Speech by Senior Minister of State for Defence, Mr Heng Chee How, for Joint Grant Awards Ceremony of the Robust AI and AI for Materials Discovery Grand Challenges on 26 July 2023.

Introduction

A very good morning and welcome everyone to the Joint Grant Awards Ceremony of the Robust AI and AI for Materials Discovery Grand Challenges this morning.

I am delighted to be here to join you. This is the first time MINDEF is partnering AISG to conduct Grand Challenges in Artificial Intelligence (AI) and I am glad that many talented scientists and engineers have participated in this event.

Pervasiveness of AI in Society

The field of AI is advancing at a rapid pace and that we all know it. Since November last year, Generative AI technologies like those in ChatGPT has taken the world by storm, answering all kinds of questions, and making it easy for anyone who aspire to be a poet or an artist and more.

AI has clearly transcended automation, to enable capabilities that are more powerful such as the creation of novel designs and generation of new content. AI is transforming many industries, for example in healthcare, AI is enabling better diagnoses and personalised treatments; in transportation, companies are gearing towards self-driving vehicles to improve road safety by reducing the risk of human error; and in education, AI-powered chatbots are being used to provide personalised learning support to students.

As AI advances, its adoption rate has also skyrocketed. Instagram took 30 months to reach 100 million monthly active users; TikTok took 9 months to reach 100 million and ChatGPT took just 2 months. The spread is very rapid, the reach will be very pervasive.

AI for Defence Applications

At MINDEF, we are actively seeking ways to use AI to intelligently serve both our warfighting and organisational needs.

The Army will be fielding unmanned ground vehicles for logistics support; this will free up precious manpower to many other important tasks. The Air Force is developing automatic detection and classification of runway damage with the use of drones; this will increase operational efficiency and enhance personnel safety. The Navy has deployed coastal security systems that are equipped with advanced video analytics to spot very small objects such as fishing boats or swimmers at distance, much faster and more accurately than the human eye. Our fourth service, the Digital and Intelligence Service, is actively employing detection analytics to hunt cyber threats for the SAF; and to address servicemen attrition due to musculoskeletal injuries, the SAF Medical Corps developed DASH which is an analytics tool that identifies causes of musculoskeletal injuries from a huge body of clinical notes so that timely interventions can be enacted. This is good both for the soldiers as well as for the overall fighting strength of the SAF.

MINDEF's investments in AI will continue to grow to support our Defence needs, and our larger national needs where there are synergies. The advancement of AI will open up new possibilities and MINDEF is poised to harness its power to build important capabilities for our future.

The use of AI, especially Generative AI, is however, not without its challenges. One of the biggest problems of ChatGPT is the phenomenon called "AI hallucination", where the AI might invent false or misleading information based on their incorrect understanding of the question asked. This could result in serious consequences when the technology is used to inform critical decisions. The power of Generative AI can also be used by nefarious actors to write malware and generate fake news at scale. Paradoxically, AI may also be the antidote to such AI-generated attacks. For example, our partner, DSO, has developed an automated tool that ranks retrieved documents according to relevance and uses natural language inference to highlight possible fake news. As MINDEF pushes the boundaries of AI technologies, we are also concurrently fortifying our defences, ensuring robustness of our AI systems against potential attacks.

Partnering IHL/RIs and Industry

Harnessing the power of AI requires a different approach from our traditional defence R&D, where a lot of the work is done in-house. AI has permeated various industries and sectors but no single entity holds a monopoly over this technology. While commercial entities such as OpenAI and Microsoft have brought large language models to the fore, academia and AI enthusiasts around the world have rallied together to build upon open-source LLMs and this has given rise to small but powerful models that are almost as good as, and sometimes even better than commercial ones. MINDEF's approach is therefore to collaborate extensively with local and global partners from across research institutes, and institutes of higher learning and industries. This is where MINDEF's Defence Technology Community, which designs and develops technological capabilities for the SAF, will play a key role in this effort.

On our global AI partnerships, we have recently established a Joint Lab for AI with France's Ministry of the Armed Forces and continue to enjoy good technical exchanges with the United States. On our local partnerships, we continue to explore new avenues to tap on the growing local tech ecosystem. We have collaborated with National Research Foundation to launch open grant calls, and that has yielded promising ideas on collaborative decision making, trustworthy and explainable AI. Moving forward, we will also be expanding our partnerships with industry, to accelerate progress in developing AI technologies and applications. I encourage the various communities to come together to co-ideate and co-innovate. I would like to invite all researchers and practitioners with bold ideas to reach out to us. Together, let's collaborate, innovate and build an exciting and better future.

AI Grand Challenges for Defence Applications

This time, we are glad to work with NRF again to shape the two Grand Challenges, one on "Robust AI" and the other on "AI for Materials Discovery".

These two challenges are not only important to MINDEF, but also to the rest of the world. As AI becomes prevalent in our daily lives, trust issues are bound to surface. How can I be sure that the car I am travelling in, that is on autopilot mode, will get me to my destination safely, without say, misrecognising a stop sign that has been inconspicuously tampered with for a speed limit sign? The first Grand Challenge on "Robust AI" aims to address the vulnerabilities of AI models in computer vision systems for autonomous vehicles. This is a tough problem that many advanced companies in self-driving technology are still grappling with and have yet to comprehensively resolve.

One of the key advantages of AI over its human creators is the ability to recognise patterns hidden within copious amounts of data very quickly. We can use this to achieve important breakthroughs in scientific fields such as in advanced materials. For example, are we able to find a material for the chassis of a drone that is lighter and tougher, more resistant to heat and mechanical stress? That leads me to the second Grand Challenge on "AI for Materials Discovery". This challenge aims to accelerate inverse material design, to discover advanced materials that are 50% lighter while retaining or enhancing their functional properties. If successful, AI would not only speed up the discovery of new materials but could also bring down R&D costs.

Conclusion

This joint grant awards ceremony marks the start of a very exciting journey for all five teams. We will be awarding funding to these five teams to carry out research to solve these challenges. I thank the participants, our partner DSO, the organising team at the Future Systems and Technology Directorate from MINDEF HQ, and AISG for making the Grand Challenges possible.

Last but not least, I wish all five teams a fruitful and gratifying experience, and look forward to your solutions.

Thank you everyone.