China train AI-general to predict ‘enemy humans’ on the battlefield

In an enthralling development that feels like it's straight out of a futuristic sci-fi saga, reports from the South China Morning Post reveal that scientists in China are on the brink of a revolutionary breakthrough in military technology.

These innovators are allegedly fine-tuning an artificial intelligence (AI) akin to ChatGPT, but with a twist – this AI is being designed to anticipate the actions of potential adversaries on the battlefield. The minds behind this cutting-edge project belong to the People’s Liberation Army (PLA) Strategic Support Force, employing Baidu’s Ernie and iFlyTek’s Spark, both large language models paralleling OpenAI’s renowned ChatGPT, to bring this vision to life.

Imagine an AI that can predict human behavior in combat, a concept that once resided purely in the realm of fiction. According to the SCMP, this isn’t just imagination but a tangible project where AI is fed a treasure trove of sensor data and frontline reports, either in text or images. This data is then processed and sent to commercial large language models for interpretation. Following confirmation of understanding, the military AI crafts prompts for further discussions, such as those concerning combat simulations, all without human intervention.

The Project

The project, led by Sun Yifeng and his team from the PLA’s Information Engineering University, stands as a pioneering acknowledgment by the Chinese military of employing commercial large language models. They argue that their work could elevate both human decision-making and the AI’s combat knowledge, refining its understanding of warfare to new heights. Their goal? To create a military AI that mirrors human intuition, capable of grasively grasping the unpredictable and adaptable nature of human adversaries – a challenge that has long puzzled machines.

The implications of this are staggering, drawing parallels to scenarios right out of "Star Wars: The Clone Wars," where tactical droids face defeat against cunning human strategies. By integrating large language models, the team hopes to better anticipate human actions, an ambition that has already shown promise in simulations like a reimagined US military invasion of Libya in 2011, where the AI successfully predicted US military movements.

Yet, as groundbreaking as this may be, Sun’s team openly acknowledges the limitations of current commercial LLMs in warfare contexts. Their solution? A multi-modal approach that enhances predictions with practical, detailed analysis and reports, pushing the boundaries of AI's application in military strategy.

Despite the leaps in innovation, this venture into militarized AI, as outlined by a Beijing computer scientist to the SCMP, invites caution. The potential for AI in military applications, while inevitable, brings with it the shadow of sci-fi doomsday scenarios, a reminder of the fine line between advancement and overreach.

As we stand on the precipice of this new frontier, one thing is clear: the race towards integrating AI into military strategy marks a pivotal moment in the evolution of warfare. With each step forward, we inch closer to a world where the line between human and machine strategy blurs, heralding a future where the chessboard of global conflict may be forever changed.