The United States Army Research Laboratory is experimenting to see whether OpenAI’s generative AI solutions can help battle planning, but within a military video game.
The New Scientist on Wednesday reported how US Army researchers are using OpenAI’s GPT-4 Turbo and GPT-4 Vision models to provide information about simulated battlefield terrain and details on friendly and enemy forces, as well as military lessons on attacking and defending. They also used two other AI models based on older technology. After that, they gave the AI assistants a mission to destroy all enemy forces and seize an objective point.
The AI assistants immediately outputted many courses of action, at which point, a user playing the role of a commander asked the model to refine that output. While OpenAI’s GPT models worked better than the other two models, they also caused more casualties while carrying out mission objectives.
The use of generative AI is just one small part of the US Army’s push to take full advantage of artificial intelligence in their strategy. Project Maven, the US Department of Defense’s flagship AI effort, has located rocket launchers in Yemen and surface vessels in the Res Sea and helped narrow targets for strikes in Iraq and Syria, reported Bloomberg in February.
But this potential use of AI on battlefields also raise many ethical concerns. The prospect of leaving decisions that could potentially kill people to machines is more reminiscent of the Terminator series of movies than that of a bright future espoused by AI evangelists. But that has not stopped the military. The Pentagon asked US lawmakers for billions of dollars to develop its artificial intelligence and networking capabilities, reported DefenseScoop in March. It has also established a Chief Digital and AI Officer positions to help use and spread the technology across the department.
ChatGPT-maker OpenAI confronting a mountain of legal challenges
After a year of basking in global fame, the San Francisco company OpenAI is now confronting a multitude of challenges that could threaten its position at the vanguard of artificial intelligence research. Some of its conflicts stem from decisions made well before the debut of ChatGPT, particularly its unusual shift from an idealistic nonprofit to a big business backed by billions of dollars in investments.
It’s too early to tell if OpenAI and its attorneys will beat back a barrage of lawsuits from Elon Musk, The New York Times and bestselling novelists such as John Grisham, not to mention escalating scrutiny from government regulators, or if any of it will stick.