A timeline on developing AI solutions The recommendation of the National Security Commission on AI mandates that “ by 2025 , the foundations for widespread integration of AI across DoD must be in place ...” has introduced a heightened importance on time as a factor in the US Navy ’ s AI calculations . This is a daunting goal from any perspective , and , according to Vaughan , necessitates an expansive adoption strategy across a broad range of disciplines .
It ’ s a challenge – and , with only three years to go until the deadline , there ’ s still much work to be done . “ The clock is ticking ; it ’ s a very short game clock . We have to rapidly and broadly progress to an elevated posture that is conducive to delivering AI at speed and scale , and all the services have that challenge .”
But scaling itself is tricky , he explains . “ The algorithm we build for a submariner is going to be different from the one we build for a jet pilot or somebody on a surface Navy ship . They are different environments and have different data ; the transport layers are different ; the operations are different . All those elements comprise the ecosystem you need to deploy AI effectively , so it ' s very important that you tune AI development within the intended use mission or problem set . At the same time you don ’ t want to get pigeon-holed or siloed in AI approach ; you want to be able to extend AI-enabled solutions across the enterprise . In order to bridge this space , to effectively craft AI for the warfighter while setting conditions for broader exploitation , the US Navy has established AI Task Forces in every one of its warfare enterprises – an accomplishment that Vaughan is especially proud of as it provides purchase on the cultural challenges of AI adoption .
Scaling AI means that “ if somebody has developed AI capabilities , say for submariners , I can now take that solution and extend it to other pieces of the enterprise and potentially boot-strap other warfare enterprises as well . Scaling means efficient recapitalisation ; not reinventing the wheel every time you develop an AI pipeline or a piece of code , you can leverage or recapitalise that work across the enterprise .”
Vaughan says that it ' s essential to have the team skilled enough to undertake the task . But that team must also have a viable , resilient and robust digital fabric underneath . He points out that AI is very heavily reliant on compute , such as cloud and highperformance compute assets , transport layers , and the network .
“ You have to move ones and zeros around from where they ' re collected to where they need to be processed , to a decision maker or to a robotic system . All of that substrate needs to be in place , ready and resilient , so there ’ s a great deal of work ahead of us ,” he concedes but maintains confidence that the critical benchmarks will be met .
technologymagazine . com 47