Evolutionary Adaptive Machine Intelligence - EAMI. A bassic form of ancient ML that reveals it is ideal.
- Although it is less-advanced, it still reinforces the question: What purpose is best-suited an ideal system?
Due to the age of the books and what computing hardware/software was state-of-the-art at the time of their printing, the demonstratable focus of Evolutionary-Adaptive Machine Intelligence (EAMI) is contained in Basic scripts run on the TRS-80 computer. The low-level of these algorithms are an excellent place to begin with metal.
For the purposes of posterity, the scripts as written in these books have been captured in their original TRS-Basic scripts. In "machine intelligence for your home computer" (1982), there is also code in AppleBasic. As it is super-inconvient to try to assemble such a computer (the Rodney robot is certainly difficult enough), an amazing project by George Phillips that has a group on Discord is a TRS-80 emulator that allows running of the scripts. It runs in Linux, Windows, and Apple machines. By using the "Tandy Color Basic" VsCode extension, one can have the same programming experience as the modern languages.
A recent Windows version used in past experiments (v.2.4.8) is contained in the emulator folder. The most recent published version is contained in the zip archive and it contains the versions for the various platforms.
In the emulator folder are some binaries of the application. What is being currently used for testing is a stand-alone application in the arm64 folder. Run a script in the following manner:
./trs80gp experiments/alpha1/games/03-alpha1-paint.bas
And you will see
Another:
./trs80gp experiments/alpha1/data/04-alpha1-compilation.bas will ask you for parameters, such as the number of cycles to run:
Then run the analysis:
Perform its intuitive-themed task:
And output the results of the agents' work:
The meaning implied in the data, Chapter Six of (1982), is:
<-talk about the data and its meaning/context here->
Last summer (2022) a set of Alpha-1 experiments were run with the emulator. It was discovered due to the length of runtime of some routines, it was a good idea to run it on an embedded board. The first choice was a NVIDIA Jetson that kept shutting down about four minutes into the running script. The next choice was a raspberry pi. As the version 3s are not suitable, an order was placed for a pi-4 that only arrived a few weeks ago. This project will revisit these experiments using the emulator on the pi-4.
The Rasperry-Pi 5 is suitable to run the code cleanly where animated gifs can capture sessional data.
It was defintely worth it. Having tried since 2016, the hardware is sophisticated enough to run in an energy-efficient context.