganizes and reduces the effects of less important real memories, by creating many smaller false memories that take up space without being consciously recalled. The surreal nature of dreams (in the rare cases they are remembered) is only a side effect.
Some dreams simulate thoughts, emotions, or situations too stressful to have while awake. The mind becomes more paranoid while asleep.
Even so, the brain can retain a million rarely used circuits that only fade slowly or never, like for forgotten foods, playing Pac Man, certain phobias.
Short term memory requires a vast number of weak connections. All the relative strengths summarize temporary, provisional, and incomplete conclusions. Thousands of random perceptions try to form a higher realization.
Long term memory forms by the strengthening of frequently used temporary connections. This happens much faster in stressful or intense new situations - like while flying through the clouds of Jupiter, or on the first day of a new job.
It may also require REM sleep to happen first.
Awareness is strongest when forming new permanent thought-chains, such as new rules or syllogisms. Humor can describe flaws through absurd metaphors. Another example is a sudden flash of insight or understanding.
What are the atoms of thought? Bits or something like neuro-bytes? Neurons fire in groups that may be sentient by themselves. Incomplete but real awareness can exist independently in different brain regions, as patients with split brains show.
The brain is a mostly static reality description. The furious activity of awareness represents only a small change in its permanent structure, an effort to generate a few long term memories. Personality, intelligence, and language tend to be stable over decades.
Neuron connections have complex rules embodied in their physical and chemical shapes.
A neuron might only deign to fire when the incoming signal strength between two types of dendrites almost touching its surface has some fixed ratio, like 0.75 or 2.24.
Any vague or detailed concept, any type of object or person represented in memory, has literally millions of different neurons associated with it (but most of them only very weakly), with many different firing strengths.
The black hole of knowledge
Part of awareness is forgetting: most data is implicitly stored in the gaps, things that can be safely ignored or forgotten. Social information is also hidden throughout the environment and society.
No one even understands or knows their own motivations. Humans can't even solve problems like homelessness or the simplest family conflicts.
Smarter people and brain states are indeed more elaborate and specialized, but they are also more cluttered and constrained (the best books about physics are written by smart non-experts). The more someone knows, the harder it is to represent all their undefined and incomplete insights.
Unlike most devices, the brain can't be turned off. Its status awareness is constantly being updated. People always know where they are, how they feel, and are forever modeling their immediate future. An unbroken line of sentience could be interpreted as a single life-long thought.
People don't know their own names most of the time. Any fact can be retrieved but instantly vanishes again.
Incoming stimuli create many weak associations, a cloud of vague impressions. Semi-forgotten facts form random connections in a search for new meaning. A nagging association suddenly becomes crystal clear once it's pointed out, or when a higher brain function stumbles upon a connection.
Each thought is made of many approximations at ever-higher resolutions, until the next insight arrives.
The most profound impressions are the hardest to put into words, the indefinable knowledge commonly called the subconscious.
Why does pain feel real?
Pain requires an agony component. The deep brain tries to override all restraints, while the cortex desperately tries to generate some solution. Pain can be 'pre-membered' - as if it were being felt in advance. Any painful memory can start an elaborate flashback sequence.
Awareness and emotions - both weak and ultra strong, good and especially bad - are unbelievably cheap, in the sense that they are easy to create in a brain-like system. For that reason, post-human minds will be designed to be literally unable to feel certain undesired emotions.
The human cortex is a framework for pre-assembled thoughts; top level logic enforced by bottom-up emotions.
Awareness is an endless parade of tiny, disposable thoughts; thousands of oversimplified models linking up in random chains. Unfortunately, it appears that no one is really in charge: the human mind has no central control system.
Ideally, all the chaos will cancel out, leaving one new insight.
Perhaps people are only fully aware when they think about awareness.
The reason math is difficult
Like self-awareness, it's one of the few things that can't be simplified: too many concepts are randomly connected, requiring multitudes of scripts stored in different brain regions.
Mathematicians try to learn to 'visualize' numbers, which is itself a great mystery of perception.
While the problem of human consciousness can be reduced this way, it can never be eliminated.
The last and most crucial insight is always another step away.
Ultimately, awareness may be an extension of an age-old problem: why does anything exist at all?
(The above is explained in greater detail in the book 'Anthropic Intelligence'.)
Probably the best hard SF novel ever written? Infinite Thunder by Jack Arcalon.
Buy the book
Read the chapters