i don't think any member of the police department is moving with the speed force of action needed to get our fighters what they need.

we established that the security is integer on rapidly embracing new technology,. from a platform centric combat to network-centric. there doesn't seem to be any as formidable. this is not only to hear about your plans for the coming fiscal year but gorsuch understand the obstacles and barriers that you have encountered on a day-to-day basis. as i have said before in this room. you can almost copy and paste the speech. we are going to kill the valley of death. the same people have the same conferences and then it does not get any better. why it seems to this committee that the implementation of

policy -- the directives on ai from your bosses -- we deftly made a little progress. the department of defense's achieve information officer. and the director of the defense systems agency and the commanders join forces for the primitive defense network. he is set to depart his position on april 15. before we hear from the witnesses, i go to the ranking member for any comments you may have. >> think you for your bipartisan focus on how we can make the department are innovative, agile and capable of adopting the

latest technology software and working with the -- i hope we can explore not just what the apartment is doing but their perfectibility is needed. the reality is we have a challenge to bring this ecosystem up to the beat of modern information while operating in the area of rate power competition. securing the industrial base. accelerating the rollout of a robust identity credential is fundamental. the subcommittee had a very interesting hearing on software developer and software acquisition. a month ago, the committee held a hearing on exhilarated the base of innovation across the department. as we approach the end of this

term, what would be helpful to me is a few, to her three very concrete recommendations. particularly of what the committee can do or needs to do legislatively to help with making the department more innovative. it is more a matter of implementation. you see iraq for the committee to continue to do more legislatively and if so, what would those recommendations be question mark thank you mr. chairman. good morning chairman gallagher. thank you for the opportunity to testify before you today. i am happy to be here with dr. martel as well as general skinner. lieutenant general skinner and i lead a team to secure and

modernize the department information technology, enhanced were fighting command control, and cultivate a digital workforce. we look forward to sharing the progress on the department's digital transformation efforts. as we have seen in asia and europe, we must keep pace to support the nations work fighters and protect national security capabilities. we are laser focused on zero trust implementation and we expanded eligibility for the defense industrial base armed security program. cloud computing means -- remain central. i testify last or that the -- that that apartment was just influencing the cloud journey.

through the joint -- joint were fighting cloud capability, they have successfully awarded more than 37 task orders over the last year and within 50 more on the pipeline now. we also published clout contracting and reduced contracts brought across the department. in today's environment and as you are focused in on your hearing last wednesday, it is critical more than ever that we provide dod personnel with secure and resilient software when and where they need it. we need to recognize the urgency of this issue and we are working hard. more than 55 software factories are transferring the way that dod develops. this requires changes to or processes, policies, or courses and technology. accelerating the ability to operate and reciprocity are key

to this effort. just as important is the ability to operate in any environment. that is critical to defend the nation. we understand increasing commercial demand. this will explore ways to address increasing federal and commercial demand without compromising national security. 5g is critical to both dod and civilian industry. my office became responsible for this mission. each of these missions would be impossible without the right people. this past year, we laid the foundations to enable dod to grow a dynamic and innovative

workforce to succeed in the 21st century. to ensure that we work with industry and crude to retain the right people with the right skills for the right jobs. as people are our greatest resource, we understand the interaction of the department's ip instructor -- if a structure that directly affects their mission and morale with each work better. civilian and contractor. to enhance user's parents, we have established a customer experience office. i look forward to answering your questions. >> thank you for allowing me to testify here today and to share the work that the cda was doing.

we spent the last year and a virtuous cycle. to answer the direct question at the beginning, we have to continue the cycle and that is what is going to increase the exhilaration. this is accelerating analytics in ai. we create sustainable changes at scale, weeding and overseeing this printable staff assisted function. mostly through policy guidance but also actively delivery capabilities across the full range. this enables change.

and running the business of the dod. think about it this way. if our job is to build all of the air department, that does not scale. we will build monday and tuesday solutions. it is really the tools and policies and best practices. they need that ai. they are the ones that understand the problem. we have been tackling both of

these fronts. the second allows for sustainability. we do this through the eye -- through the hierarchy of needs. you have to get the data right. the foundational air is do we have the right data? it is understandable. can people use that data? think about a separation between data and apps. the next level above that is analytics and metrics. spent a greater our time this year getting people to move from effort based metrics to outcome based metrics. that has allowed for the running of the department. on the top of that is ai.

you measure the past to protect the future. you need the quality data. this hierarchy of needs combined with the agile approach is how we're going to drive this sustainable change to our virtuous cycle. what do i mean by an edge approach? this is what is continuing to drive. we deliver by learning by doing, shipping fast and iterating quickly. this is done on real networks

with real data to learn and deliver fast. we could not have done without working closely. >> think. >> good morning. i am honored to be here today. i am honored to be alongside dr. craig martel. a key ally and partner in the campaign to innovate, modernize, secure and defend the department's network.

we the people's republic of china giving all cyber has to strengthen our digital technologies. through the support of this committee, we look forward to highlighting our future plans. resiliency, a jodie comer survivability and velocity are key tenants in describing our initiative started with the department flagship which was awarded in december of 2022. it provides access to multiple global cloud fabrics. along with the 47 passwords, we

have successfully deployed pay commissions. we also had joint operational edge. it is critical that we ensure that dod personnel have secured software we got this where they're located. with templates, tools and automation in our software platform. equally important is perimeter defenses. we have two keypads ongoing with commercial companies that are increasing our ability to see encounter activity which are increasing each and every day. we are in lockstep with the zero

step. thunder down which is part of our next generation network and data environment while also increasing cybersecurity by knowing who is accessing the network, limiting individuals to only the data they are authorized to access. in haste -- enhanced data analytics like fires and houses that key players in spreading cash from spreading. our work is not done. i believe our new strategic plan outlined the goals. a final area to highlight is our commitment to novella mission such as the national leadership command capability. this will deploy and integrated multiple of the secure voice and

video communications and conferencing capability to the -- to apply direct support to senior leaders including the president, secretary, chairman and nuclear command and control community. nerve these initiatives are possible without our bold and quickly thinking workforce. our workforce when -- workforce 20 35 strategy is to retain and professionally -- professionally develop that the rape and nessel -- the right personnel are in the right positions. strengthen and resilience and work better success relies on the support subcommittee has provided for many years. i'm grateful for your support and the opportunity. >> we were number of the questions. what's the joint fires network in the absence of a solution or operational capability, how will

you contribute to the j fan and by extension command and control. how do you see the joint operating system and cedia efforts fitting into the j fm? >> thank you for the question. we were cannon hand with arnie and with research and extremity in. the underlying data later which draws data to build to make the decisions, to be able to quote change comes from the ceo. we have a team that sits with them and when we need more data, we help deliver that data. they were a key part of our last experimentation.

the success was due to that partnership. we are tightly aligned. questing used elaborate for us on the difference between the joint operation system and the data? >> out there it is part of the data integration layer. think about as two major components. a catalog to help you find where data is and apis that allow the data to be served. that data has to be so the work with cio is really important there. the data is in the red cloud place. the data has an api that is -- that allows an application builder to access that data when needed and it is discoverable. not only discoverable but when it is discovered, it is understood.

just us specific things that we originally contracted them for and they are now part of that data integration there. they have data they are providing. that has an api layer. you want to think about the data integration there as a multivendor heterogeneous entity that is the data that is needed is discoverable and accessible. that is what we are driving four. >> on 25, it says that the joss is moving to production. i like the idea of using all sorts of acquisition authority in commercial technology. could you elaborate more on how you see the johnson production? >> and may have to take that for the record or for the closed session because i don't actually know specifically if we can talk

about what it delivers. they deliver specific information that was necessary for the joint fires network. is there hardware that is distributed across pay, theater that allows that data to flow easily and it is discoverable through our catalog and our apis. i think i answered the question more specifically that did before. >> you are leaving soon. what are the metrics you would your tenure on? >> massive increase in demand for getting it right. we just did a presentation across multiple to parents about the six -- the successes we had. a year ago there were a lot of i don't know what this means. i don't know what you mean by

getting the data right, i don't know what you mean by having the data accessible in the right time and right place. i was very pleasant surprised. there is no longer an ambiguity in our work about what this is. it isn't a system, not even a destination. it is a set of behavioral patterns. this was command and control for that. the demand for that data flow has geometrically increased over the past 1.5 years. if there is a wind that i am strongly willing to claim, it is that. >> how would you respond to the articles about the poor climate service? >> that is great. a great question. really important one. we did a very hard past. we took four organizations, we merged them with very distinct

cultures and we had to break a lot of expectations in doing that. there were lots of people that wanted to do particular things and we set no. we will look for economies of scale and we will do these things instead. that is a monumental task for any merger and acquisition. it's going to upset some people. we worked really hard over the last year to work with some industrial partners on getting our culture right, on making sure our teams are being heard, that communication flows better, we did a bias as fusion new to the expense of giving occasion with our people in the beginning. i am looking for to next month when the new poll service are released. i fill confident they will say good things. >> take you, mr. chair. see were to drop -- you worked at dropbox. can you describe some of the collaboration with silken valley

technology companies you had and how that can improve? >> we could not have done the job without a tight partnership with key companies. palantir is one. data bricks is another. almost everything we do and then all of the cloud providers. who call -- google, and herzog -- google and amazon and microsoft. every silken valley company. we could not have done it without them. we spent a great two of time talking to the leaders of those companies to get buyers for the stores i am telling you now. what i try to do was take best practices from silken valley witches -- that is more important. the data has to proceed the app that is on top of it.

quality data has to proceed ai. i vetted this view with all of my colleagues in silken valley. i had many conversations not just with the breakup is but lots of startups to make sure that they were on board and that they would be willing to engage after we built out this multivendor integration layer that would allow for separation of data for maps. >> are the things we can do to improve collaboration? you ticket is in a pretty good place? >> i think we have -- we still have some work to tackle. doug beck is a really strong higher. he is on board. i've seen an increase in the way we have been interacting with small businesses for example. because of the market was we

built, we are now up to 65 nontraditional complete contrast. we are up 75 small business contracts and many of these are done in 30 to 60 days. that is because of the authorities you have all given us. we are seeing some growth. we are seeing some inertia increase. but we need that to continue. i think the biggest take away -- if i'm candid, the thing i feel best about is that at the end of my tenure, people are demanding more. 100% agree. pastor has to happen. hundred percent agree but i think it is not clear what has to happen faster witches getting the data right, having the data flow correctly, being able to integrate with companies as quick as possible and it shouldn't be something we built. it has to be multivendor and hybrid. >> i see you have a computer science phd degree from the university of pennsylvania. any people rightfully have been

concerned about anti-semitism and there is no place for that on because campuses but some of the rhetoric in this building has gotten pretty excessive. one number of congress that i'm going to start defining these universities. he talked about what it would mean for our national security and ability to have a lead in ai if we just started defunding m.i.t. university of pennsylvania, harvard and many of these universities? >> they can for that landmine question. i think it is a tough call. the environment of a university in order to be effective has to allow for everyone to be able to think freely. but i do agree that if we don't continue to fund stem the way we need to fund stem and continue to fund technology the way we need to then that is going to

put us behind. continuing to fund has to be the case. quotes general skinner, are you a gamer? >> i am not. >> no madden, no ea sports, call of duty? ? son is. i am not. >> i'm not either. but it seems a lot of our servicemembers are. a lot of ussr building these great gaming complexes. i took a great interest in a report that we were feeding some of the starcraft two game models into integration with chatgpt and ai surpasses humans. i guess what they did in this case was they told jet -- told chatgpt to function as an

assistant to a military commander in this engagement so i imagine like in ai to its root, assistant to the regional manager. in changing, dynamics, military assistant turned out to be quick capable. dr. martel, you seem to have some familiarity with this. what can we learn from the integration of giving models and ai in military strategy? >> i am not a gamer but let me talk about how we might effectively use tools. we have been working really hard to figure out where and when generative ai is going to be useful and where it will be dangerous. the danger is it takes a high kind of look to fill it in the upper this model. there is a very large demand signal for ai to replace efforts

and allowed novices to replace experts. where ticket will be most effective is helping experts be better experts. question i don't know. i find a lot of novices showing capability as experts when they are able to access these large leg which models. >> i think the reason is it is it's really to validate output. as long as there is a way -- i am totally on board. hallucination has not gone away yet. there is lots of hope that this will go in. there is some research that says it never will go away. that is a question we really need to pay attention to. most importantly, if it is difficult to validate output, i am very uncomfortable with it. >> i have just used in my modest way the ability to audit the

chatgpt. i think you are right. i think it is interesting -- think about this. you think about that in the air domain. the space domain. i want to get to the inputs as well. that is about the quality of the data. i envision a circumstance in this room, marking up the ndaa. there is a big fight about who owns the data and just as we see in a large language model, will you trained on is our data and our stuff so we have some ownership interest in the work product that comes out. as you depart government service and onto something us, what advice can you give on how to have as much of that data open source and acceptable so we don't have a circumstance where

locket market is saying we have to protect our source code on the f-35? >> the tactic is to separate the solutions witches is did all the way up to the and solution. and then create two separate market places. who's going to open my word doc? that is the app layer. you can also build a marketplace and a data layer. i invested a great deal of ip and worked to build up that data layer. they do we pay them for access to the -- access to that data. selling that data not just to themselves but require that data be accessible. i think that is a really important point.

i hope we're able to get back to it at some point because when what i worry about is creating the cost of that data. so as to vertically integrate all of the features of the contract. i think this is a gouging that we are going to have to confront that. >> i think that's right. quick thank you. the first year and half, calls are up over 50%. texts are up over a thousand percent. saving thousands of lives. what he has heard from ariz about what difference it has made in their lives and their units. i know the leadership department is working on producing our shocking number of suicides. it is a legal requirement that you can piled on 11 but not yet 988. can we do to make sure we get there?

>> thank you for sponsoring that legislation as secretary alston said. mental health is health and this is imperative. so within that department defense, we have validated its use on the pentagon, fort meyer. weaving call from my office to make sure. the i want to be honest about the first few times we had done this. working with general skinner and then also personnel and readiness. making sure the fed guidance is out there. we are going to continue to validate. i want to make sure we are clear on this. this will be an ongoing process for whether it is at camp anderson or edwards air force base. wherever a servicemember or civilian is that needs to get that number. we will continue to validate this. this is all that we haven't in talking about recently. this will be an ongoing process to make sure that 988 is reachable.

calling from the dearborn down at 588. this is something that got her attention. we want to make sure we can validate that. >> a lot of troops know this number by heart. it is going to be able -- is going to be important to people to access this. dr. martel, thank you very much for your time with the department. we all clearly agree that we need to invest in ai capabilities for the future and we also understand that ai is dependent on data. it is kind of insane to me that american taxpayers have spent more money than any defense system in history and yet we downloaded data from many missions and literally throat out. >> i'm getting all the great questions. that is a great question and i

appreciate it. we call that data exhaust. valuable data is an example of it. i don't feel i have the expertise to address that. we can continue to deliver solutions. i believe that virtual cycle will start creating the demand to understand how the data can be most effective. >> if my laptop or my phone told me i was out of space, i could solve the problem in a couple of hours. i don't expect beauty to that fast. the fact that he has taken years and used to can give us an answer is a real problem. that is a question for the record. how are we going to make it safely available?

quickly think are two parts to that. from the available perspective, the technology is there. the right apis that have the right limitations but it is more about contracting to make sure the people who produced -- it is fundamentally the governess data. that is ip you can pay for. we have to make contracts clear about what the payment looks like and how it must maintain accessibility. that is a really big set of work we have to do. on the safely adjustable part, i refer to my colleagues about how you can use zero trust to make sure all of this is the right estimate. quick they have their top ai

companies working on their militarily -- working on their military problems. this is where you have to have the urgency to get this done. how are we going to find someone to a place dr. martel and bring the level of academic expertise and private-sector experience? >> we will certainly miss dr. martel. he has been a huge supply. >> i am glad that you're nodding your head yes. you have your engineer sitting with your operator. i am sure they spent sickness of that is growing.

usually the data integration later -- data integration layer. this is likely the most critical position in that department. moving and exchange data between many services. how do you build the data integration layer question mark >> i can tell you what we have been doing. we have been building out the prototype of what we mean. combatant commands can have a unified picture of what is going on in the world. we are doing that as the key learning exercise. we do these to understand what

combatant commanders would need to see and what all of the components would need to see and exchange and how data would need to flow in order for it to go from powerpoint and email the digital data flows as that information goes across the combatant commands. the next step which we should be doing within the next three to six months is building up a set of requirements so that industry -- we have been doing this with key partners. it needs to be -- the technology we have built, we are leaving behind, it is there. it is available. that is what we call a minimum viable capability. it is viable. we then need to build out the requirements that allow other industrial partners to join in and expand that data integration

layer but also expand those capabilities. quick sells ice on the wing may need to tweak. that is something you need to address to this committee. >> absently. i would prefer to do it because it is a question for the record. it is alien definition of what we need. i could use about 45. could you explain what would be the consequences of affording your position with some sort of directive authority over these services and their development? >> i am not a fan of that. i am not a fan of hard authority there. i may be saying the wrong words because i have not gotten quit up on all the bureaucratic lingo i should. the center should provide

oversight policy and best practices. the edge knows the problems best. solving the problems from the center and imposing upon the edge i think is dangerous. it will create one-size-fits-all solutions. >> you should write that down and put it on a plaque so can hanging on every room in this building. and the services understand what they need better than ost understands what they're going to. they do have authority about the interface. the right kind available have to flow up out of the services.

force the right questions. one day it should flow. we can't answer title x and title 50 combinations. we can start with that. we have to watch the data flow and then there will be an increased demand for dataflow and the right data to flow and we can say that is a policy issue we can tackle or there is a small version of a policy issue or just one piece of data. that is the way the changes going to happen. it is not going to happen by some large due -- philosophical view of how the world is going to be. quick thank you, mr. chair. i find it fascinating. we just had our first ai task force meeting. this is literally right in line

with us. there are a must focus on this technology right now. it is unbelievable from every aspect. whether it be cybersecurity or mechanical applications, laborforce educations, law and will be considered legal and how you verify facts. this is a comprehensive issue that everyone is focused on right now. i want to give more insight into how you are approaching it. dr. martel, much like the data storage capacity, they have increased at an initial rate. what challenges does this rate? we are trying to keep up legislatively. whatever the dod and how it applies mark >> it is

remarkable. i will say this is the most exciting time to be an ai expert. the science has been really remarkable in the last five years but if you go back 15 years it has been amazing as well. i don't know if the science supports the marketing. i think one is to -- skeptical is the wrong word. i am very bullish on this technology. i think it is important to be skeptical about or to verify -- trust but verify. -- verify the claims of the marketing. that is fine. number two, i think it is extremely important to not see this as a monolithic technology which is how it is sold. get this thing. if we have it, we win and if they have it, we lose. i think it is fundamental he flawed.

if we have it, it doesn't mean every problem we want solved will be solved. if they have it, it doesn't mean every fear that we have will come to pass. we always have to think about it on a use case by use case basis as we always have with any technology. the reassurances -- it is just technology, it will work and some use cases. it won't work and other use cases. we have to have quality data and evaluation metrics to be able to say it is working that were not working now. i get is truly frustrated with my colleagues who pitch it as magic. i actually think that is the case often but when you ask about the use cases they have a clear set of small use cases they know how to measure. information retrieval or a number of things which we all feel fairly confident are working well. the sort of pitch that it is going to reason for us, we are

going to be able to do away with brains and use this instead, that is just way overhyped. last year i think that hype was braided. medic we have learned in the last year where that hype is starting to come down. close you mentioned a couple of things. i would spew one thing you said. we talked about us not having it. when they talk about producing 100% of our ai chips and we don't have a backup, that is a denial. >> the underlying hardware of the chips act is the right direction to go because -- that i think is a national security issue. the ability to generate the underlying infrastructure. in the underlying data. i just meant the models themselves. we have a gap between the science and the marketing.

they are trying to rationalize that gap. we are building immaturity model. very similar to the autonomous driving maturity model. that is a really useful model because people are playing level five and abjectly speaking we are really at level three with a couple people doing some level for stuff. >> i am all for kids not driving in cars at figuring out how to do it safely. i am very concerned about the coaching and what he could be to us what we have disruptive chinese technologies that are competing with us at the cybersecurity level. >> if i could respond, i am in a hundred percent agreement. i'm not saying we don't have to worry about these cases. we do have to worry about specific use cases but just shouldn't worry that it is monolithic. we should do what we always do.

evaluate what our adversaries can do. invest in ways we can do it better and invest in ways that we can't contribute to our adversaries not doing so well. my only caution was don't think about it as a scary, monolithic technology, thing about the use cases and do the right valuation of where they are and where we are. quick thank you, mr. chairman. i knew it was going to be a good day when i saw the former air force here. dr. martel, what concerns me about the conflict talking about china and taiwan is the tremendous potential loss of life. the only thing that really bothers me is when you look at wargames when there are some scenarios where we lose in a very short timeframe a fighter aircraft. that is devastating and to avoid that kind of catastrophic loss i

think we need to employ an unmanned aircraft. to do this, they need to invest not only in hardware or software. that exits in the commercial companies like amazon don't have the requirement to operate that way in the military. i want to ask you what you think the significance of autonomous unmanned air platforms will have in helping us do enough to get to where we need to be. >> i completely agree with you. i think you're going to be strongly important. i think we have to continue to drive that forward. one of the things they were doing was supporting the replicator initiative by creating a sandbox workup of these can come using our data to evaluate the success of their

unmanned aircraft software. with respect to the scenarios you think will be effective. we are strongly supporting this. part of this ai scaffolding. this way in which we are trying to help scale the department by giving tools to people closer to the edge. we will provide data to that. we will provide scenarios. we will provide some software for that and then the people who were part of the replicator initiative will come and try out their software, their detection on our data. we think it is really important and we are strong supporting it. >> can you talk about -- you mentioned the ai scaffolding. >> the job is twofold. one is immediate help to the war fighter and building out tools,

processes, best practices that will allow other folks to easily build us a -- filled the solution to the problem. the centralized team cannot build everything. part of our support to be able to scale is what we are calling ai scaffolding. that includes things like the ability to do labeling as a service. knowing what you want to detect means you have to live with that data. that is very difficult. most teams post to the edge won't have to do that so we want to provide contracting and expertise to allow them to do that. data transformation is a source. the data that has to come to your problem will be in a thousand different formats. the amount of work to transform a pdf or word doc into the structured data you need will take up for the people on the edge of their time. the building of the model -- we

strongly believe the government ip is the data. we believe the mother should be contracted by the instream but on the others out of it we will provide those tools, those contacts, those abilities to build initial models. that is part of our ai scaffolding. the final models were probably built by absolute experts in this field. and then on the right side of the model building we speak about modeling. the industry just are tackling this five years ago. and the government is behind it. after we ship a model, does it fit -- does it continue to bring value? we gather data from the past to predict the future. the future changes in a war fighting scenario. it doesn't look like the past. a bottle trained on free war fighting scenario is going to start to degrade as the world changes.

how do we measure that degradation? how do we retrain to get that model back up to where it needs to be. that is the global monitoring piece. that will be my biggest charge for dr. plumb. focusing on how we measure the value of models over time. we don't forget, we just continue to believe. the way we deliver the model should include the ability to monitor it and so we have done a lot of thinking about this. we need to move out faster. westway. we will not go to the classified session which will be upstairs in 2337. we look forward to getting into those. we will see you up there in a few minutes.

with that, the classify portion is clear. -- unclassified portion is clear.

this is about 25 minutes. [applause]