14 Sep 2023

USI

EMPLOYMENT OF ARTIFICIAL INTELLIGENCE (AI)

AT TACTICAL LEVEL

Why AI took off since 2010

• Exponential growth in computer processing power
• Increased availability of large datasets of “big data” sources upon which to train machine learning systems
• Decreased cost of storing data, cloud.
• Advances in the implementation of machine learning techniques
• Significant and rapidly increasing commercial investment

Data Policy

Availability of Data

• The Army is a data-rich organisation with vast amounts of information on the OE, training events, combat operations and soldier readiness. Data rich does not mean data ready
• Much of the Army’s data can be characterized as “dark data,” sitting in a silo and accessible only for limited, single-use purposes
• To get AI ready for the future, the Army requires a Service-wide effort to break down these silos and import all data into an open and shared architecture that is accessible by a range of AI tools.

DOTMLPF-P Integration

Select Social Media Manipulation Tactics and Potential Implications of Generative AI

LLM, Chat GPT

• Large Language Models (LLMs) and generative artificial intelligence (GenAI) are the latest and potentially most significant technological advances in this time of rapid technological change. They represent a step change in AI’s capabilities and a fundamentally different and promising path in AI’s development.
• IBM PC first shipped in 1981, and the iconic Apple Mac followed in 1984. It would take nearly a decade, before there were 100 million PCs in use. Six months after its release, there were already 100 million ChatGPT users.
• Gen AI is on pace to achieve the speed of diffusion in one year which the Internet took seven years to realise.
• We have seen OpenAI’s release of GPT-4 in March 2023, with significantly advanced capabilities compared to GPT-3.5, released in November 2022

​

Generative Artificial Intelligence

• Generative artificial intelligence (AI) describes algorithms (such as ChatGPT) that can be used to create new content, including audio, code, images, text, simulations, and videos. Recent breakthroughs in the field have the potential to drastically change the way we approach content creation.
• Even in its current nascent stage, generative AI is already powerful enough to help military personnel access faster and more data and tools; generate text and initial recommendations for memos, plans, orders and information campaigns; and prepare elements of planning for legal and operational review.
• As generative AI models become increasingly sophisticated, they will further help synthesise and display data from different datasets and multiple modalities.
• Finely-tuned models could deliver advanced battlespace awareness, including visualisations of the battlespace and provide commanders with a detailed and updated common operational picture.
• Generative AI could also conduct real-time fusion, correlation and pattern analysis of massive volumes of data and adaptively orchestrate novel attacks or defences with minimal human intervention.

​

​

Technical limitations to building improved, next-generation models and� to improving some of their basic architectural limitations.

• Foundational LLMs such as GPT-3 and GPT-4 are not cheap to build; the latter reportedly cost $100 million to develop
• Deploying them can also be extremely costly: running ChatGPT might cost $100,000 per day
• Training the next generation of larger and higher-performance models is likely to be very expensive
• Technical limitations on cost for building and deployment are specific to LLMs. Text-to-image models, such as Stable Diffusion, are much cheaper to train and can be deployed on personal computers
• It is not clear just how much real-world impact social media influence campaigns have, regardless of how convincing they might be.

​

China

• PRC’s publicly released LLMs have lagged behind those of American companies – such as OpenAI, Google, and Anthropic – but access to open-source platforms and cloud infrastructure could enable PRC firms to accelerate the development of their domestic models.
• Beijing currently relies on U.S.-designed AI chips despite actively pursuing self sufficiency for years. Training and inference for LLMs and AI applications currently requires large-scale supercomputers which are powered by clusters of thousands of specialized AI chips called Graphical Processing Units (GPUs). American export controls have aimed to limit PRC access to these advanced microelectronics. U.S. firm Nvidia accounted for as much as 95 percent of China’s GPU market.
• PRC’s domestic chip manufacturing capabilities at advanced nodes (e.g., 7nm) are presently operating at low yields, though Beijing remains determined to catch up
• PRC faces a number of systemic challenges in developing and deploying its own LLMs. Scarcity of data available for training LLMs in Mandarin Chinese, as less than two percent of the Internet is in Chinese compared to nearly 60 percent in English.
• PRC’s domestic Internet domain is firewalled away from outside influence, minimiSing the exchange of digital information with the rest of the world.
• China maintains access to foreign data via products like TikTok, potentially an advantage for its AI industry.

Elephant in the Room

China‘s embrace of intelligentisation�

PLA is fusing mechanisation, informatisation, and “intelligentisation” into a new way of war purpose. PLA has set its sights on leveraging AI and big data, use influence operations and employ swarm attacks.

​

​

​

How Does Generative AI Change the Game for Chinese �Social Media Manipulation Against Taiwan?

Use of AI in Ukraine

• Ukraine’s highly dynamic battlefield where opposing sides adapt their methods and systems on a daily basis. Traditional military equipment is paired with the newest technologies in innovative ways
• Ubiquitous sensors, big data processing, automated orchestration platforms and drones employed at scale are dramatically compressing detection-to-destruction are introducing new ways of warfighting.
• Software systems such as GIS Arta, Delta and Kropyva have allowed Ukraine to carefully marshal its resources across large and geographically disconnected fronts
• Ukraine has achieved local numerical superiority at a place and time of a commander’s choosing – significantly enhancing both lethality and survivability at the tactical level.
• Dominance in information and thus speed and precision, can win over dominance in numbers.

​

Implications of Artificial Intelligence and Machine Learning for Cybersecurity

Use of AI by Bad Actors

• Advances in artificial intelligence (AI) have lowered the barrier to entry for both its constructive and destructive uses.
• Low-cost, commercial off-the-shelf AI means that a range of nonstate actors can increasingly adopt these technologies
• Acquisition of AI-based technologies by nonstate actors threatens to destabilize existing state-nonstate dynamics on the battlefield.

​

​

Regulating the AI

• Not why but when.
• U.S. government has turned its attention to GenAI, seeking outside expertise on how to address the novel risks and opportunities presented by the technology.
• White House announced a “voluntary commitment” from the seven foremost GenAI companies to ensure safety, security, and trust in their models prior to public release.

​

Talent Mgt

ERIC SCHMIDT IN DEFENSE INNOVATION BOARD MEETING,�ON U.S. TECHNOLOGY TALENT RECRUITMENT

“The talent is there, but I have a feeling they’re not coming to you. Trying not to be blunt . . . Why aren’t they coming to the Department of Defense? There is the obvious money problem.

Silicon Valley artificial intelligence specialists are paid salaries up into the millions and tech giants are increasingly competitive as they vie against each other to recruit the best and the brightest in hard to fill technological fields.

This has turned the current civilian technology arms race into a race for

human talent, which is far more scare than either data or computing power.”

Examples of AI Projects Undertaken

Role of Big Tech

Conclusion

• War in Ukraine revealed that cognitive warfare and cyber warfare conducted in non-physical domains — do not alone provide strategic advantages.
• Neither Sun Tzu, who idealised subduing the enemy without fighting nor British strategist B.H. Liddell Hart, who advocated the indirect approach strategy, gave specific advice on how to put it into practice.
• In the long history of warfare, it has been physical battles that subdued the enemy’s will.

WHO HAS WON

THANK YOU

SELECTED AI DEFINITIONS

Points for Consideration�

• What current Armed Forces policies are prohibiting or hindering our adaptation of AI systems?
• How can the Armed Forces allow for the authorized routine access of data to support AI systems while ensuring necessary security and privacy?
• What network or bandwidth challenges must be overcome to realise the Armed Forces routine use of AI systems?
• Will preparing Armed Forces personnel to use AI systems increase or decrease training requirements?
• What must we do today in our recruiting and accessions programs to ensure that we have the right human capital for the future?