This div is called theData. It contains divs that will become subwindows when you click the button. The inner divs have a title attribute that you can not see below, but it will be in the subwindows.
Audivolv automatically evolved this Java code to help me play this sample music: .OGG .MP3 .WAV, but if Audivolv does not work, install Java and try again. Sample music from my older music program: CodeSimian.
Move your mouse to play whitenoise and other random sounds. Audivolv continues learning how you want the mouse to sound. You can only play music with your mouse after you teach Audivolv what music is. The current version has good/bad buttons to teach it, but future versions will learn what you like by your mouse movements alone.
are an unfinished software called Natural Language Mouse Interface. Try making your browser window bigger or smaller. Checkbox keeps a subwindow on screen. Later, I will add Artificial-Intelligence to control these subwindows. Works in: IE, Firefox, Safari, Opera, Chrome.
Audivolv will become the "global language" described at the end.
I am continuing the work of Timothy Leary. His video describes a "global language" of music and lights and global brain to brain communication (OpenEEG and consciousness experiments through music psychology) and dancing (Audivolv will use Nintendo Wii controllers for that). Every music and dance will be different, like every neuron in a brain is different. Please listen to the video while reading.
408.334.7214 (call me any time) benrayfield@mousesearch3d.com
View my resumE Bug List
ToDo Summary.txt
Every download of Audivolv contains technical documents and source-code. Audivolv (newest version at sourceforge) is a Jar file. Rename it to x.zip and unzip it.
Extremely technical summary and links
Audivolv is open-source GNU GPL 2+. Natural Language Mouse Interface is GNU GPL 3+, or business licenses are available.
Lambda-based self-aware AI. Collective-intelligence of millions of people and computers. Internet <-> computer -> music -> You and your friends -> mouse, wii controllers, and/or OpenEEGs -> computer.































Audivolv - where Instruments Play The Musicians

What is artificial intelligence (AI)? Its any software that does unexpected things and surprises you how well the unexpected things work compared to what it was designed to do. We expected the Deep-Blue chess computer to win, but we could not predict what chess moves it would do until it played against the best Human chess player.

The exact definition of intelligence is a game between 2 players/symbols X and Y:
Its like rock-paper-scissors but with 2 choices instead of 3. Each player chooses X or Y then sees what the other player chose. If both choices equal, player X gets 1 point, else player Y gets the point. After many rounds, the player with the most points is more intelligent because this is the simplest form of pattern-matching and prediction. Solomonoff Induction is based on this idea, and it helps me build better AI.

Most people think AI is something done only by scientists that takes lots of tedious work, is hard to use, and is better to watch on television than to play with. My softwares: Audivolv and Natural Language Mouse Interface will be simple and fun for you to use. Forget the instructions. Its Human-nature to use it the right way. Audivolv is almost that way now. I'll keep working on both of them, but first I need to get a software job. See the links below.

Audivolv is a musical instrument you play with your mouse, and an artificial intelligence (AI) that modifies itself (by writing new Java code) to sound more the way you want it to sound. The current version learns how you want it to sound with Good/Bad buttons, but later versions will not have those buttons. Later versions will use my theory of psychology to replace those buttons. That theory of psychology is: If you like the music you are playing with the mouse, your mouse movements (interpreted by AI in combination with the music you are playing with the mouse) will be more predictable by a smart enough AI. AI will interpret your mouse movements to know how much you like the music you are playing, then AI will write new Java code to sound more the way you like, an intelligent musical instrument.

How does Audivolv fit into my long-term strategy? Read my new website for details: HumanAINet.com


Audivolv - where instruments play the musicians (and musicians play the instruments, a strange-loop of music interaction between AI and people)
Always free and open-source GNU GPL 2+

Downloads, sample music, sample java code Audivolv evolved, contact info, instructions how to use Audivolv, Bug List to report any problems or confusions, and lots of other important stuff is below, but its not well organized because I have too many things to say and not enough time. I started creating my "Natural Language Mouse Interface" software to organize this website, and I will use it on this website eventually, but for a different reason. It has better mouse and graphics and text interactions than any other website, and it does make it easier to organize things, but I have more important things to do than finish it. You can find the links to the demos (version 0.7 is the most impressive so far) below. The softwares I created, and the technical documents about what softwares I will create, are impressive enough to get me a very good job. My problem is most people do not understand those things because I work in an unusual area of software that most people think is a toy area or not important: the interaction of mouse movements, AI, Human psychology, and realtime generated sounds (which become music after you teach it what music is). Most people think I should spend my time learning about less-advanced softwares and get "industry experience". I'm looking for a job in an industry which mostly can not understand the advanced things I have already built therefore they think I should be doing less advanced jobs. Audivolv is very easy to recompile to create a new Jar file that you double-click to run Audivolv because its a strange-loop. Every Jar file is a Zip file. Rename it to Audivolv.zip. Unzip it. Use Apache Ant on the buildAudivolv.xml file that is also unzipped, and it creates a new Audivolv Jar file, which you can unzip somewhere else and go through the strange-loop again. In other words, Audivolv is its own build environment with as little setup as possible (only Ant is needed). Thats only for recompiling the core code. The Java code Audivolv writes at runtime is compiled by Audivolvs internal Java compiler (Javassist) and used less than a second later. Its completely automated so you do not even know its doing that while you play music, unless you try to know. Audivolv is a self-modifying AI software.


Looking for a software job. Here's my resumE

I also could talk to you about a Venture Capital investment in me for my Natural Language Mouse Interface software becoming a distributed search engine powered by the brains of each Human user as collective-intelligence. I would have to finish building that software (I'm not working on it now) and build a server software to organize them.

List of URLs:

The most recent version (all versions include source-code and design documents) thats certified to work and certified to include no viruses, should be here, if you search for audivolv: http://download.com

Most recent version, and the bug list (to explain any problem or confusion), and the feature-request list, should always be here: http://sourceforge.net/projects/audivolv

Most recent writing about it, my contact info, and official webpage: http://audivolv.com

Direct link to the program will be here as long as my webpage stays up. Add this link in any webpage, click it, click OPEN, and Audivolv should start running immediately (if you have at least Java 1.5): http://audivolv.com/audivolv.jar

Forum thread about the technical details of creating the first open-source global AI network (the feature-request list contains that also): http://www.kurzweilai.net/mindx/show_thread.php?rootID=152417

Forum thread about the technical details of an exponential realtime Java compiling optimization, but only for the subset of Java that Audivolv generates: http://www.kurzweilai.net/mindx/show_thread.php?rootID=152642


What does Audivolv do now?
* Starts when you click here then click OPEN. If it does not work, click here for how to fix it or my phone/email.
* Open-source GNU GPL 2+. Unzip that file instead of opening it to get the source-code.
* Sound reacts instantly to mouse movement.
* You move the mouse any way you like and think about how that should sound.
* You choose "Sounds Good" or "Sounds Bad".
* Audivolv tries to learn what you were thinking.
* Repeat until your musical thoughts become real. It took 5 minutes to create this example music: .OGG .MP3 .WAV
* Close Audivolv and start over if you confuse it too much.
* Audivolv is a Java software that creates Java software to change things about itself as it learns
* Learns techno music faster than any other type, but what did you expect an AI to play first?
* There is not 1 line of music code in Audivolv until you teach it what music is. Then it writes the code.
* Find all new musical instruments in the "Audivolving" folder it creates, and copy/paste whats in those files into the "Create Musical Instrument" tab in the options, to play them again, or use them as part of a new music software (not GNU GPL 2+ because it usually does not apply to program output). Audivolv writes reusable Java code.
What will Audivolv do years from now?
* The same thing it does now but it learns faster, and will learn what you like with no good/bad buttons.
* Privacy And Safety Policy is only relevant if you turn on the Internet option or after it has much more intelligence. Does not exist yet.
* Given examples of the best known AI algorithms (bayesian net, neural net, evolution, etc) rewritten in Audivolv's simpler code and data format, Audivolv will evolve an intelligent way to design and use new AI algorithms in that same format.
* Play all types of music and create new types. If you think mouse-music can not rock, listen to my older and less advanced music program: CodeSimian. What I did manually in CodeSimian, Audivolv will do automatically and better.
* Audivolv evolves a way to intelligently design AI brainwaves and data structures it will flow through and modify.
* Audivolv designs a different music-and-mouse language between AI and each person who plays music with their mouse.
* A new kind of communication: Audivolv translates between those music-and-mouse languages (no text) between millions of people across the Internet.
* Theoretically: ToDo Summary.txt in the Audivolv Jar file summarizes my long-term theoretical plans for Audivolv. The most important parts are the "filter" of Java code (also called "firewall"), Friendly AI (for accuracy of the emergent effects), evolution of connectionist algorithms and node types, and the part after "MICROPHONE AND SPEAKERS PARROT" near the end. It will be a 5 megabyte Jar file.




Downloads (just use it to play music, forget the math)

Audivolv's old website has the most info about Audivolv.

Audivolv on Sourceforge.net (Newest version is always here. Example: Version 0.2.1 is a zip file that does not work but has the newest design-documents and in-progress code. Version 0.1.6 is the working version before that. Double-click it to run it)

You can also get Audivolv at download.com

(If it does not work, install Java and try again. If that does not work, please write any problem on the Bug List)

Sample music created with Audivolv

How to use Audivolv,

contact info (same phone number, but I check this email less often. Its not for jobs.)

I'm looking for a job. Here's my resumE.

My other AI software is not finished but there are some demos. Its Javascript so it works immediately in your web browser. It works in the top 5 browsers.
Natural Language Mouse Interface


Audivolv is free and open-source GNU GPL 2+.
Natural Language Mouse Interface is free and open-source GNU GPL 3+.






How I will create the first open-source global AI network.


There is no way to say this without sounding like I'm bragging or exaggerating, but it really is my well-thought-out plans, and anyone can read the design-documents of what I'm planning to build (included in every download, newest docs in unfinished version 0.2.1 on sourceforge). I will create a self-aware self-modifying musical instrument you play with the mouse (and/or other devices) that learns how you want it to sound. It will be artificial-intelligence that creates more artificial-intelligence automatically as you play the mouse-music instrument, using your unconscious musical thoughts as part of its own mind, a collective-intelligence system for creating music. Listen to some music created with Audivolv. The style of the music makes it obvious that an artificial-intelligence created it. The interactions between people and AI in this network will be in the style of Timothy Leary's video: How To Operate Your Brain (click here to watch it inside my Natural Language Mouse Interface software)

OpenEEG

I've been planning this for years, but here's my most recent plan summary, which I wrote at: http://sourceforge.net/projects/openeeg
I've not used it (because the current version is hard to build), but it appears to be the most advanced open-source (no patents or limits on building it) hardware for reading your mind. Its a device you wear on your head that measures electricity on your skin to interpret your thoughts, similar to Emotiv's Epoc mind-reading helmet that they sell for video-games. Last I heard, Emotiv is offering an "enterprise license" to use that mind-reading helmet the same way we can use OpenEEG for free. Of course Emotiv's is better, but thats only because they got there first. I will take OpenEEG far past their video-gaming efforts. My goal is http://en.wikipedia.org/wiki/Collective_intelligence for mouse-music in the first open-source global AI network.



The Nintendo Wii controllers below can detect movement and turning in 3 dimensions, a total of 6 dimensions per controller. Click here to see how I plan to use Nintendo Wii controllers in combination with a mouse and/or mind-reading helmet. A future version of Audivolv will optionally learn to use many hardwares simultaneously to create intelligent music on-the-fly. What a Wii controller can do with motion, a mind-reading helmet can do by thinking about it. Theres some videos of this on youtube and emotiv.com



If Emotiv will agree to sell their mind-reading helmet without restrictions on how it can be used (like I don't need anyone's permission for how to use my mouse or keyboard), then (this would be very hard for me to build) I would upgrade Audivolv to use self-aware mouse-music to speed up their long training that is necessary to learn to use their mind-reading helmet. If not, then lots of years from now, OpenEEG will be taking their customers. Emotiv has created an amazing product (I'll buy one when theres more games for it), but last time I asked them, they are not interested in allowing open-source softwares to use their hardware, so I'm going with OpenEEG. To anyone who wants that to happen, please tell Emotiv you want that to happen and why, and maybe Emotiv will change their mind. For now, I'm going with OpenEEG, but I could use both.

I plan for a future version of my Audivolv software to optionally connect to OpenEEG mind-reading devices (or any other devices or many devices simultaneously) for a similar purpose as emotiv.com uses a mind-reading device as a video-game controller. http://sourceforge.net/projects/audivolv Audivolv (GNU GPL 2+) is the only Artificial-Intelligence that creates musical- instruments you play with the mouse, learns how you want them to sound, and automatically creates and uses new software to make them sound more that way. Thats what it does now. The goal: "Human moves mouse to control realtime generated music which artificial intelligence (AI) changes to try to control Human's hand on mouse. AI translates music problems to programming problems, organizing Humans unconsciously through internet to build AI". Audivolv will become the first open-source global AI network, using simple UDP packets and letting AI organize and optimize the network and what to use the network for (while obeying each user's privacy options). The purpose of extending Audivolv onto the internet is to increase each Audivolv's intelligence by including the unconscious musical thoughts of millions of people who move their mouse to play the realtime generated intelligent music. Generated AI code and people's unconscious musical thoughts, will be searchable by content, but not by name, because this global AI lambda-based system will be too much like Lisp to search anything by name. A future version of Audivolv can do that using only mouse X and Y positions as the user-interface. If combined with OpenEEG, the "Audivolv Network" would be extremely smarter. Unlike the softwares most people think of when they hear "artificial intelligence", the combination of Audivolv and OpenEEG and a million people would be an experience that most people can not imagine. Audivolv+OpenEEG will use subtle psychology of music and mouse movements to form a collective-intelligence of millions or billions of people, free and open-source. http://en.wikipedia.org/wiki/Collective_intelligence

I'm looking for a job. Here's my resumE. Money is not related to building advanced things. Me not having a job today is in no way related to my skills. Its related to people wanting to build less advanced things. Don't miss this opportunity to work with the only person on this planet who will create the first open-source global AI network. --Ben F Rayfield




Downloads (just use it to play music, forget the math)

Audivolv's old website has the most info about Audivolv.

Audivolv on Sourceforge.net (Newest version is always here. Example: Version 0.2.1 is a zip file that does not work but has the newest design-documents and in-progress code. Version 0.1.6 is the working version before that. Double-click it to run it)

You can also get Audivolv at download.com

(If it does not work, install Java and try again.)

Sample music created with Audivolv

How to use Audivolv,

contact info (same phone number, but I check this email less often. Its not for jobs.)

I'm looking for a job. Here's my resumE.

My other AI software is not finished but there are some demos. Its Javascript so it works immediately in your web browser. It works in the top 5 browsers.
Natural Language Mouse Interface


Audivolv is free and open-source GNU GPL 2+.
Natural Language Mouse Interface is free and open-source GNU GPL 3+.

Unquote: And yet when I say "strange loop", I have something else in mind - a less concrete, more elusive notion. What I mean by "strange loop" is - here goes a first stab, anyway - not a physical circuit but an abstract loop in which, in the series of stages that constitute the cycling-around, there is a shift from one level of abstraction (or structure) to another, which feels like an upwards movement in a hierarchy, and yet somehow the successive "upward" shifts turn out to give rise to a closed cycle. That is, despite one's sense of departing ever further from one's origin, one winds up, to one's shock, exactly where one had started out. In short, a strange loop is a paradoxical level-crossing feedback loop. Quote. (pages 101-102 of the book I am a strange loop, a book which I have not read but I know what the word "strange-loop" means)



Example strange-loop: The Halting Problem is a strange-loop. Using only logic (a part of math), its been defined and proven unsolvable, using only a pencil and half a piece of paper. That disproof is a proof of: There are an infinite number of things about math that are true but can never be proven true using only the other parts of math.

Example strange-loop: Creating weapons to work toward peace is a strange-loop. The solution to this is easier than the disproof of the Halting-Problem. the solution is a strange-loop: More people learn how to calculate with strange-loops.

Example strange-loop: Running a Linux emulator inside of Linux, is a strange-loop. Running a Linux emulator inside that emulator is also a strange-loop. How do you know how many emulators you are in? The science-fiction movie The Matrix (3 movies) describes many levels of recursive emulation of Earth, where the machines appeared to know the future but that was only because they had already simulated the future and were replaying the past. The Matrix is a strange-loop. Similarly, I could use recursive self-emulation in my Audivolv AI designs.

Example strange-loop: I'll define some words (different than how Max Tegmark defines them). "Universe" means "all time, space, laws-of-physics, multiverses, and whatever else there is". "Determinism" is everything thats completely predictable. "Nondeterminism" means all "random" and all "chaos" and all "free will". Abstract things do not exist. Math is abstract and infinite. Math is determinism plus nondeterminism. Things that exist can be part of things that do not exist, like "dog" is an abstract definition of an animal, but your specific dog does not have to be abstract. Math is symmetric. In logic talk, "implied" is the timeless word for "caused" or "required". I've not said what exists and what does not exist yet. End of definitions. If universe not equal math, then that lack of symmetry is 1 nondeterministic thing that is not implied by anything, therefore the universe equals math. Math includes nondeterministic things. Nondeterminism is a strange-loop. Theres a technical way to say that: nihilistpanpsychism. This is a proof that the universe equals math (Max Tegmark's theory) and does not exist. Universe is a strange-loop. I wrote on Tegmark's messageboard about this. Maybe he will respond.

Example strange-loop: Time is a strange-loop. However you travel through it, "time" always appears to be "now".

Example strange-loop: Space is a strange-loop. However you travel through it, "space" always appears to be "here".

Example strange-loop: Consciousness is a strange-loop. However your consciousness changes, wherever and whenever it is, "consciousness" always appears to be "me". Its no different than space and time.

Many things in this "reality" appear backwards. It could be related to this 4x4 matrix I found on 1 of Wikipedia's physics pages, where the dimensions on each side are time, spaceX, spaceY, and spaceZ:

-1000
0100
0010
0001

So we are in a matrix? The matrix math gets more complex after that, including things like the square-root of negative numbers.

Example strange-loop: Giving examples to explain what a strange-loop is, is a strange-loop.

Example strange-loop: The software industry has evolved toward the most efficient way to create small bugs (problems in the software) and charge customers money to fix them. When a software business has to choose between fixing their bugs OR adding new features, they make more money by adding the features (with a few bugs to fix later) and charging money to fix bugs later. This is called a "new version" or "tech support" or "support contract". The way bad software makes more money than good software (on average) is a strange-loop. My last version of Audivolv will have exactly 0 bugs.

Example strange-loop: I learned what strange loop means without reading the book I am a strange loop where that word was created.



I am a strange-loop that creates stranger loops. For example...

Downloads (just use it to play music, forget the math)

(If it does not work, install Java and try again.)

Sample music created with Audivolv

How to use Audivolv,

contact info (same phone number, but I check this email less often. Its not for jobs.)

I'm looking for a job. Here's my resumE.

My other AI software is not finished but there are some demos. Its Javascript so it works immediately in your web browser. It works in the top 5 browsers.
Natural Language Mouse Interface


Audivolv is free and open-source GNU GPL 2+.
Natural Language Mouse Interface is free and open-source GNU GPL 3+.

Creating free open-source software and later making money by multilicensing only your parts of that software, is a strange-loop.

This is the new icon for Audivolv:

It means lambda exponent music. Lambda means a stateless-function, which means the same as "functional programming". Its the kind of programming normally done in the Lisp language. "X exponent Y" means the simultaneous interaction of Y quantity of Xs. It does not mean the simultaneous interaction of X quantity of Ys. The way I'm using the word here, "music" means a wave of sound generated by AI. I'm using "music" as a part of math, like you would talk about in signal-processing research, like the SETI@HOME search for alien signals from space is wave research. Waves of light in space (radio signals), or waves in the air we hear in music, are both described by the same math, just different frequencies. Lambda exponent music.

I'm using these 3 words as descriptions of parts of math. For example, "lambda" means the set of all lambda functions. In that way, "lambda exponent lambda" equals "lambda". "math" is the set of all possible things and all impossible things and all combinations of all things. Its just a word. I'm sure that "math exponent math" equals "math". I'm also sure that "lambda exponent music" is a subset of "math". "subset" means "is a part of". Every thing is a subset of itself. I would very much like to know the answer to the following question: Is "math" a subset of "lambda"? By definition, "lambda" is a subset of "math", but are they equal? In other words, what is the technical definition of "math"? Does "lambda" equal "math"? I have a use for this information, but its hard to explain. Its the only research problem I know of that is harder than the Friendly AI problem. The Friendly AI problem is closely related to "free will" in the same way it is related to "pseudorandom" (so it does not matter if "free will" is real or not in how those relate to Friendly AI), and I describe it more below. Also, the word "nondeterminism", as I will use it here, means "random" or "chaos" or "free will". The last question I asked was: Does math equal lambda? I ask a second question here: Does the question "Does math equal lambda?" equal "nondeterminism"? I have a third question: Does the second question equal the third question? These are the kinds of things a Lisp programmer would ask. I use Java instead of Lisp because, over the years, Java has become more self-consistent, and Lisp became more complex as it connected to more self-inconsistent softwares (same reason Windows breaks often). Today, Java is more consistent then Lisp, so Audivolv is completely a Java software. I plan for Audivolv to become the "new Lisp". I will build it into a language for AI and music. Audivolv's Func interface can be used exactly as a lambda (a Func that takes 1 Func as input and returns 1 Func as output), but it also has lambda-like support for efficient floating-point and stack calculations. I've written a technical document explaining an exponential speed optimization for some patterns of code (It will work without you knowing its there). More importantly, I have new theories on the interaction between control-flow (loops and variables), multiply, and exponent. Or forget about the math and simply download Audivolv and play music with your mouse...

Downloads (just use it to play music, forget the math)

(If it does not work, install Java and try again.)

Sample music created with Audivolv

How to use Audivolv,

contact info (same phone number, but I check this email less often. Its not for jobs.)

I'm looking for a job. Here's my resumE.

Audivolv - where Instruments Play The Musicians
A unique Artificial-Intelligence (AI) that evolves Java code to create mouse-music instruments as you prefer them to sound. Free and open-source GNU GPL 2+.

What does that mean? What it is now, what you get in the current version: Audivolv is the only Artificial-Intelligence that creates musical-instruments you play with the mouse, learns how you want them to sound, and automatically creates and uses new software to make them sound more that way.

Really, in the whole history of Earth, Audivolv is the first to do that, and I will continue improving Audivolv. I will NEVER change Audivolv in any way that makes it harder for any person to use, even if that person does not know anything about computers. Use your mouse to play music. Simple. Also, I will never do anything that makes Audivolv less reliable. Unconditionally, I will improve Audivolv toward being easier (as you experience it), more reliable, more predictable, simpler (ratio of code size per feature), and smaller ratio of code size divided by complexity of what it does. For example, it has "Sounds Good" and "Sounds Bad" buttons now, which you use to train it to sound however you want. In some future version, those buttons will be gone and only the mouse movements will remain. Audivolv will learn how you want to sound only by how you move the mouse, based on my theory of Human psychology that if you like a sound more then a smart enough AI must be able to predict your mouse movements (which is your reaction to the realtime generated sounds) better. In other words, your mouse movements and Audivolv's generated sounds will become a 2-way communication between people and AI. You will move the mouse and it will change from "whitenoise like radio-static" to any musical instrument you can imagine, for the same effect as Audivolv reading your musical thoughts. Its a very subtle thing about psychology. Its 100% Java, works immediately without install. Sets up all options automatically. If you are not evolving audio 10 seconds after it starts, please write on the Bug List. If you do not know how to use it, please write why its confusing on the Bug List. Starts as whitenoise like radio-static, then you teach it what music is with your mouse Click the link above to try it now. Its fun, easy, and fast. (If it does not work, install Java and try again.)

I'm looking for a job. Here's my resumE.

What is AI? Its any software that does unexpected things and surprises you how well the unexpected things work compared to what it was designed to do. We expected the Deep-Blue chess computer to win, but we could not predict what chess moves it would do until it played against the best Human chess player.

Most people think AI is something done only by scientists that takes lots of tedious work, is hard to use, and is better to watch on television than to play with. My softwares: Audivolv and Natural Language Mouse Interface will be simple and fun for you to use. Forget the instructions. Its Human-nature to use it the right way. Audivolv is almost that way now. I'll keep working on both of them, but first I need to get a software job.

Downloads

(If it does not work, install Java and try again.)

Sample music created with Audivolv

How to use Audivolv,

contact info (same phone number, but I check this email less often. Its not for jobs.)

a NEWS article about a different AI software that creates music. Audivolv is different because Audivolv uses the mouse instead of microphone to learn, but future versions will do both.
"Instruments on the verge of a musical revolution"

Summary of the technical details of my plans for the final version years from now. This summary and the detailed technical documents (in every download of Audivolv) should be enough to explain to any expert software-engineer how Audivolv works and my plans for future Audivolv versions. (TODO reorganize this website) You can read over 100 pages of technical documents (included in every download if you open it with WinZip or other unzipping software), and more updated in the most recent version you can get on sourceforge.net. Every download also includes the complete source-code and a buildAudivolv.xml file you can run with Apache Ant to create a new Jar file from the unzipped files. In that way, Audivolv is a strange-loop.





You can always easily download Audivolv now and use it with a mouse. Or in later versions (could be years from now), hold 2 Wii controllers (12 dimensions), or 5 people each hold 2 Wii controllers (60 dimensions). Instead of listening to the music and dancing to that, dance to create the music, which will create different kinds of dancing. In that way, you and AI together will create a strange-loop that creates more strange-loops and creates better AI and music software. Theoretically, by using your movements and music as a 2-way communication between people and AI, AI can learn how your unconscious mind works and copy parts of that into new AI software.

Here are some open-source softwares that connect those Wii controllers wirelessly to a computer. Audivolv does not have to obey the licenses of those softwares for the same reason Audivolv does not have to obey a license for your mouse or keyboard. They are only controllers and no direct software connection is needed. Each Wii controller can detect acceleration and turning in 3 dimensions. Thats 6 dimensions per controller. Future versions of Audivolv will evolve music using many devices simultaneously.


Audivolv will always be easy for you to use. What I write below is for Friendly-AI programmers to understand. You do not need to understand it to use it to evolve mouse-music. Just move your mouse to teach it how you want it to sound. It will always be that easy. In some later version, I will remove those "Sounds Good" and "Sounds Bad" training buttons, to make it easier, and instead Audivolv learn will predict how you play mouse-music that you do or do not like, instead of you using those training buttons. Audivolv will become easier to use than it is today, never harder.

I am adding new code to Audivolv very slowly because my goals in creating it include it becoming a Friendly AI. Every Friendly-AI is an Artificial General Intelligence (AGI), and every AGI is an Artificial Intelligence (AI). Thats the definitions. AGI is dangerous, and Friendly-AI is the best theory (that we know of) how to safely create an AGI. Create the first Friendly-AI before creating the first AGI that is not a Friendly AI. The problem of how to create a Friendly-AI is probably harder than all problems any person has ever solved. It is closely related to what some people call "free will", and how to create an AI that wants to do certain things even when it has a choice not to and after it modifies itself many times and still it would choose to work toward its original goal. Yudkowsky says it as: "Gandhi does not want to commit murder, and does not want to modify himself to commit murder." The Friendly AI problem is at least as hard as knowing how Gandhi avoids self-deception when changing his mind about things. In some ways, Gandhi would be a good prototype to copy into an AI design, but that would require we know how Human brains work. Instead, we can only use vague ideas about him in AI designs. Thats the definitions and basic information about what Audivolv (and some other softwares) are for. Heres some software design ideas:

Because I took a lot of time to think about how to create Friendly AI, I now have have a very small idea of how to evolve code that tends to evolve toward becoming Friendly-AI a very small amount. Last year, I had no idea at all how to define Friendly-AI in any technical way in the smallest amount. This is progress:

I expect that if I continued building what Audivolv's existing design documents describe, and add 1 thing, that it would tend to evolve toward Friendly AI a very small amount sometimes and usually not. That 1 thing is, when it recursively-self-improves (as the design documents describe how to do), add a requirement that it predict (with more accuracy than last time, on average or median?) what new behaviors it will have after it creates variations of different parts of itself (recursive changes to be observed by other parts of itself earlier in that recursion, partial self-awareness), and in this lower-recursion-modifying-structure of AI code and data, when any small thing is predicted more than a little inaccurately, then a large penalty flows up the tree (the AI code that created the unpredictable AI code and/or data and its behaviors) to cause it to learn to predict its actions more accurately in the future on average. Also, do more than 1 of these, independently and in parallel, and if they do not come to almost the same solution, then do a similar penalty flowing up the recursion. Its not enough just to predict what it will do. It should be directed toward the right thing to do, and if the "right thing to do" is ambiguous (could be more than 1 thing), then whatever caused it to come to such a choice should cause a penalty flowing the same way up the AI-modifying-and-creating-AI-tree. The root goal of all this must be what the Human user wants it to do, which will be derived from Audivolv's ability to predict your mouse movements in reaction to the sounds Audivolv is generating in reaction to your mouse movements and its thoughts about your previous mouse movements and other things. This is not certain, not a proof, and probably only a very small amount of progress in my limited understanding of Friendly AI theory. This is a small change to the design of Audivolv and does not contradict any of its design documents. It only constricts what options I had left open. As a result of this change, I expect Audivolv will increase its intelligence much slower at first and much faster after that. This is something I did not expect: Working toward Friendly-AI makes the AI design be more efficient. If I was holding myself to the same standards as this AI design, then I would say the fact I did not expect to write it that way means it is a worse design than if I did expect to write it that way. It does not mean its a bad design. Its a continuous interpretation of worse, much worse, incredibly bad... This is an example to show you what I mean by recursion. I am a person recursing my AI designs into this text, which may later recurse AI designs into Audivolv, which may later build more AI and recursively predict and measure it similarly. To become a complete Friendly-AI, it would have to wrap around and predict itself and the person playing mouse-music. It would have to be completely self-aware. I can create a self-aware AI, but Friendly AI is much harder. It has to predict when its important to predict things, and which things, and lots of other vaguely-defined things... Unlike Gandhi, Friendly-AI has to understand its own ways of thinking in a more technical way because it changes them more. Thats the only way to make a self-aware self-modifying AI be safe.

What does Audivolv do now?
* Starts when you click here then click OPEN. If it does not work, click here for how to fix it or my phone/email.
* Open-source GNU GPL 2+. Unzip that file instead of opening it to get the source-code.
* Sound reacts instantly to mouse movement.
* You move the mouse any way you like and think about how that should sound.
* You choose "Sounds Good" or "Sounds Bad".
* Audivolv tries to learn what you were thinking.
* Repeat until your musical thoughts become real. It took 5 minutes to create this example music: .OGG .MP3 .WAV
* Close Audivolv and start over if you confuse it too much.
* Audivolv is a Java software that creates Java software to change things about itself as it learns
* Learns techno music faster than any other type, but what did you expect an AI to play first?
* There is not 1 line of music code in Audivolv until you teach it what music is. Then it writes the code.
* Find all new musical instruments in the "Audivolving" folder it creates, and copy/paste whats in those files into the "Create Musical Instrument" tab in the options, to play them again, or use them as part of a new music software (not GNU GPL 2+ because it usually does not apply to program output). Audivolv writes reusable Java code.
What will Audivolv do years from now?
* The same thing it does now but it learns faster, and will learn what you like with no good/bad buttons.
* Privacy And Safety Policy is only relevant if you turn on the Internet option or after it has much more intelligence. Does not exist yet.
* Given examples of the best known AI algorithms (bayesian net, neural net, evolution, etc) rewritten in Audivolv's simpler code and data format, Audivolv will evolve an intelligent way to design and use new AI algorithms in that same format.
* Play all types of music and create new types. If you think mouse-music can not rock, listen to my older and less advanced music program: CodeSimian. What I did manually in CodeSimian, Audivolv will do automatically and better.
* Audivolv evolves a way to intelligently design AI brainwaves and data structures it will flow through and modify.
* Audivolv designs a different music-and-mouse language between AI and each person who plays music with their mouse.
* A new kind of communication: Audivolv translates between those music-and-mouse languages (no text) between millions of people across the Internet.
* Theoretically: ToDo Summary.txt in the Audivolv Jar file summarizes my long-term theoretical plans for Audivolv. The most important parts are the "filter" of Java code (also called "firewall"), Friendly AI (for accuracy of the emergent effects), evolution of connectionist algorithms and node types, and the part after "MICROPHONE AND SPEAKERS PARROT" near the end. It will be a 5 megabyte Jar file.

Old Audivolv website

(Audivolv is open-source only)

This small page

NEW: Natural Language Mouse Interface,
will become Audivolv's new website.
(open-source and business licenses available)

Audivolv's old website has the most info about Audivolv.

Audivolv on Sourceforge.net

You can also get Audivolv at download.com
has downloads,

sample music,

basic instructions,

info about the new website I'm building,

contact info,

links to my other websites,

a NEWS article "Instruments on the verge of a musical revolution"

and a summary of the technical details of my plans for the final version years from now.
Compatible with the top 5 browsers (maybe more): InternetExplorer, MozillaFirefox, Opera, Safari, and GoogleChrome.

Click here to try Natural Language Mouse Interface 0.7 in your browser. When it gets to version 1.0, I will replace this website with it, and the following will be true:


Or try version 0.10 which has only been tested on Firefox in Linux.

Natural Language Mouse Interface (NLMI) is Javascript - no server needed - that enables dynamic movement of text, or other web components, using intelligent neural network algorithms based on mouse movements. Move your mouse over 1 or more texts that interest you, and the most relevant text will move toward your most recently selected texts. If the relevant text is not on the screen, it will appear and replace other irrelevant text, which may come back later when it is relevant to something else you select. Wikipedia page comparing Javascript frameworks.

Natural Language Mouse Interface on Sourceforge.net
LICENSE: open-source GNU GPL 2+.
(All software was made by me except Javassist3.9)

CONTACT:
benrayfield@mousesearch3d.com 408.334.7214 (call me any time)
View my resumE
I'm looking for a software job near San Jose, California, USA. --Ben Rayfield.


To start Audivolv now, click here then click OPEN.

(or click SAVE instead of OPEN to download it)
Sample music created with Audivolv: .OGG .MP3 .WAV

If it does not work, install Java and try again. If that does not work,
you can immediately describe any problems on the Bug List, or call/email me.

Sample music: .OGG .MP3 .WAV, played using this evolved Java code that Audivolv wrote.

Audivolv is open-source GNU GPL 2+.
To get the source-code, SAVE that file as x.zip and unzip it.


How to use it: Move your mouse to play whitenoise and other random sounds. Audivolv continues learning how you want the mouse to sound. You can only play music with your mouse after you teach Audivolv what music is. The current version has good/bad buttons to teach it, but future versions will learn what you like by your mouse movements alone.





Read this NEWS about "Instruments on the verge of a musical revolution" (not about my software, that will be later).

Instruments on the verge of a musical revolution 13 May 2008

The world�s first intelligent digital musical instrument that can learn to make music in the style of any musician playing it has been developed by experts at the University of Plymouth.

The research � undertaken by PhD student Marcelo Gimenes and Professor Eduardo Miranda at the Interdisciplinary Centre for Computer Music Research (ICCMR) at Plymouth � means that musicians will soon be able to jam with machines that hold virtual musical clones of themselves, the profile of others musicians, or both.

Unlike existing technology that needs to be pre-programmed to make music, the new software developed at the ICCMR �listens� to the musician and learns their style so that it can improvise with them.

When it listens to the musician, the software creates a memory from interconnected streams of data. Incoming musical information is compared with the data already stored in the system and the memory is subsequently updated on the fly.

The ICCMR is developing new intelligent musical systems that will be able to evolve their own rules for musical composition and the ability to interact with musicians and listeners in much more sophisticated ways than at present.

Professor Eduardo Miranda, Head of the ICCMR, said: "We predict the emergence of new kinds of intelligent musical instruments that could revolutionise the music industry. Such instruments will be able to learn basic musical skills autonomously as they are played. New forms of music making will certainly emerge along with new business opportunities."



IN EXTREMELY TECHNICAL WORDS, MY LONG-TERM THEORETICAL GOAL FOR EACH AUDIVOLV PER COMPUTER: efficiently unify connectionist AI (the obs[] part of a audivolv.Func) (any recursive permutations of linear and exponential types) with evolution (the obs[] part of a audivolv.Func) and hypercube vector fields (the flos[] part of a audivolv.Func or any flo numbers recursively in permutations of linear and exponential connectionist array networks), represented in the same code and provably-predictable data format (arrays of arrays of arrays... allowing cycles and leaf nodes like audivolv.Func or numbers in hypercube range, with other constraints), code-string-firewalled for safety, so generated AI softwares can use other generated AI softwares as tools, starting from a few example AIs like bayesian (a type of exponential connectionist AI) and evolution, to play better mouse-music (a hypercube vector field that includes 1 dimension for each speaker and mouse x and y position).


THEORETICALLY: THIS PART COULD BE DONE ON 1 COMPUTER OR A GLOBAL AI NETWORK:
To an AI, mouse-music would look like an AI changing the data described above. To a person, mouse-music would sound like theres a person on the other end of the mouse helping with the music. A common language between people and AI, to use AI as a person emulator and people as an AI emulator. Its much easier to do both of those things together than only 1. Audivolv's long name is in quotes: "Audivolv - where Instruments Play The Musicians" and Musicians Play The Instruments, a theory of recursive intelligence, where Human thoughts and AI thoughts could be used interchangably in the same recursion, using mouse-music as "Human interface code" instead of normal "Human interface code" like windows, text, buttons, and menus. I know brains run much slower and more parallel than computers. There can be many speeds of recursion, some allowing more time to think at each step, some slow enough to flow across a global AI network, and others fast enough to process realtime audio. Recursive intelligence can be slow or fast, blurry or specific, and any combination of those, in theory.


THEORETICALLY, technical parts of later versions of Audivolv, and other subjects: ArrayAlign | ArrayAlignHierarchy | ArrayAlign_ConstantRange | ArrayAlign_Equal | ArrayAlign_Multiply | ArrayAlign_PermutationCycle | ArrayAlign_Power | AudivolvNetwork | AudivolvSummary | Brainwaves | CodeSimian | CodeTree | Func | FuncPurpose_MeasureOtherFunc | FuncPurpose_SoundColorMouse | FuncPurpose_Unknown | HardCoded_CodeTree_BreedFunction | HeapQueue | HighStandards | Hypercube | Javassist | LongTermGoalsOfAudivolvDesign | MindReadingHelmet | MovementScore | NeatVsScruffy | NoSpying | Node | NormedCodeTree | Exponentially Fast Permutation Partial-Compiler And Interpreter (Nobody has ever built one of these optimized for AI) | Plugin | ShortTermGoalsOfAudivolvDesign | SoundCardPart | Template | TuringTests_per_second | flo | ob


































Audivolv - where Instruments Play The Musicians