service of realtime high dimension pattern match powered by hive of Human minds

[BenRayfield]

There are many artificial intelligence (AI) software which do that on grids of computers, but that would not be nearly as intelligent as a grid of Human minds interfaced to the system similar to how http://en.wikipedia.org/wiki/FoldIt turned protein-folding into a game and some people outperform the computers.

A normal Human brain has 10^15 connections between neurons. A normal computer runs 10^9 cycles per second and runs much dumber software than Human intelligence, so even if it was fast enough to run a Human brain as software, that software doesn't exist. If we went to third-world countries and hired everyone who is starving or lacking other basic needs of life, we could create a billion jobs which produce far more value than they would have to be paid.

That value could be used in predicting stock markets, patterns of global events, or anything else AI is already set up to predict. This could potentially obsolete all existing infrastructure used to predict those things and turn it into a free market where people in the system hire eachother recursively to help them predict various patterns in the system.

The realtime part of the service would be useful for adding Human intuition and strategic intelligence to Internet games. Non-player-characters acting in a predictable and boring way? Throw in a few bits per second of real Human intelligence from this cloud, charged at a free market fluctuating price per bit, normalized against actual accuracy of such predictions and assembling of patterns.

Think of the artistic uses of real Human intuition and intelligence sold by the bit at fluctuating market prices, accessed through standard AI software interfaces. Evolve code that uses recursions and fractals and other math and logic to paint great works of art, compose music using evolved code (like in my Audivolv software which evolves Java code to define wave amplitudes calculated 44100 times per second for 44.1 khz audio quality) and Human minds together, and things we can't even imagine yet, accessed seamlessly as if the people are just another function in the software. On the other end, the people solving these puzzles of flowing patterns are having fun playing the game which automatically adjusts to the difficulty level where they most efficiently earn cryptocurrency. Research into questions of consciousness. Can a hybrid software made of AI and real Human intuition (indirectly through the game networking our minds together statistically) experience the same consciousness of the person alone, or of 5 people who are trying to sync their actions in the game so that any 4 of them can reproduce most of the actions of the missing 1, for any 1 of the 5 that temporarily leaves? If a hand is only a machine thats connected through bio nerve wires to our brain, and we feel like we are our hands, then would we also feel like we are the software and other people in this system which our thoughts about the game flow with? Can consciousness flow through an Internet wire the same as it flows through simple well understood bio nerves between our brain and our hands? Are brains, sufficiently networked together at a subconscious pscyhology level, still completely individuals? If you learn to imagine recursive actions into chains of people in this system which happen as fast as any other network call into the Internet, as people play the game in realtime like any other game but more flexible and flowing, then where is the line between the consciousness of the second and third step in that recursion, the second and third people it jumps through in the patterns flowing through the system? My answer is consciousness is like and ocean and we are all drops of water, statistically localized in bodies, a problem easily solved by the intelligent and smooth and intuitive dance-like flowing of our actions and movements in this game. Identity is ok as an approximation, but when you look deeper into it, the contradictions in that theory build up and we should instead find a different way to divide reality. There are many things this Internet cryptocurrency prediction service could be used for, some very practical and others just barely imaginable.

If I flip 2 coins and a least 1 landed heads, whats the chance both landed heads? Most people say 1/2 or 1/4, but there are 4 ways 2 coins can land, I only excluded "tails tails", leaving "heads heads", "tails heads", and "heads tails", and then I asked what is the chance of "heads heads", so its 1/3. That is a simple bayesian statistical question, and most people get it wrong because it goes against how brains naturally think. For that reason, the metacortex which networks our minds together should be some kind of bayesian statistics. It should also react gradually in realtime to flow with our thoughts.

I created http://sourceforge.net/projects/bayesiancortex to network our minds together, but its still an early version. Its what led to this theory of a new kind of economy for prediction based on a grid of Human minds.

A paint program where the canvas is the visual cortex of a simple kind of artificial intelligence. You paint with the mouse into its dreams and it responds by changing what you painted gradually. There will also be an API for using it with other programs as a general high-dimensional space. Each pixel's brightness is its own dimension. Bayesian nodes have exactly 3 childs because that is all thats needed to do NAND in a fuzzy way as Bayes' Rule which is NAND at certain extremes. NAND can be used to create any logical system. In this early version, I'm still working on edge detection and its understanding of the same shapes at different brightnesses. This will be a module of the bigger Human AI Net project and will be used for adding realtime intuitive high dimensional intelligence in audio and visual interactions with the user.

The big players have their own cloud computing infrastructures, but this is more like a very weak form of Borg Collective, without the medical implants or mind control, just a job of playing a flowing puzzle game where you earn cryptocurrency by moving your mouse, thinking into a mind reading game controller (like Emotiv Epoc or OpenEEG), or other user-interfaces, toward higher accuracy of predicting and putting together the patterns.

Think about what the creation of stock markets did for global productivity. This would be that level of improvement over stock markets as they exist today. You could even run statistical software on it. Want to simulate a Human brain? What better platform to run it on than a hive of Human minds through a statistical interface? Or find patterns in SETI's data from space. People are natural pattern matchers as http://en.wikipedia.org/wiki/FoldIt proved.

Just a proposal for a project. I have more ideas than time to pursue them all, but I am building the part that networks our minds together for free. We could use some simple cryptocurrency as a plugin, but as I explain here, I have other important things to do, one of which is this system to network our minds together... https://bitcointalk.org/index.php?topic=77332.0 If I do that part, will some of you add the cryptocurrency? You're the experts.

[Sukrim]

WTF did I just read...  

[BenRayfield]

Want to get filthy rich? Help build it.

"Make things as simple as possible, but not simpler." --Einstein

Version 0.1.4 of BayesianCortex is a 24 kilobyte executable Jar file which contains its own source code if you unzip it. It does things that programs 1000 times its size have difficulty with.

I'm going to build the parts where we network our minds together, but cryptocurrency and politics and Emotiv Epoc and OpenEEG and other ways it connects to the world take very much work because of their pointless complexity. Complexity is offensive to me, and I have far more important things to do than interface an advanced system to complex systems. The question is not will I change the world, but when and which ways. I want to work with equals, not to boss or rule or command over anyone. We each do the work we're capable of, or in how the world works today we each do work under a rediculously low glass ceiling. Hammers. Get your hammers here.

...

But we have to start small. All it knows is Bayes' Rule (a simple fact of conditional-probability) and an intuitive way to do it gradually between the many pixels and hidden variables, designed to flow with how people think.
chance(X, given Y) * chance(Y) = chance(Y, given X) * chance(X)

From Bayes' Rule, we can derive NAND, AND, OR, XOR, 0, 1, and other boolean logic, which are some of the extremes of the floating point math.

Then we teach it to communicate in sequences and combinations of 0 and 1, like teaching a child its first 2 words.

Then we teach it to add 1 to an integer, like 0101 leads to 0110 then 0111.

And so on. This is completely new stuff because its building a computer inside the minds of people playing a flowing puzzle game.

Without the crowdsourcing of intelligence into the Internet hive mind, Eurisko is a good example, but its not open source so we can't verify more than it won some AI competitions of certain strategy games long ago. Supposedly Eurisko was the first and only AI ever to learn math from the ground up, similar to what I'm saying here, and reach the level of figuring out what prime numbers are and that they're important. Eurisko didn't have a hard-coded multiply function. It figured out what multiply means in terms of many plus, and it learned plus by looping over a list, and there was some simpler stuff below that, like a thing is equal to itself.

Lets build a small Turing Machine, or Conways Game Of Life, inside the minds of people playing the game, as a proof of concept. Any logical system can be created from NAND.

We could teach it Conways Game Of Life as a grid of 9 squares and the rules of when a square becomes 1 or 0 based on the 8 outer squares, and use neuromodulation (built into the intuitive way to use Bayes' Rule) to route attention paths to overlap that 3x3 grid of pixels onto itself at each pixel, the same way you can think of 3 pencils visually while only having 1 model of that pencil in your mind. It gets routed to 3 different places on your visual grid of neurons. This behavior has to be built in, and Bayes' Rule, but that's all a general intelligence really needs.

Then, people can teach it how to hate. Amazing skill in that area. The technical definition of love is multiple systems flowing more as 1 system, being in sync, resonating, forming into interesting patterns that flow smoothly. Hopefully the Human species will teach it both sides of that, but its open source so I can't and wouldn't choose to control what others create. Be careful what software you hook into those military robots, for example.

But first we start small... create more versions of BayesianCortex, or whatever other metacortex system you may think is better, which we can play as puzzle games through the Internet, then in our minds flowing together statistically more as 1 mind connected through Bayes' Rule, we teach it to communicate as sequences and combinations of 0s and 1s. You can't just hard-code what 0 and 1 are. Then it would never understand. Its about the learning.

[opticbit]

busy, just found this.  will read it all later.