The need to maximize performance permeates military spending worldwide. In the US, Lockheed has been at the forefront of technological innovation for over 60 years and is leveraging AI to save American lives and optimize the country’s military outcomes.
Lockheed’s focus on AI is evidenced in its creation of LAIC (Lockheed Artificial Intelligence Center), a centralized organization to facilitate the company’s development of AI tech. One of the latest developments sponsored by LAIC is the deployment of the AI Factory, which was announced in October 2022 and will help Lockheed implement AI at scale. Through its AI Factory, Lockheed will standardize tools and processes for all divisions and functions so they can efficiently build AI models, essentially streamlining access to ML/AI. The components of Lockheed’s AI Factory are:
- InfraOps: Enables the AI Factory to be scalable and run across different computing infrastructures, hence training and productizing models flexibly.
- DataOps: Allows engineers to manage complex data life cycles and features such as data storage, labeling, and versioning. For example, automated software developed by the AI Factory allow data labelling processes to be reduced from two weeks to under one day.
- MLOps: Includes standardized training and model registry functionalities.
Lockheed’s goal through the deployment of its AI Factory is to make model development cycles shorter and provide all the necessary tools for AI and ML engineers to focus on model development rather than worrying about infrastructure and operational needs.
With the base layer infrastructure that the AI Factory provides, Lockheed expects AI to be fully integrated company wide, and plans to build AI models that create value across multiple use cases. These include:
- Internal Operations and Manufacturing: With the goal of increasing productivity, reducing waste, and minimizing costs, Lockheed is implementing AI models in its manufacturing facilities. For example, they have developed an AI-enabled scheduling tool that optimizes machine and operator workflows and furthers factory automation. Their goal is to create a digital twin of the production lines, which will facilitate (virtual) AB testing of new initiatives with minimum capex and schedule disruptions while allowing for automated prediction of maintenance needs.
- Mission Execution Services: Lockheed is focused on enhancing human performance by leveraging AI as a capability multiplier.
Generally, Mission Execution Services allow for the deployment of AI across 3 types of teams:
- Humans Working Alongside Machines: These are teams in which there is a constant radio frequency connection between the controller and the machine. The latest example of AI implementation in these types of teams is the Squad Mission Support System, which is an autonomous vehicle loaded with supplies and gear for infantry in long-range missions, essentially allowing more supplies to be carried with lower physical strain.
- Humans Remotely Piloting Machines: The most common use cases for AI in these scenarios are surveillance, safety, and efficiency. An example is Indago, a 5lbs drone operated remotely which can be used for automated mapping and firefighting.
- Humans Travelling Within Machines: For example, the Sikorsky MATRIX Technology provides helicopter pilots with a virtual second pilot. This means that forces deploying Black Hawk helicopters will have the option to send the vehicle to a mission completely unmanned, even if the default setting is for pilots to be in control. This will allow logistics and resupply missions in conflict hotspots to be carried out without risking human lives, despite obstacle-rich environments. MATRIX also reduces lifecycle cost and operating cost per flight hour. The Black Hawk in the video below was operated by MATRIX without human intervention for over 80 miles:
As evidenced by the use cases above, Lockheed faces tremendous opportunities if it fully integrates its AI Factory company wide. First, through the Internal Operations and Manufacturing use cases, Lockheed will be able to push the frontier from an efficiency and cost perspective, hopefully minimizing defense costs for taxpayers while building a more robust and flexible defense supply chain. Similarly, Lockheed will be able to position itself as a defense provider that enhances capabilities of human team members in America’s armed forces, making operations safer and cheaper. The integrated characteristic of Lockheed’s offerings provide a positive network effect: the more sensors and devices a government procures from the firm, the more integrated all combat systems will be and the better the data and AI models deployed in combat operations.
However, Lockheed’s journey is not without challenges. First, the use of AI in defense applications is still a regulatory grey area and can be significantly limited by both domestic and international regulation in the same way nuclear, chemical, and biological warfare was regulated. Weapon systems with the ability to hunt, choose to engage, and eliminate human life, without human inherence, can be extremely dangerous if they were to fall into the wrong hands. Moreover, who determines which are the right hands to wield this kind of power?
Another challenge Lockheed faces is the lack of AI technical talent willing to work in defense applications when compared with more traditional tech giants such as Google and Meta, which are in cities with a higher concentration of recent graduates, offer a potentially more inspiring mission when compared to military applications, and have higher compensation.
To de-risk the potential for regulatory impact and allow for greater AI/ML development through increased talent attraction, Lockheed should centralize all of its AI efforts into an internal unit and spin it off under different branding. This will allow the new firm to attract the technical expertise required by isolating the AI function from Lockheed’s more defense-oriented mission, while providing AI services to Lockheed and other companies (not necessarily defense oriented) on contract basis. Similarly, the new firm should be able to lobby government officials independently, in order to make sure the appropriate regulations are in place to allow for a safe rollout of AI-enabled defense technologies to lower cost and improve the safety of America’s military operations.