Artificial Intelligence (AI) is in its early stages of development, but in the next 15 years, it will touch nearly every business and consumer operation. One of the central features of AI development is the various software frameworks that make it possible. We are starting to see significant rationalization of frameworks; however, it is unlikely a de-facto framework will emerge, given the many AI use cases such a framework would need to cater for. TensorFlow is gaining momentum in terms of developer use, but the framework is not optimized to support all use cases and work over any hardware technology. As a result, if not implemented well, the framework could induce huge performance issues.
This report describes the fundamentals of the current state of AI and explains how it is being applied and developed. In this report, ABI Research introduces some benchmarks to assess which frameworks are likely to be most used in the long run, as well as giving a top-down overview of the framework ecosystem, highlighting some critical issues for those player involved to address. The report also covers how frameworks and hardware vendors are beginning to address the issue of shifting AI out to the edge and onto devices. Not all frameworks are designed to support deep learning (DL) and AI implementation on the edge. As AI is moving from the cloud to distributed intelligence, frameworks need to adapt to the new environment.
Table of Contents:
AI and Framework Overview
What Makes a Good Framework/Commercial Support
Review of Select Frameworks/Libraries/Toolsets
Edge AI 54Trends/Conclusions