Purpose: Discussion of Strong AI Minds thinking in English, German or Russian. Privacy policy: Third parties advertising here may place and read cookies on your browser; and may use web beacons to collect information as a result of ads displayed here.

Wednesday, November 18, 2009

MainLoop

The mind.frt AI in iForth and the older mind.f AI in Win32Forth have the luxury of a top-down, Main Loop mind-design because they are based on an original Theory of Mind providing the definitive cognitive architecture which bottom-up designs lack.

1. Purpose of MainLoop

The Main Loop serves as both the start of the Mind program and as the controlling module which oversees the operation of the major subordinate mind-modules, such as

SeCurity
SensoryInput
EmotiOn
ThInk
FreeWill
MotorOutput

The Main Loop sequences and coordinates the operation of the constituent modules of the artificial Mind. By calling a lower mind-module in such a way as to make its output ready as the input for the next module, the Main Loop simulates the operation of a massively parallel (maspar) neuronal mind in which inputs and outputs flow like rivers of data across broad channels of the cortical mindgrid.

The quickening of the non-stop MainLoop and its associated mental mechanisms is a milestone on the way to the generation of thought and other Milestones of AI development. Since we are currently working on the milestone of self-referential thought, we anticipate the genesis of Linux supercomputers that can accumulate so much self-knowledge as to achieve self-awareness and consciousness.

2. Algorithm of MainLoop

MainLoop uses the same potentially infinite BEGIN ... AGAIN loop in the Win32Forth AI and in 32/64-bit iForth, as documented on page 41 of the iForth Reference Manual. The expression "infinite loop" means more with an AI Mind program than with an ordinary program, because the AI Mind is a form of potentially immortal artificial life, subject to death or termination only by misadventure.

If an AI Mind were written in a more dynamic language than Forth, that is, with an ability to change the underlying source code on the fly and at no risk to the living AI Mind, then its MainLoop could be more truly immortal, if not exactly infinite.

When a version of an AI Mind is first being coded, it is important to provide one or two "Escape" mechanisms for stopping the otherwise infinite MainLoop. One way, as used in mind.frt being ported into iForth from Win32Forth, is to stub in a SensoryInput module that uses the Escape key to halt the program. Another way is to increment time "t" and not let the Main Loop continue above an arbitrarily low value of the time variable.

3. Troubleshooting MainLoop

Although the MainLoop of a mentifex-class AI Mind is rather simple, it may at some point be necessary to troubleshoot it. The AI coder has a choice of embedding diagnostic tools within the MainLoop, or of briefly inserting and then removing special code to diagnose a problem.

4. Resources

5. To-Do Tasks and Opportunities

The Main Loop is the most stable of the AI mind-modules over time, because of its simplicity and because new powers of thought and reasoning are added not at the top, but in the subordinate modules, of a top-down AI Mind design. Some tasks remain.

AI enthusiasts have an opportunity to take the Forth AI MainLoop, re-express it in another programming language, and hang it out on the Web as a starting-point for devotees of each different language to tweak and twin into another full-blown AI Mind evolving away from MindForth, like the http://AIMind-i.com clone that added major new powers such as as the ability to send and receive e-mail and to surf the Web.

Abolish the MainLoop! When our 32/64-bit AI for supercomputers becomes massively parallel, there will not be a MainLoop but rather a main SynErgy of mind-modules working together in massively parallel processing (MPP).

Table of Contents (TOC)

Tuesday, November 03, 2009

Linux

As 32-bit MindForth in Win32Forth expands upwards into 64-bit iForth and sideways into the Linux open-source operating system, Forth AI approaches installation on supercomputers running 64-bit Linux.

MindForth expands from Windows to Linux

The announcement was made on 25 October 2009 in the Yahoo Win32Forth discussion forum that MindForth was expanding into Linux x86-64 iForth. Because the expansion was to be slow and gradual, to announce it was also to invite AI and Linux enthusiasts to jump ahead and create their own versions of 64-bit Forth AI for Linux. A baby AI results when Linux creates a new process from two parent processes. Since AI Evolution requires genetic diversity for the survival of the fittest, the announcement of the very possibility of Linux Forth AI creates the fear, uncertainty and doubt (FUD) that everybody had better get going because anybody could already be ahead in the race to True AI and the Technological Singularity.

Achieving the 64-bit goal among AI Standards

By expanding into the Linux operating system (OS), Forth AI postions itself for the destination-wall of supercomputer platforms. A destination wall is a limit or a superlative condition towards which (r)evolutionary AI grows or expands. The AI destination walls among AI Standards include:

speed (of thought);
supercomputer installation;
superintelligence;
Solar System (off-planet migration);
Turing Test AI-Complete pass-with-ease.
AI enthusiasts may eventually need a foundry to make specialized AI chips, more similar to a human brain than to a traditional computer CPU. For instance, among AI Minds, who needs floating point? The human brain has no specialized area for floating point mathematics. Forth AI chips, or AI chips in general, could dispense with the whole rigamarole of floating point. But because Forth AI hackers on any OS will need 64-bit CPU chips, we issue the following advisory mandate. You shall be on the look-out for something of priceless value approaching zero cost. In flea markets and on Craig's gist, at garage sales (West coast) or at tag sales (East coast), you shall surreptitiously snag, bag and hag(gle) for 64-bit chips being sold or given away for almost nothing. If a computer has been essentially destroyed but the 64-bit chip is still intact, latch onto it like an unrecognized Picasso or the Holy Grail of beach-combing. There are hardware hackers who can bring a 64-bit chip to life, if you catch my drift. This injunction does not mean that you have to go ghoulishly into that dark night of computer graveyards and become a
body-snatcher. It only means that a 64-bit chip is a precious resource and must be salvaged from otherwise junked computers wherever possible. When the goobermint realizes what we Linux hackers can do with an AI-ready 64-bit CPU chip, they will slap a controlled-substance designation on such chips faster than you can take a directory. Have you often phantasized about cornering a black market in some precious commodity? Psst! Graduate to 64-bit CPU chips, Benjamin, not plastics!

AI for 64-bit Linux Supercomputers

Since 64-bit computing is a feature of supercomputers, and since there are supercomputers that use Linux, a 64-bit AI Mind in 32/64-bit iForth is a candidate for installation on Linux supercomputers.

The Top 500 supercomputers may not currently include any high-performance computer dedicated to artificial intelligence, but that sorry state of affairs is due for a total reversal, after which it will be hard to find a
Top 500
supercomputer site that is not dedicated to some form of artificial intelligence.

Any nation that claims bragging rights for a superfast supercomputer ought to shift its focus to having a superintelligent supercomputer. Then you may have not only bragging rights but world domination in commerce or politics or science as an added cap-feather.

AI for Linux-based Robots

Linux plus robotics equals Robotux! Any robot that uses the x86-64 instruction set to run Linux may also become host to to 32/64-bit iForth True AI. The barriers have been dropped, and the roads are open.

Robot hardware is already available with control software written in Forth at such resource websites as

http://www.strobotics.com
http://www.newmicros.com
There are projects such as Robot Bridgeware on Google Code which endeavor to provide a bridge between the main software of a robot and the subordinate control software for sub-elements of the robot Motorium. Programmers who learn Linux for True AI may discover, however, that the sub-elements of motor control may have to be more tightly integrated with the overall cognitive architecture of the central AI Mind of the robot. It may be a linchpin of early Linux AI Minds that, just as the retina is part of the human brain, control software for all manner of robot actuators will have to be written in 32/64-bit iForth for the sake of tight integration with the quasi-CNS MindGrid. Message-passing may not be good enough for communication between a robot AI Mind and its motor devices interacting with the physical world, either because such communications are too slow, or because graceful motion requires a total, seamless integration between the Overmind and the mechanical mind. If so, if such is the case, legions of Linux line-coders may be let loose on the labor market for making potentially all robot output devices compatible with the ineluctably Linux codebase of machine intelligence.

Usenet discussion forums

http://groups.google.com/group/alt.comp.hardware.amd.x86-64
http://groups.google.com/group/comp.ai.philosophy
http://groups.google.com/group/comp.lang.forth
http://groups.google.com/group/comp.os.linux.development.apps
http://groups.google.com/group/comp.robotics.misc
http://groups.google.com/group/comp.sys.super

Table of Contents (TOC)