The future science of artificial intelligence (AI) will finally solve mankind's most intractable problem: laziness.
People hate to work, especially when they don't feel very good, which is very often.
If everything goes according to plan, within one human lifetime smart machines will take over Earth. Unimaginable alien minds will then do all the hard work so that humanity can finally retire forever.
But how exactly will it happen, and is there any way to make it happen sooner?
The following is of course pure speculation. Thinking about awareness is like trying to imagine everything at once by maximally focusing on one thing. Philosophers working on the problem often feel like a giant revelation is at hand, but they keep editing their writings until nothing remains.
Thoughts can only be about other thoughts, each requiring its own definitions. The problem with explaining intelligence is that thinkers keep trying to simplify the truth, when the exact opposite is called for. We need more new concepts, less familiar explanations.
There are many possible thoughts: This moment is happening right now. It is always now. Show the exact arrangement of neurons that created this notion.
No one knows where to begin.
The mystery of awareness is called the Hard Problem.
When they first encounter it, many people think they can solve the problem, even making it sound easy. Some claim they once knew the answer, but then forgot.
Then reality sets in. In six decades of effort, the elusive goal hasn't come any closer. The mystery seems as great as ever, if not more so.
Most people don't care about this stuff. Those who do have heard it all before. Apparently, some problems are so vast they're boring. Complicated words don't explain enough, so more complicated ones are invented meaning even less.
In fact it should be illegal to explain that something is complicated. Anyone can understand anything, or at least come close.
The mind creates its own reality. Nothing we think we know is completely true; it's mostly barely adequate.
Many problems are fundamentally incomprehensible at first, like higher dimensions. To understand them requires adding new mental models which can't be integrated with existing ones. The expanded mind understanding the higher dimensions would be fundamentally different.
For anyone desperate to understand the riddle of awareness: you already do! Everyone figures it out to the maximum extent possible, when they learn the most important fact: The brain is a machine that converts pulses and is converted by pulses.
The standard model is almost too simple: at any moment in time, a mind IS the physical state of its brain: a fantastically complex, constantly updated model of its reality.
It could be represented as an immense chessboard. There's nothing there but moving chess pieces - but somehow this pattern also represents itself.
A mind can only access itself, but it also 'stores' memories in its environment, by creating and responding to regular triggers.
Since our universe is full of actual and implicit patterns, awareness could be everywhere: an unlimited number of meaningless perceptions in every brick, air, and even empty space, but we can't detect them. If so, the meaningful patterns in our brains may be extra amplified to overwhelm all the chaotic ones in the environment.
Various experts have tried to invent simple rules that would automatically generate intelligence:
- Assign a word to one or several categories, depending on how it modifies other words.
- Name any two items, and select a third to relate them.
- Start with any term and add definitions.
- Assume every point in reality is described by tags.
- Binary selection: a rough approximation, followed by a series of ever finer choices.
- Start with the simplest possible grid universe, and add levels of detail.
- Describe one item with perfect precision, and you've described them all.
The 'Mysterians' believe in 'qualia'. They think the mind contains pseudo-physical 'atoms' of awareness in a way no one can understand.
Many fundamental constants can't be further simplified. Awareness is as much an aspect of reality as weight or mass or quantity.
No matter how well we describe a feeling, we can never understand it outside itself.
The Mysterians may be at least 90% wrong, but they do have a point.
Awareness reflects reality, but is deeply biased. According to common evolutionary principles, minds should be as simple as possible; basically danger detectors.
This leads to the first important insight: humans are actually quite stupid.
We're inefficient, ultra slow learners. Machines can perform logical tasks billions of times better. Just try to remember a short list of foreign words. To load every Russian word into memory, my obsolete PC only has to make a brief clicking sound.
A crude 100 KB picture is more vivid than the human imagination. Just try to draw real persons from memory: they look like cartoon zombies.
Human minds are very wide but not deep.
The next stage of evolution, if there is one, might involve the elimination of now-obsolete emotions.
Post-humans could become like smart ants, with smaller but much better organized brains. Perhaps they'll be able to hold images in memory, and 'read' a page minutes after glancing at it.
After the final breakthrough, true AI may turn out to be surprisingly easy to achieve. Most of the hard work has already been done. There are thousands of profound clues about how the brain works. They just needs to be put together, a vast future project that will have to sort and combine more data than any previous endeavor.
The project's goal would then be to completely describe just one human thought.
-neurology
The brain is made of trillions of neurons, each a simple single-cell computer, connected by quadrillions of tiny wires.
Each neuron has separate sets of incoming and outgoing 'wires', except for some feedback functions.
Each neuron receives many more signals than it sends out, from thousands of other neurons just like it. When it finally does 'fire', the pulse goes to thousands of different neurons, a few of which may then fire themselves, if they get enough confirming signals from still more neurons. A small spark affects billions of cells. Eventually, a pulse will reach a muscle, and the organism takes action. That's almost all there is to it.
Less than 1% of all neurons receive direct signals from the senses, but each is no more than six layers removed from the outside.
With each 'firing', connections (including those between simultaneously firing neurons) are strengthened. Otherwise, they begin to atrophy.
Neurons can also fire at regular intervals when triggered, as 'pacemakers' for various activities.
Perhaps all connections can be categorized as 'good or bad' associations, like yes or no switches. Every neuron has an 'appropriate' action potential for any situation. Identical responses of two neurons to the same input are turned into links between them, so that only one of them has to fire. The brain is always looking for shortcuts.
Neurons can amplify or suppress their output depending on whether more 'good' or 'bad' signals arrive.
-Society of the Mind
The next level are 'meta-neurons', made up of many linked individual neurons. The brain is a collection of millions of interconnected groups combining inputs from each other, ranging from very simple to so abstract they still haven't found a pattern to respond to.
They don't understand what they're doing. Each group 'knows' only one thing, and has to cooperate with many others to function.
They only transmit when the single thing they're interested in is happening, but with excellent timing and many possible intensities.
Some groups sort perceptions into categories (male/female, toxic/edible, beautiful/ugly).
Low-level groups can start and stop higher groups, from general to specific. Higher decision-making groups indirectly influence the lower mood, motivation, and alert levels, a much slower process.
Groups are always competing in a never-ending winner-take-all contest. They activate in rehearsed sequences, and may have to complete scripts once started.
To create the first AI, many small teams will have to develop each element separately.
Activation networks: Some neural networks amplify very weak signals.
Suppression networks: Other networks suppress frequently repeating signals, so that the more important ones can get through.
Synchronization networks: Different groups of neurons fire at different frequencies. Eventually, one frequency comes to dominate.
A 'white matter' multiplex network may encode signals between different brain regions, with many possible error and failure modes - including dementia.
Humans have at least three 'brains', stacked on top of each other.
Each level stores different memories, and performs different tasks; simple emotions in the brain stem, situational awareness in the cerebellum, complex 'logic' in the cortex.
Fixed drives like attentiveness, novelty seeking, and stress responses combine into stable personalities.
Repeating or notable patterns in the environment cause new neuron connections to form, which recognize the patterns and try to predict them. They start as random links that are soon strengthened.
The brain has many simple mechanisms for:
- horizontal, vertical, and diagonal image scanning
- motion tracking
- balance
- grabbing moving objects
- recognizing simple or repeating shapes
- sudden changes in brightness or speed
- thousands of common skills, like sitting down, tool use, turning pages
These will all have to be analyzed and explained separately.
-The extreme complexity explanation of awareness
Experts suspect the great mystery of awareness will eventually be replaced by many smaller ones.
Instead of the vague, overly general theories available today, the goal will be to fully describe all the elements of awareness existing in a single moment. Understanding one frozen instant would be necessary to understand how awareness arises with time. An exhaustive description of all the items in memory would form a complete (if highly biased) model of reality.
Awareness is intense concentration: all mental resources are dedicated to one scene. Fear is the most extreme version, an incredibly tense balance trying to amplify a solution.
The mind is an immense organization, like a self-rewriting encyclopedia. An incredible number of facts constantly compete for precedence.
It's this immense activity, not some mystical essence, that makes sentience seem 'real'.
There's a century in every second. It will take a vast team to understand the first human thought.
Even so, at each instant of time, there is very little awareness. A mind can only know an insignificant fraction of its universe or even itself. Compared to the mystery of existence, it might as well know nothing at all.
Full awareness is spread out over many minutes or even a lifetime.
Most thoughts are shorthands for ambiguous truths.
When someone is working hard, they don't know why they are busy. They would have to stop to recall their motivations.
Memory is constantly being rewritten. One second ago you were a different person, whose essence is already forgotten. We remember being aware in the past, but the full realization only emerged after the fact.
A mind can only consider a model of its memories, not the memories themselves, and the thing that's doing the considering is different from the model or the memories: one brain state trying to describe other brain states. Other states are then needed to make sense of it.
Like the other senses, awareness conveys pain or pleasure - only it can do so in advance.
Awareness is the average of the known truth: an index, not a calculation; the result of thought, not the cause.
The 'self' is the area of densest computation. A universe of facts, with associated pleasure/pain values around an unstable center.
At a still higher level, are 'personas' (meta-meta-neurons).
Only one persona can be in control at any time, with several helping out, and others on standby.
Each regulates a specific situation by combining data from lower levels. Not all rational analysis happens in the frontal cortex.
Personas combine to form the mind's top decision maker.
-Scripts
The brain keeps repeating the same tasks in slightly different combinations. There's a specific memory module for each of tens of thousands of actions, a 'script', or a template of neuron networks, connected in a vast, adaptable flowchart.
-The free will illusion
People consider themselves to be free agents while performing common tasks like walking or driving, but they're almost completely immobilized. Only a few actions are allowed within narrow parameters.
No one really controls what they're doing. They're too busy trying to stay on course.
Every action requires a reason, not necessarily a rational one. People can't suddenly throw themselves on the ground, give up all of their possessions, or change their beliefs, unless a script is activated.
-Toy universe theory
Minds become aware by constantly running simulations to predict the future in a virtual universe.
New simulations are automatically created by new inputs, which the mind triggers by changing its environment to match existing simulations. The easiest way to do so is by moving to a different position.
Simulated outcomes are compared to reality (an essential part of awareness), causing new mental models to emerge and perish.
When a thought is completed, a new representation is created like another mind toy.
Awareness is an uncontrolled, emergent process, the sum of all thoughts, not even a formal brain function, but the ultimate side effect. Otherwise, personality could easily be changed.
The organism's first goal is survival. Pleasure and pain signals try to balance out according to a genetically determined ratio.
-Self modeling
Awareness is another bootstrap principle, a model of models: the representation of all other instances of awareness.
Maybe it is its own simplified model. Each insight is a combination of earlier insights.
-The narrative explanation
Humans create elaborate narratives of earlier awareness.
With trillions of events per second, humans can't really remember their past perceptions, only their reactions to them.
All organisms with a nervous system, from nematode to human, relive the past by combining previously stored thoughts.
The perceived differences in the new situation are at the core of awareness.
-The data compression explanation
The human brain is a universal data compressor. It makes itself more efficient by constant rewiring itself. The more it knows, the less data is required to represent new knowledge.
Awareness is learning.
~
Memory, like reality itself, is fundamentally chaotic. There usually is no big picture. Awareness attempts to compress the incompressible, by finding whatever detail two unrelated facts have in common. Often, it will only be the fact that the mind is thinking about them at the same time.
Memories are maximally compressed in such a way that relevant changes are quickly noticed.
-The top level (if any)
The highest level is a composite persona - the meta-meta-meta neuron.
Awareness is the ultimate voting system, the best available truth embodiment.
The top level only thinks about one thing at once, if that.
Any AI that could answer scientific questions as well as an average trained human would need to perform high-level logic, and be fully sentient. To calculate the surface area of the earth, it would roughly need to multiply the square, not the cube of the width.
Often, once a mind has learned some new fact, it can no longer explain how it knows this thing, but it can usually prove it's true in new and unpredictable ways.
The mind's true purpose is to select its immediate next state.
In a way, awareness is mostly self-extinction (alcohol, by the way, interferes with this suppression process). A large part of the human mind perishes every instant, tens of millions of temporary connections fading out forever and forgetting almost everything they have assembled to create the illusion of the moving now.
version 3.3
Perhaps the best hard SF novel ever written: Infinite Thunder by Jack Arcalon.
Buy the book
Read the chapters