HUMANS REMAIN AT THE CORE

These developments raise questions about whether AI will threaten human roles in the military. But there is much to suggest that the human element will remain crucial for the foreseeable future.

Minister Chan said at a Jun 24 dialogue that, although AI accelerates decision-making in defence, such as in acquiring targets and selecting weapon systems, there must still be a “man in the loop”.

“It cannot be so autonomous, whereby you just take decisions, and you don’t know what you are hitting,” he said. He explained that an overreliance on technology is tantamount to a “dereliction of duty... to an algorithm that you do not understand”.

This is especially so when such decisions in a combat situation can have literal life-or-death consequences. For a conscript armed forces such as Singapore’s, the vigilance, training and commitment of military personnel are the deciding factors of mission success, more so than just having the best-in-class equipment and weaponry.

Another factor to consider is that most AI expertise currently resides with the private sector. Any military seeking to adopt AI must build deep partnerships with the broader industry, companies that are developing AI, and third-party evaluators who are experts in AI capabilities and limitations.

In particular, the effective use of AI will require addressing many of its current flaws. These include a tendency towards subtle biases and hallucinations, potential for sycophancy or telling the user what they want to hear, and an inability to make decisions in line with ethical considerations. These have a significant impact on the outcome of military operations.

As such, the military will need to set policies, guardrails and procedures for AI systems, and also provide training for human users to ensure that AI does not lead to damaging mistakes.

In these instances, it is impossible, and potentially foolhardy, to take humans out of the loop, given the need for human-to-human interactions and the ability to make nuanced decisions.