 Thank you. It's very good to be here today. I'm happy to be here to celebrate all your accomplishments. I have to tell you as an old M60 tanker I'm both amazed and excited by all the things I've seen around here as an old tanker I'm also rather intimidated by the mathematics involved. I hope there'll be no quizzes On behalf of General Kaufman who today is on the Hill advocating for funding for the types of technologies we're interested in and For all of our partners at the Detroit Arsenal. Thank you again for inviting me to talk here today This country's recent jump into autonomous ground vehicles can be traced directly to the ARL and DARPA Grand Challenge in the Mojave on March 13th 2004 which challenged industry to develop autonomous vehicle technologies While this challenge asked for applications to military vehicles It helped create a research community that has developed fleets of autonomous car prototypes That are driving our roads today. Your work here has set the foundation for our future military efforts But while we have come a long way Commercial application is still well off and we have a much further way to go for robotic combat vehicles Getting an autonomous car to pick up your groceries or deliver you to the airport is hard It's harder than we predicted Adverse weather conditions diffuse lights and pesky Inconsistent human interactions are preventing us from seeing autonomous cars take over our streets We're seeing the start of a heartbeat and I have no doubt that the commercial industry will figure out the last 20% To provide an autonomous fleet of vehicles while that road is hard our objectives are much harder Combat vehicles need to be able to operate cross-country in all weather conditions Robotic combat vehicles can't get stuck in the sand mud snow water or thick vegetation But they also cannot pick the optimal route for navigation because that's where our enemy is lying in wait For centuries, we've been training our soldiers how to do this. How do we train our machines? The US Army and our peer threats operate in other spectrums as well. These spectrums are contested We cannot use technology that broadcasts our positions if I put a commercial automotive lighter on a tank for navigation I might as well say I'm over here shoot at me Nor do I want my robotic combat vehicle to be hacked and turned on me That would be the 21st century definition of a bad day at the war The future robotic combat vehicle purpose is to hunt Not to deliver people or products in a highly map destination It has to be robust and resilience and perhaps the biggest obstacle boils down to one word trust Just as we will not trust autonomous cars until they drive as well or better than humans Our soldiers will remain skeptical of unmanned ground combat vehicles until they prove that they are a trusted agent To conduct their share of the mission as well as if humans were standing in their place So how do we get there? As you know our autonomy efforts are a scale We do not seek to remove human decision-making from warfighting. No one wants a Skynet What we need to do is to efficiently automate human tasks to allow humans to spend more time on what we are Intrinsically good at and less time on what tasks and machines can do Theoretically that will allow for a higher level of human and machine output Right now we are at a two-to-one control ratio of a tele-operated vehicle Over time we want to increase the autonomy and reduce the human interventions Until we are able to provide a one to twelve human control to robotic combat vehicle ratio While we do many things better than robots, we must realize that machines do things we cannot Machines do not need sleep nor do they require food or water or uncontaminated air They do not feel fear or exhaustion and when they die nobody will miss them except maybe the accountants And that is perhaps their greatest quality. They can be destroyed instead of our sons and daughters being injured and killed The goal is to develop AI to support multi-domain operations against peer Russia and China primarily Threats if you don't know they are very serious about it. We must be too Our current military commanders have digital systems that provide a certain level of friendly force tracking and shared understanding It's better than radio calls and putting flags in a map But we rely on those too because the digital system is not robust More broadly than robotic combat vehicles a challenge you to help us move our situational awareness and targeting capabilities from these simple map displays that required 96 hours to synchronize a brigade or division targeting cycle to 96 seconds with the assistance of AI We need an AI enabled web that is resident in the next generation of combat vehicles and command posts That allows a commander be it in a turret a command post or in his goggles Well dismounted to see the battlefield and decide the most lethal way to destroy our enemies and protect our soldiers Simply stated. We are asking you to help us see this battlefield to put robots into harm's way first and to optimize our ability to Distribute the American firepower. We call our effort project quarterback All of the vast array of current and future military sensors aviation assets electronic warfare assets cyber assets unmanned aerial systems unmanned ground systems Next generation man vehicles and dismounted soldiers will detect and geolocate an enemy on our battlefield We need an AI system to help identify that threat Aggregate it with other sensors and threat data Distributed across our command and control systems and recommend to commanders at echelon the best firing platform for the best effects Be it an F 35 the Urca or an RCV and that the task goes to the most appropriate We want the system to be able to do this for commanders not in 96 hours But 96 seconds and allow that brigade to fight more efficiently In conjunction with the network APNT and the AI task force NGCV CFT is working on this tactical effort at the brigade and below echelons of command LRPF is working on the division core and theater army echelons of command shaping the effort However for AI to work for tactical and operational commanders. We cannot produce two different systems Leaving the technical details for other conversations. Here are some simple operational AI requirements for the community This is what we need to make AI an enabler and not a distractor for the army first Operationally deployable systems that aggregate data from the vehicle level ultimately to the theater and theater army level We need to start small and work our way up To targets to platforms We need an open architecture that allows integration of multiple platforms and sensors We need a robust and resilient cyber defense that can't be targeted or hacked in a timely manner And it has to be self-healing We need robust and resilient systems soldiers are really good at breaking things and our next fight has no room for contractor support Thank you all for all the work you've done to make autonomy a reality in every American's life It's coming and thank you very much for all your efforts. Thank you