The Use of AI in the U.S. Military - Timeline

Militaries will need to be cautious about employing AI systems. Getting an AI trained doesn’t take long; it only took 40 days for AlphaGo Zero to train itself to beat the world champion beating AI, AlphaGo Lee (Hassabis & Silver, 2017). AI will need to be well-trained and notably better than existing systems before they are used in war. Despite the hype, most implementations are still in the playing-around phase. We are still at the dawn of AI adoption by the military. Missy Cummings, director of Duke’s Humans and Autonomy Lab, plainly states that regardless of the increase in interest, research, and development of AI systems by militaries around the world, adoption has been “incremental” and organizations are “struggling to make the leap from development to operational implementation.” (Cummings, 2017)

During development, integration testing will reveal potential issues. Some of the obvious concerns have already been addressed in this article. The risks of these issues may be deemed acceptable. There is a huge risk of falling behind in this sector. A lot of the issues will be resolved on production systems with rapid and automated patch management. Militaries will have to determine the trade-off. Sure, there are notable reliability issues. However, the capabilities and enhancements brought by AI are notable. Hopefully, they’ll lean towards reliability before using these new systems on the battlefield. However, as we say in the Information Technology field, “you won’t find all the bugs until you put it in the wild.”

References

• Cummings, M. (2017). Artificial Intelligence and the Future of Warfare. Chatham House.
• Hassabis, D., & Silver, D. (2017, October 18). AlphaGo Zero: Learning from Scratch. Retrieved from DeepMind: https://deepmind.com/blog/alphago-zero-learning-scratch/