At a military installation in central California, autonomous all-terrain vehicles are being put through their paces—not for human operators but to train advanced AI systems for potential use in conflict zones. This initiative is led by Scout AI, a startup founded in 2024 by Coby Adcock and Collin Otis, which recently secured $100 million in Series A funding following an earlier $15 million seed round.
Scout AI’s flagship project is an AI model named “Fury,” designed to manage military assets, initially focusing on logistical operations and potentially extending to autonomous weaponry. CTO Collin Otis likens the training of these systems to the preparation of soldiers, asserting that it’s vital to start with a foundation of intelligence similar to what a young soldier would possess.
The startup has already attracted $11 million in military technology contracts from agencies like DARPA and the Army Applications Laboratory. Its technology is among those being trialed by the US Army’s 1st Cavalry Division, with expectations for further deployment in 2027. Scout’s testing involves real-world missions simulated by former soldiers, utilising challenging hilly terrain to assess the vehicle’s capabilities.
Emphasising advancements in autonomy, Scout is employing innovative Vision Language Action models, which enhance the decision-making abilities of robotic systems. These technologies, first developed by Google DeepMind, are intended to improve the military’s pool of autonomous vehicles.
While autonomous vehicles are increasingly common in urban settings, operating effectively in the unstructured, unpredictable contexts of a warzone presents unique challenges. Otis, a former executive with Kodiak (an autonomous trucking company), realised existing technologies fell short in terms of reliability for battlefield conditions, driving him to co-found Scout.
The firm’s first planned product is “Ox,” a command and control software that will enable soldiers to manage multiple drones and vehicles more effectively. The training of these autonomous systems is conducted in real scenarios, where operators refine the models through a reinforcement learning system, revealing their limits and refining performance.
Scout’s ambitions extend into the realm of munitions with drone systems that could autonomously engage targets under defined parameters, an evolution of longstanding military technology practices. While the deployment of AI-driven weapon systems raises ethical concerns, Scout asserts that programmable constraints can significantly mitigate risks.
Adcock and Otis envisage a future where their advancements might outpace traditional contenders in the quest for Artificial General Intelligence. The duo maintains that the continual interaction with real-world scenarios will yield a more practical application of AI compared to solely theoretical knowledge. However, Scout aims to complement existing technologies rather than compete directly, positioning itself uniquely within the military tech landscape.
Fanpage: TechArena.au
Watch more about AI – Artificial Intelligence
