You are here

Technical report | Recent Advances in Artificial Intelligence and their Impact on Defence

There have recently been a variety of high-profile demonstrations of artificial intelligence (AI) — with significant progress being made in fields as diverse as self-driving cars, game-playing machines and virtual assistants. In this report we discuss some of the recent breakthroughs in AI research, and explore some of the opportunities these provide within the Australian Defence Force (ADF) context. This paper is intended to contribute to both the dialogue around the use of AI in the ADF, as well as to provide a useful resource for ADF members to enhance their education and understanding about the technologies of artificial intelligence, with a particular focus on deep learning.

We begin with a high-level summary of the history of AI research, to provide some context to the current wave of AI development.

We discuss the drivers for the current growth in AI interest and, in particular, introduce the field of deep learning and the reason for its exponential growth and dramatic successes over the last decade. The success of deep learning has been driven by three main catalysts: data, computation and algorithms. The availability of increasingly large data sets, coupled with readily available and massive computational resources, has enabled the development of a variety of algorithms to solve real world problems, which only a decade ago seemed intractable.

We present five significant problem-domains that have seen rapid advances during the last decade, and discuss the drivers for these developments and prospects for future successes. These application areas were not practical for machines prior to the recent growth in deep learning. These are:

  1. image understanding
  2. intelligent decision making
  3. artificial creativity
  4. natural language processing
  5. physical automation.

This list is not exhaustive, and does not reflect the breadth of the AI field, but each of these areas has shown significant and rapid change over the last decade and are likely to see further successes moving forward.

We also discuss the potential applications for these techniques in the military domain. We argue that, to avoid losing its capability edge in the future, the ADF needs to invest in a number of areas that will be critical for future AI systems. In order to embrace the potential of AI, there will need to be a significant cultural shift in the way that military data are generated, captured, stored and processed. In addition, Defence needs to invest heavily in high performance computing. However, given the inability for Defence to replicate the data or computational resources of the commercial AI industry as it currently exists, the ADF also needs to invest in research into algorithmic improvements that maximise data and computational efficiency. In addition, with its legacy systems and complex environment, the ADF needs to carefully consider elements of system integration to fully employ AI technologies into the future. Finally, to ensure the ethical use of AI and contribute to the worldwide debate, the ADF needs to carefully consider social and ethical issues around the employment of AI in military operations.

Key information

Author

Glennn Moy, Slava Shekh, Martin Oxenham and Simon Ellis-Steinborner

Publication number

DST-Group-TR-3716

Publication type

Technical report

Publish Date

May 2020

Classification

Unclassified - public release

Keywords

Artificial intelligence, machine learning, deep learning