AI OS Introduction

Aichoo AIOS edited this page Jan 25, 2018 · 5 revisions

AI OS provides an intelligent framework for data science process, feature engineering and for creating AI/ML based applications driven by data.
The benefit of having a standardised framework is that it simplifies the process allowing non-experts to perform advanced tasks and experts can achieve more in less time by concentrating on more advanced activities.
It can also be viewed as a unified test harness for creating and recording ML experiments. Most data scientists create their own test harnesses, which is a significant exercise on its own. Our objective it to make AIOS in such a way that it will save data scientists time and effort as well as make their work portable and shareable.

Problem: modern enterprises collect enormous amount of structured and unstructured data expecting that one day that data will be useful. To make that data useful is still a big challenge despite the availability of machine learning algorithms, big data tools, Google/Microsoft/Amazon/IBM Watson and other vendors’ cognitive APIs. There is a wide disparity between what normal businesses expect from ML/AI/cognitive tools and what is actually being provided. Clients expect a truly intelligent solution but in reality they are presented with a set of unrelated and not compatible blocks, without a well-defined path how to put them all together and achieve the desired outcome.

Solution: our product is a complete new platform for creating cognitive solutions in an enterprise with minimum effort, which we call “AI Operating System (AI OS)”. It is an Operating System because, like its counterparts Linux/Windows etc., it is a kernel whose purpose is to intelligently combine and orchestrate other data processing and cognitive blocks which, within the AI OS context, are called Agents.
AI OS is an operating system for the next generation of computing. Previous generation of OS, examples of which are Linux\Windows, was created due to the need to manage a variety of hardware resources by abstracting the complexity of hardware in such a way that users could create end-user applications without knowing all the intricacies of hardware. We are now entering the next generation of computing, whereby it is all about data analysis and machine intelligence, and a similar need has now arisen in this generation.
Computing power is now not on local hardware but distributed globally. Data processing services are also web services that can be located anywhere. There are thousands of libraries and services that do data processing, all with unique interfaces. Writing bespoke software to manage all of the above is akin to programming early computers that had no standard OS. Such exercise is simply beyond capabilities or strategic objectives of most companies. AI OS is an attempt to solve this need for the next generation of computing.

Outcome: AI OS consists of a kernel, which resembles an organic brain whose objective is to continuously build and improve internal representations of available data. On its own the kernel is limited in its cognitive capabilities. However, its power lies is in its ability to utilise knowledge and expertise of agents connected to it. Agents themselves are algorithms that implement specific data processing.
AI OS is the kernel and a collection of customizable agents. An agent is any process that transforms data. Agent could be as simple as one that simply applies a basic math function to input data and produces an answer. On the other side of the scale, an agent could be a very complicated procedure which looks for and classifies images/videos and provides an output back to the kernel.
All Google/Microsoft/Amazon/IBM Watson services can be connected to the AI OS as Agents, thus providing their expert knowledge. Any other data processing service can also be connected as Agent of the AI OS. More agents AI OS has available, more intelligent it becomes and more capable it is in creating better representations of all the data provided to it by an enterprise. Better representations naturally lead to better ability to make decisions on the data needed for fulfilling enterprise’s objectives.

Typical AI OS users no longer need to spend time and effort on the myriad of complex data science activities because most of those will be performed by the AI OS. Users just need to point to data sources, make sure they have agents connected, provide computing power resources and specify decision making targets.

Advanced users of the AI OS may develop additional agents for specific data processing functions which have not yet been implemented. This can easily be done in Python within the AI OS templates, or connected as external services (like Google/Amazon/Watson services) again via provided templates.

Source code of the AI OS Kernel is currently kept private. Nevertheless, the AI OS will soon be available as SaaS on our website, free to use for evaluation and for non-commercial purposes.

This repository contains code for Agents that is open for everyone to develop and publish for others to use. All Agents developed by contributors will be available within AI OS but will have a flag indicating whether an agent is checked and approved by us.

You can’t perform that action at this time.
You signed in with another tab or window. Reload to refresh your session. You signed out in another tab or window. Reload to refresh your session.
Press h to open a hovercard with more details.