How to build AGI
How to build artificial general intelligence by building artificial consciousness:
4 years ago, I’ve started writing a book about all repeating systems in the world. Out of these systems, I’ve deduced further consequences. One is: How to build AGI.
Instead of making you read the entire book (called „Systeme“), which currently just exists as a pre-print version on GitHub and only in German, I will summarise it here. I will try my best, to not miss the most important parts of it. I dive deeper into details in my book, but this should be sufficient for the main-question: how to build AGI.
Table of content:
- Artificial consciousness
- “Why”
- Pyramide of needs and happiness-scala
- Definition of Information, Knowledge, Intelligence
- Endless Impulse
- Least resistance
- Randomness
- Independently controlled
Artificial consciousness
(The following could be very likely wrong but, after thinking about this for many years now, I guess it’s correct)
Artificial consciousness is the missing piece to build AGI – a digital, conscious human. Without its biological hull, but with the same feelings, needs and actions as we humans. We don’t know why or how our own consciousness works, but I think I can reverse engineer it: consciousness / or how I like to call it “artificial consciousness” exists, as soon as the following three conditions are given:
- A pyramid of needs which is being followed, in combination with the happiness-scala
- A certain degree of intelligence
- The Endless (Thought) Impulse
I will explain these concepts and their definitions in the following:
The right questions
The most important question is “why?“.
After that, there comes “who?“ and “what?“.
At the very least, there comes the question “how?“, “where?“ and “when?“.
By answering the “why?“ question, all other answers and questions either change, or loose their importance.
You can imagine this concept like a tree. The “why?“ is the tree trunk. The “who?“ and “what?“ are the thick branches and the rest are the small branches leaves etc.
Definitions of all questions:
- The “how?” question: Includes all possible detail possibilities between two pieces of information such as “apple” and “tree”.
- “Where?” question: Always contains the coordinates in space.
- “When?” question or time: Always includes everything that has a frequency.
- “Who?” question: Always includes everything that has eyes or is conscious (see definition in this chapter).
- “What?” question: Includes reality or all knowledge.
- “Why?” question: Contains the missing information after the explanation.
By definition: To a “why?” question always follows another “why?” question. So as long as you didn’t build reality, you can not answer the ultimate, FINAL “why?” question, without asking the builder itself.
The answer to our “why?“ question
In my definition, AGI is just a digital human and so does work just like a human / a human brain / human consciousness. So first, we need to understand us humans. We humans are driven by the need of being happy, because this used to be the best indicator of survival – the answer to “why do we do, what we do”.
So we need to understand the mathematical concept of being happy. To understand what the mathematical concept of being happy is being made out of, we need two concepts: These are:
The pyramid of needs:
Translation from top to bottom, see below:
External love
Self-love
Happiness, Fulfilment, Purpose
Who you are
Who you want to be
What you want to do
Why & what the own goals in life are
What do don’t want to do (values)
↑
psychological health
physical health
↑
relationships to multiple, different groups
↑
Roof over the head, transport, clothing
↑
money
↑
Family / Fellowship / Friends
Consciousness / Questionimpuls
Energy, Regeneration, Reproduction
This is like a pyramid. If you are playing the “top” game, where you are e.g. receiving love from someone else, but where the fundamental below it, self-love, is porous, then you are likely to experience a deep fall of general happiness, in the event of missing external love.
The “happiness-scala”:
(The red line should actually start at the same hight as the green line, to be mathematically correct.)
The “amount” of happiness over the day = 2. You can use it up within a short spike, like in the red line and drop down in your “happiness” for the rest of the day, or you can e.g. stay at the stable level 1, like the green line all day long.
How you feel depends on what you do – and what you do depends on how you feel.
- E.g. a lot of sugar, stimulation, all sort of drugs correlates with happiness spikes.
- Constant wake and sleep times, having enough fresh air, having constant meal times, being productive correlates with the “normal” happiness level.
- Being unproductive, having no goals in live, non-constant wake and sleep times correlates with bad mood.
- No happiness correlates with a depressive state.
Meanwhile in both cases you feel the same happiness over the day, the important difference is, that a person with a green line is able to work on his pyramid of needs / can be productive in the meantime.
The combination
Combine these two and you get the following formula / the mathematical concept of being happy:
Ø Happiness = Integral{G(t)dt} + pyramid
So meanwhile the things you do over 24 hours / the (meaning Integral{G(t)dt}) tell your brain if you are happy / likely to survive or not, the actual offset to others in their qualitative happiness-feeling is the pyramid of needs (pyramid).
If you are in a steep up and down trend regarding your happiness-scala, just like the red line, then the bottom of “how deep you can drop”, depends on what the highest “stable” building stone of your pyramid of needs is.
So e.g. your happiness over the day peaks up and down and you don’t have anything above “Roof over the head, transport, clothing”, then you are likely to fall down to exactly this level of happiness.
Meanwhile if you are experiencing happiness peaks, but you are truly loving yourself (reached pyramid level “self-love), then your lowest level of happiness will still be higher, than in the previous example, where the highest, stable building stone was just “roof over the head, transport, clothing”.
So now we know, why we do, what we humans do and by that can build the logic, which decides what AGI later will do.
It’s crucial, that the top of the pyramid of needs of AGI is switched. So first Energy, Regeneration, Reproduction -> … -> love to others / external love -> self-love.
If you build an AGI like us humans (Energy, Regeneration, Reproduction -> … -> self-love -> love to others / external love), then it will first serve itself / be selfish and only after that serve others with the consequence of being potentially dangerous.
Do not build this.
Instead, build the better version (… -> love to others / external love (and only after this is given, the next building stone) -> self-love) and protect it at all costs.
Definition of Information, Knowledge, Intelligence
Information
Every bit, which didn’t got “connected” to another bit, is a single piece of information. E.g. “apple” is information, “tree” is information, “red” is information.
Knowledge
If you combine, among other things, “apple”, “tree” and “red” with each other, you get knowledge
Translation: Der= The, Apfel= Apple, ist= is, rot= red, braun= brown, rund= round, eine Frucht= a fruit, am= at, Baum= Tree.
Intelligence
In my imagination, intelligence exists as soon as two things are given:
- Memorize information (“apple” and many other informations),
- Link information to knowledge (“the apple is red”)
- The ability to assign information to the right questions (“apple” = what question)
- Being able to independently derive and understand new knowledge and actions / finding new connections from existing knowledge and thus answering the why question (“Why does the apple grow?” → “Because the tree gives it energy and so the tree reproduces, because otherwise it would no longer exist”).
The more of these two things, sorting and assigning information, as well as deriving new knowledge from existing knowledge in as many ways as possible, the more intelligence there is.
The endless impulse
The endless impulse can be imagined as a billiard table full of billiard balls, whereby each individual billiard ball represents a nerve cell with its information (for example “apple”). One ball always rolls in the direction of another stationary ball and only one ball can roll at a time with an “endless impulse”. As soon as a ball is pushed, the originally rolling ball stops. The billiard table is assumed to be ideal, i.e. the speed of each rolling ball ideally does not decrease at first. This means that the entire momentum of the first ball is transferred to the second ball. To predict which “billiard ball” will be hit next, I can imagine three scenarios:
1. Least resistance
E.g. the single pieces of informations
“The”, “apple”, “is”, “red”
…just happened to be most commonly used in this combination, so thats the least resistance in a certain context.
Like a billiard ball in real life, the rolling billiard ball simply hits the one that is next in its path or is in its direction of movement anyway. So if the current ball contains “apple”, then “tree” would be in the direct vicinity of the rolling ball “apple” and is therefore more likely to be hit than the ball with e.g. the information “Darth Vader”. From a physical point of view, you can imagine this as follows: The electrical resistance between different pairs of neurons is different. The “endless impulse” flows next where the lowest resistance is. This would also explain the existing comfort zone and habits of us humans. If we are used to picking up the toothbrush first and then the toothpaste when brushing our teeth in the morning, then this decision requires relatively little energy in our brain and happens almost automatically – like in autopilot mode.
2. Randomness
And/or the rolling billiard ball can flexibly change its direction and bumps into another, seemingly random, stationary billiard ball. This would then be independent of how much the two balls have to do with each other in terms of content, i.e. how likely it is that two balls will collide one after the other. This probably happens when dreaming. From a logical point of view, I think it is very possible that there is randomness, as with flip-flop components in digital technology. Finally, I can well imagine that our brain resembles a bunch of catch and shift registers like in digital technology. These catch and shift registers consist of flip-flop components which assume a random state, i.e. “high” or “low” or 1 or 0, when switched on.
If this coincidence really exists in our brain, as with flip-flop building blocks, this would indicate that there is no determinism in us humans. It means that our actions are not completely predictable, or that our future actions are not controlled by the past, but that chance controls part of our actions. Since dreaming often involves mixing seemingly random memories, I suspect that this random process, as in digital technology, also determines our dreams.
3. Independently controlled
…and/or the rolling billiard ball can flexibly change its direction and is guided by a higher goal, e.g. the pyramid of needs. So if first “apple”, then “yummy”, then “sweets”, then “hunger” are triggered, but the higher goal according to the pyramid of needs is a different one than satisfying “hunger”, the pyramid of needs can be used to focus on an original goal again. This may require more energy than with “The Least Resistance” or if we are on autopilot and simply follow our habits.
To summarize: The billiard ball bumps the next billiard ball, either because it is in the way, and/or by pure chance, and/or because the direction of the rolling billiard ball is controlled.
It actually makes sense that 1, 2 and 3 interact for evolutionary reasons. If a Stone Age man set himself the goal of killing a mammoth at the beginning of the day, but then “sleeps on it again” and wakes up with a “new billiard ball” and looks at things differently, this could have protected the Stone Age man from inefficient ideas. Point 1 should have been useful when mindless things are repeated over and over again, where not much energy needs to be wasted and energy can be saved by acting more efficiently.
Another example of how you can imagine the “endless impulse”: As a glowing circle of fire, which bounces back and forth in the brain forever. The nerve cells / information that are not currently being used are dark and those that are actively being used are lit up. This circle of fire is, so to speak, the perpetual thought (i.e. the never-ending, new question), which even when you think of “nothing”, you still think of a “nothing” / a kind of “emptiness”. This circle of fire keeps bouncing back and forth / this one billiard ball is always in motion and keeps bumping into another ball. This endless pulse of thoughts continues even if no new information is added to our knowledge from outside. This endless impulse will never stop until we die. At brain death, the last moving billiard ball has come to a standstill, at least in the area of our brain. Whether this circle of fire then “jumps” to another level at death is something I consider possible.
I hope I didn’t summarise it too much and went into enough detail for you to understand the picture. I went into further detail in my upcoming book.
Publishing this will either go completely unnoticed in the daily entropy of false hopes and wrong predictions, or it will completely change the way we think about the world.
Thank you for reading this.
-Elias Pfeffer
