AI Meets Open Source

(Source:  klss/Shutterstock.com)

Artificial intelligence (AI) is being used by many to accomplish great things beyond human intelligence. Open source platforms, data, frameworks, and models are increasingly used in conjunction with AI development to improve and enhance AI projects.

The premise of open source is that everything is free and available to all. Source code, designs, and related intellectual property is shared and may be redistributed at large. It represents an open exchange where users participate and collaborate in a communal effort. Programmers use source code to program a software application. With the source code, programmers and developers use it to perform and reach desired objectives. How does the open source movement impact AI applications? Let’s explore. 

Interoperability and Communal Sharing for Progress

In AI applications development and the emergence of machine learning, the open source movement is more important than ever; major software purveyors and software developers across a wide variety of industries are contributing their own source code and using other’s code.

Open source code meets a common open source standard. As an enterprise develops open source code with open source standards, and without proprietary data formats, the resultant software is compatible with other software and applications. This enables interoperability.

A communal effort ensures interoperability. Interoperability is key. Interoperability means compatibility and an ability to integrate an application with other applications, allowing enterprise networks to grow and prosper from software.

Open source code and software is often loaded to a common platform, available to all, such as GitHub. While it's a Microsoft site, it's meant to be a working parking lot for open source code and self-described as: " a development platform inspired by the way you work. From open source to business, you can host and review code, manage projects, and build software alongside 50 million developers.

Another popular platform to share code is via the Apache Software Foundation (ASF).  It provides framework for source code where those committed to the open source credo of sharing intellectual property may do so, without worry of infringement or liability. 

The foundation touts "The Apache Way," as one where "more than 730 individual Members and 7,000 Committers successfully collaborate to develop freely available enterprise-grade software, benefiting millions of users worldwide."

In AI, a data framework uses data to create predicted outputs. Such prediction is used to learn data and continuously train the model. This takes place via a feedback loop. Data is learned. Prediction is enabled. What’s more, data and predictions are continuously fed back and improve the accuracy of the prediction.

An example of how open source data tools are used in everyday applications is one employed by the ride-sharing company Lyft Inc. Most recently the company open-sourced a a debugging tool for artificial intelligence data it had been using for years, internal to its operations.  The tool called Flyte, has been used by Lyft in-house for the past three user and corrects and debugs data related to Lyft's pricing, locations, estimate time of arrivals, mapping and self-driving developer teams. It is open-source and can be downloaded and used by anyone.

Open Source Frameworks for AI

For AI, there are many open source frameworks. These frameworks are available to all, used by many, and result in the creation of applications that may be proprietary, but are based on an open source framework.

For example, a popular and useful open-source framework is Google’s TensorFlow. This framework is a comprehensive ecosystem of tools and libraries that enables developers to apply machine learning to and better serve their markets. TensorFlow, for example, is used by Airbnb, eBay, Dropbox and others. The TensorFlow, unbeknownst to many, is what allows rental listings to follow an order suitable to a user as that user and their preferences are indicated per their profile and search. It may aid eBay to better present auction listings. Or it may establish a file hierarchy that follows a certain pattern for a Dropbox user.

Amazon SageMaker Neo is another open source machine-learning platform. Its project code helps AI developers to build models that learn data and trains a machine-learning model via the cloud. It's compatible with sensors and other computing and connected devices used with the Internet of Things (IoT) applications. It may be used for AI developers to make accurate predictions.

For example, SyntheticGestalt (an applied machine-learning company) uses SageMaker Neo to train drug discovery models used in the pharmaceutical and life-science industries. It processes and learns experiment data, evaluates it, and produce model results.

The Open Neural Network Exchange (ONNX) is an open-source artificial intelligence ecosystem, created specifically for AI to represent machine-learning models. Its operands are common and serve as basic building blocks for machine-learning and deep-learning models. This aids AI developers utilizing formats to subsequently allow models to work with different frameworks and tools for AI applications. Like other open source platforms and formats, ONNX enables interoperability of, and between, other applications and solutions. By developing within a preferred framework, it curtails problems from incompatibility further downstream of the application.

Our Open Source Future

As AI continues to gain traction to accomplish great things, the open source mindset will continue to help accelerate it. It uses the power of a community of give-and-take who use open source data property: data, algorithms, platforms frameworks, and formats, to perform amazing feats.

In the future, open source platforms will help technology use math, science, and other technology to improve our world in new and different ways, via artificial intelligence; in this way we will move beyond what we ever thought possible.