Thursday, April 25, 2013

Serverus: Evolvable System of Systems PaaS Services

One day, certain components in the programmer's development stack will be fully self contained ai's. The programming languages they use, their compilers, interpreters, intermediate languages, and virtual machines, along with their ide's, server apps, and the networks they deploy on - down to the very computing interfaces they use - will all eventually evolve into self contained ai's that solve their problems domains completly. This article describes the Sysosys System of Systems Evolving Architecture Model and how any firm that adopts it has a stake in the AI business. Evolvable components in an evolving system of systems desire: To self extract into project starter kits any components of which it is composed, and also any systems which can be composed using it as a child component. To be fully composed of decoupled atomic modules. To be as succinct as possible. To be obvious in design. To flow as logical statements. To be as simple as possible. A.I. Knowledge Systems, of which Watson is the current forerunner, can be used to run a company.  Business processes, data, interactions, workflows, can all be thought of as programs which can be piped together by the Knowledge System as needed to run the firm's day to day operarations, under the oversight and tuning of their human stakeholders and architects. Evolving Apps from API's. Using human guided genetic algorithms to assist in app development at every point of differentiation . Apps start their lives with points of similarity, and points of differentiation. At every point of differentiation, employ a string genetic algorithm that seeks to discover the appropriate string to put there. Fitness is evaluated by whether the underlying java files build, install, run, pass test scripts, and a survive a human assisted curation step. 

Custom app development can be driven by growing interfaces for 3rd party api's, adding genes for browser automation, and minimizing the difference between native and web application screenshots, treating programmer supplied uri's as seeds. Recog: Replacing Programmers With Artificially Intelligent Programming Entities.

Or why I had to create my own dev team to get anything done.

One could argue I am simply proposing better, more intelligent tools.  Nay, I am orchestrating the creativity that flows from the void.

The why
solidarity of vision
low cost
no politics

The how
Genetic Algorithms
Bayesian Inference
Kallman Filter
Monte Carlo Method

No comments:

Post a Comment