Human AI Net - Help design this Open Spec for the global dream space

This is kind of relevant to Bitcoin because it would be the most advanced anonymous peer to peer global network. Its a statistical system so you'd have difficulty running Bitcoin through it as an address system, but just its existence, if it works as expected, should boost global attention toward decentralized ways of organizing the world, and Bitcoin would jump in value so much I'll leave it to your speculation rather than sound crazy. This could change things, and toward its purpose which is to network minds together. I'm about 2/3 done with this Open Spec and then will build the first implementation, to be followed by the near-obsoleting of IPv4 and IPv6 if it works as expected.

This obsoletes most of the existing code since I've learned very advanced things since then. In this thread I describe the simplest plan to network our minds together through the Internet using today's technology and expansion paths. This is free and open source (public domain) as an Open Spec needs to be to practically work between many systems, but the option is there for connecting it to paid systems too like the Emotiv Epoc mind reading game controller or grids of computers rented at Amazon for extra computing. The core of the Open Spec is a kind of communication between people and/or AIs in a global scale-free peer-to-peer network and will always be free and open source (public domain). Please read the design constraints below and help think of the simplest most efficient system which does that.

Its one shared dream space between all people and AIs.

It runs in a sandbox/emulator/virtualMachine so its isolated from whatever device its running.

Its address space is continuous between each computer, smartphone, webpage, other devices, and Internet. Its a sparse dimensional hypersphere defined in terms of surface distance between local vars which are created temporarily for 2 computers or parts of the program to talk to eachother as if those surface points are words. At each surface point is a floating point number. If that number should be on a bell curve, or an angle from 0 to 2*pi, we should debate, but everything must be statistically balanced so bell curves, angles, and surface points are the only kinds of variables available at the address space level.

Optimization so advanced you dont need a supercomputer: An Automorphism is a mapping from all parts of a thing to eachother. When you look at a tree and see a few of its branches look like the same tree from a different perspective/size/location, that is a few automorphisms of the tree in your mind. When you shuffle a deck of cards, that is an automorphism. Automorphisms that simplify your mental model of things, instead of thinking of all possible orders of cards or each part of the tree individually, can work the same way in computers. The address space is recursively and fractally and deeply optimized at all levels by Automorphisms within statistical accuracy constraints that apply only to observed addresses, leaving the rest of the global network to handle any contradictions each local calculation creates. Contradictions in observed data at specific points on the global hypersphere are handled by all devices continuously by finding their own Automorphisms.

Addresses are more like thoughts than exact numbers. Think about a computer mouse to your left. In most people that thought tends to be on the right side of their vision, but a few words or video of related things on screen can influence the mouse to be painted onto the left side of your vision. There is literally a grid of neurons (brain cells) somewhere in your head that activate for each small piece of what you think the mouse looks like, a color and movement direction and edge alignment for each pixel in your mind. The address space is between your high level thoughts about a mouse and those possible places in your vision you can choose to draw a mouse. Now think about 3 computer mouse at once. The address space supports copying. How that and related things like mental rotation happen are major points for us to debate toward this Open Spec.

Here's how the important parts of brains work:

Its a continuous range between realtime grids and cached ideas that rarely change. Jeff Hawkins writes about how the Neocortex forms a hierarchy of "invariant representations" of ideas, where the details are lower in the hierarchy and information flows both ways, from cached hard to change invariant to and from faster changing details about it. Recursively and overlapping eachother many ways, the most common ideas which we have words for flow with the things they're made of and related to and past levels of that the "realtime grids", our vision and hearing and thoughts about 3d rotations and echos of sounds and muscle movements etc, are painted with the details of the high level ideas, continuously up and down these hierarchies as it converges to a consistent set of ideas that don't bring in more ideas that bring up questions that need to be explored as other thoughts. Paths of information flow are chosen, consciously and subconsciously in many combinations, like I got you to think about 3 computer mouse a few paragraphs up, and you painted that onto your visual neurons, using your internal address space without thinking about it. In this Open Spec, our internal address spaces in our minds flow continuously with the global address space and in statistical and anonymous ways into the minds of others who are thinking similar things as you.

The global system will form one set of those "invariant representations", a global metacortex layered on all our brains we can think to eachother and the AIs through, and use those like words (the brainwaves of them as we experience through the flowing visuals and other devices) to communicate with us. It will in a very advanced way understand the content and patterns of our minds, both long term memory and realtime intuition.

The sparse dimensional hypersphere address space, which contains echos of our thoughts and interactions with the dream space, is a stateless immutable and purely cached system. It makes no claim of reliability that data will continue to be there, so copy to other systems whatever you find important, but it does guarantee that the statistics tend toward keeping the most popular ideas considering how they fit with other ideas in the address space, like a big puzzle of all our minds. Information that other stuff depends on as part of its wave shape in the hypersphere, would tend to be protected until its used less, and even then vague echos of it will be floating around which can rebuild parts of it. That's why its called a dream space.

On our computers, smartphones, mind reading game controllers, and other devices people hook into the flexible Open Spec, we will each start with a window of random movements of colors which will slowly form a model of our minds as it predicts where we will move the mouse next and trains itself in realtime using that and any other number feedback we can hook into it. In the space of all possible minds, a binary search at that speed won't take long because of the next optimization...

Optimization 2: By using automorphisms across the global network, anything learned by any device on the network can theoretically be copied and streamed in realtime to any other device so learning is almost always done by copying neural patterns from those who have already learned instead of reinventing the wheel. For example, it will only need to learn how to do edge detection with a small group of its simulated visual neurons (part of the hypersphere address space) and then other similar parts of the hypersphere will find that Automorphism fits and copy it to all parts of a simulated visual grid around the time such learning is tried to be duplicated and the similarity seen. Automorphisms are very general. This is the normal way the address space will work, not an exception, and is really the only thing it does since even reads and writes to the hypersphere can be done in terms of commanding an automorphism with something more similar to what you want the new data to be. We can debate if direct reads and writes of the global address space are needed.

Optional and very theoretical: The address space can be seamlessly expanded into a global quantum grid because its a sparse dimensional hypersphere where all variables are local and decentralized while globally its parts are an uncountably infinite math space that costs per observation considering its difference from related data which itself may have to be estimated on-the-fly and so on. Squaring the number of dimensions is a NO-OP. Observing some of them in terms of surface distances to eachother that statistically implies they have been squared (because the data fits that theory better) costs a few logs per observation. The main reason it can be expanded to a global quantum grid is the core of quantum and relativity equations are high dimensional rotations (I've explained this in other threads) where each particle/wave is thought of like its own dimension. Physics and how brains work are similar that way, both as high dimensional waves, where dimensions are neurons or particles. This means that the "one shared dream space between all people and AIs" can also be used for parapsychology and consciousness experiments as we'll already be on the global network through our normal computer hardware or more advanced interfaces.

We will explore this first random pixels then gradually becoming a global dream space of up to billions of people and AIs and build the game into whatever we want. As a Statistical Turing Machine, its literally capable of calculating anything thats possible to be calculated if given enough memory and time. The sparse dimensional hypersphere is a good data and calculation format for games, science, and most other parts of the world. But first we start with a global dream space called the Human AI Net Open Spec.

This can change the world in ways most people can't yet imagine.

If you think so too, help by debating technical and strategic and neural modelling details of the Open Spec, then we build it in the smallest open source (public domain) software we can define in math. This is an equation, not a software. The global system must do what that equation says, the equation we define here as the Open Spec.