Friday, September 11, 2020

Remarks by General John E. Hyten to the Joint Artificial Intelligence Symposium and Exposition

 Sept. 9, 2020

Gen. John E. Hyten

GENERAL JOHN E. HYTEN:  Greetings.  It’s General John Hyten.  I’m vice chairman of the Joint Chiefs of Staff.  Thank you very much for allowing me to be with you today talking about artificial intelligence.  It is a subject of much interest in the world.  It’s been a subject of my interest for quite a while now, and I’m going to share some thoughts with you today about that subject.

So thanks very much for this conference.  Thanks to the JAIC for holding this conference.  I really hope to get across and visit the JAIC before long, but COVID has gotten in the way of that.  COVID is also the reason why we’re having to do this virtually.  This would be much better as a conversation, because I think this is a subject where we need to learn from each other and delivering a speech, a lecture, whatever you want to call this through a video is not the best way to talk about artificial intelligence.  Nonetheless, I’m going to do my best to deliver some things for you to think about as we go forward.

So the first thing I want to do when I’m talking about artificial intelligence is I want to share with you a quote, and it’s one of my favorite quotes.  It’s from John Kennedy.  Now, the quote is about space, and most people in the audience know that I’m a space officer, have been a space officer for a lot of years.  But what you may not know is that’s not how I started my Air Force career.  I started my Air Force career as a software engineer, and I was a software engineer and a – and a programmer, and that was how I started my Air Force career.  But I moved quickly into space because space has been my passion for a long time.

So I’m going to share a quote with you, a quote that John Kennedy gave in 1962 at Rice University.  It’s not maybe his most well-known quote, but it says a lot and I just want you to think about artificial intelligence in context of this instead of space.  But the quote goes like this:  “We set sail on this new sea because there is new knowledge that must be gained, and new rights to be won, and they must be won and used for the progress of all people.  For space science, like nuclear science and technology, has no conscience of its own.  Whether it will become a force for good or ill depends on man, and only if the United States occupies a position of preeminence can we help decide whether this new ocean will be a sea of peace or a new terrifying theater of war.”

So that was about the space program.  And that speech was part of the speech that actually generated the excitement in this country and said we’ll land a man on the Moon by the end of the decade, which we did.  It started the Apollo Program.  It got my dad involved in the Saturn Program.  It got me excited about space.  It generated a lot of things.  But in it, the president – President Kennedy at the time – talked about whether it will be a sea of peace or a theater or war, whether it will be for good or ill, and it’s dependent on man.

And if you think about artificial intelligence, we’re being asked that question today:  Is it going to be a force of good or a force of ill?  And it’s all up to us.  We get to make that decision because, simply put, it’s just technology, but it’s technology that the world understands is critically important.  And because the world understands, that means our competitors, our adversaries, understand that being able to use and apply artificial intelligence in future military operations is going to be critical to the success of those operations.

So we have to figure out how to lead the world in doing that.  And we have some challenges as we walk forward into that, challenges because much of the work that is being done in artificial intelligence, much of the technology, much of the research and development is not being done by the Department of Defense.  I’ve got to be involved in tremendous research and development over my career.  I’ve seen from the very beginning the development of satellite-based navigation, GPS; satellite-based timing, GPS.  I’ve seen the development of an integrated circuit, an integrated circuit really built by our laboratories, our DOD laboratories.  And we are investing enormous amounts of taxpayers’ dollars in looking at the research that will be required for us to leverage artificial intelligence, and that’s important.

But as much as we’re investing, it’s orders of magnitude larger in the industrial sector.  I was talking to one of our industry partners last week, talking about their research-and-development budget, and their research-and-development budget was measured in the billions.  That’s one company, measured in the billions.  If you look at all of the companies and how much is being invested, our numbers pale in comparison.  The great inventions are going to come out of industry.

I was – I was in Silicon Valley visiting a small artificial intelligence company.  They label themselves an artificial intelligence company.  And I saw artificial intelligence algorithms already running, working through the data challenges that they have, applying this in a real business model and delivering capability in the commercial sector.  It already exists.  We have to be able to take advantage of not only that industry, but the research that’s going on on that side.  We have to be able to partner with them.

And I tell you, we don’t like to do that in the Department of Defense.  We like to think we’re still the big dog in the room, and in this case we are not.  Well, at best we’ll be a minority partner in the research that has to go on, but it doesn’t mean that we can’t deliver very, very critical capabilities.  We’ll focus on applying the research and development of others into the military business, because that’s not what they do; that’s what we do.

But we have to get out of our own way in many cases and apply the research and development of others to our problems, which means we have to partner with them.  Which means there has to be a business case that we – that we partner with them with because other businesses don’t just do things out of the goodness of their heart.  Many are operated by great Americans.  Many love operating in this country, want to make sure this country succeeds.  But they’re a private business.  They have to be able to make money.  So we have to figure out how to partner with them, allow them to make money, and then apply the technologies into our side.

And again, we don’t like to do this.  We like to control things.  We like to have a focus on our research-and-development budget, how those things are going on.  That is – that is not enough to make sure that we can get the job done, so we have to look at research and development a little bit different.

And then we have to look at all elements of artificial intelligence and we have to look at them effectively.  And there’s basically seven building blocks of artificial intelligence.  I’m going to spend a few minutes talking about each one just so you understand what they are.  So the seven are algorithms, data, software, computing, ethics, engagement with industry, and international cooperation.  Those are the seven building blocks that we have to be able to deal with.

And I tell you, when people come into my office – whether it’s government people, industry people, whoever it is – it’s like, all they talk about is algorithms.  And don’t get me wrong, algorithms are critical.  Algorithms are important.  But algorithms, to me, are actually one of the easier problems to solve.  We have to invest in developing the algorithms, but there’s industry partners out there that already have functioning algorithms that can take massive amounts of data, apply the algorithms to the data, and learn from that – allow the machine to learn from that, allow it to grow as we go through that.  We have to figure out how to do that and take advantage of those algorithms, develop our own algorithms.  But algorithms without the other building blocks are useless, especially when it comes to data and computing.

So of the seven building blocks I talked about just a while ago, the two most important are data and computing.

We have to make sure we focus on data.  So I’ve talked a little bit about algorithms.  Again, we need to invest in algorithm development.  We need to understand what those are.  We need to make sure we apply them to the mission space that we have.  But we really need to spend a lot of time focusing on data.  And there’s really two different pieces of the data, and it’s actually related to computing as well.

The first thing we need to think about when it comes to data is gathering all the data that exists so we can take advantage of it.  There is so much data that’s out there, and much of that data that we gather in the Department of Defense we actually throw away.  We just ignore it because it’s irrelevant.  Or the data is in the wrong format, therefore it’s unusable to our structure.  All of those things are untrue.

I watched a company in California take data in a format that they’d never seen before and they did not try to reformat the data.  They just taught the machine how to look at the data, understand the data, put it in their construct so they could apply their algorithms to that data, and with a couple weeks and just a couple hundred thousand dollars they delivered a capability that just – that just blew me away because their focus was on the data that existed.  How do I – how do I learn from that data?  How do I understand that data?  How do I take it?  So the structure and the format, we’ve tried for years to mandate a structure, to mandate a format.  That’s not the critical piece.  We have to understand how it is structured, how it is formatted.  We have to understand and be able to explain to others how to do that.  But machines can learn from that now if we just teach the machine to look for the right kind of data.  So we have to build that existing data architecture out and understand what that is from the existing data that is already out there in enormous amounts.

And then we have to look at data from a new perspective.  Call it data discovery.  And in many cases, the machine that we’re talking about – the machine that will be performing artificial intelligence, the machine that will be learning as we go through this – that machine can drive sensors all by itself.  That machine, that computer, can have sensors built into it, and it can gather that real time, discover new information – discover that information, translate it into data, put that data in the system, and learn from that data.  So when you integrate the sensing capability with the data, now, if you apply computing in the right way, you can accomplish just amazing things.

So you see, computing, algorithms, and data together are all part of the same puzzle.  They can’t be looked at separately.  They have to be looked at together.  And there’s two kinds of data – data that exists and data that can be discovered – and we have to figure out how to work in both of those.  And then the computing has to be powerful enough and fast enough to manipulate the data quickly, to pull the right information so that we can deliver the kind of capabilities that we need as we go forward to the future.  That’s what we’re going to get out of machine learning.  That’s what we’re going to get out of artificial intelligence.

So embedded in all that is the third element, and I kind of skipped over that.  I skipped over it intentionally.  I talked about algorithms.  I talked about data.  I talked about computing.  But the third element is software, because software is actually the magic that is going to make all this happen.  Somebody has to sit down and write the code, figure out how to make all of this work together, how to make sure we can look at the code and pull those things out.  The software developers of the future will be – will be the most important element of this from a skill set because that is the art.  The data is kind of an enabling capability.  The computer is the machine.  The algorithms are the magic.  But the software that puts it all together is going to be the key.

And when you look at software in the Department of Defense, oh my gosh, we develop software in archaic, ridiculous, 20th-century ways that never worked.  And I’ve been through so many different software-development approaches inside the Department of Defense.  I was part of the Ada Initiative, where the Department of Defense thought it would be a brilliant idea to develop our own programming language so that we could lead the world and teach ourselves how to write a common programming language.  Well, guess what?  The Department of Defense does not lead the world in software development.  The rest of the world has gone past us so many ways.  And the developers are completely different.

When we put a software-development team together, what do we do?  We build it like – we build it like a tank.  We build it like a ship.  We build it like an airplane or a satellite.  We put together hundreds of people and we put together these giant engineering teams and try to put them together.  That is not the way software is built in today’s day and age.  I have – I have relatives, I have friends, I have a number of classmates and others that are heavily involved in the software-development business, and they would never have a software-development team with 500 people.  We have software-development teams with 500 people all over the department.  It’s crazy.  You have to be nimble.  You have to be fast.  And you can’t be nimble and fast if you have to integrate the work of 500 different programmers.  It’s literally impossible, and we will chase that to the end of the ages and never be able to get to the answer.  And you’re seeing that happens in weapons system after weapons system, where we wait year after year for the software to deliver.  And we have to figure out how to do that across our enterprise.

But when it comes to AI, we had better figure out how to do it in a modern, quick way.  Now, I use the word “quick” because it’s really about speed.  In order for us to stay abreast of our adversaries and move ahead of our adversaries, we have to be able to move faster than our adversaries.  And right now the United States, actually, as a whole does pretty well.  If you go to Silicon Valley or Seattle or Cambridge, or even here in the Washington, D.C. area, or you go to Austin, or you go to any area where software development is a key piece – and our nation moves unbelievably fast.  If you go in the AI industry, unbelievably fast.  If you go in the Department of Defense, unbelievably slow.  Unbelievably slow.  And for the life of me I cannot figure out how we change that culture, that understanding, to move in a different direction towards the future.

Now, I tell you, I have the term “culture” and I just used it.  And I used it for a reason, because many of you think that’s the Department of Defense culture – that that’s just the way it is, and we have to figure out how to deal with the culture.  But I had a boss of mine once who, when I gave a magnificent speech about the difficulty of culture and changing culture, looked at me – and I was a general officer – and he said:  So, John, what is your rank?  I said, general officer.  And he said, well, what is your duty title.  I said, commander.  He said, well, how come you just don’t decide what it is your want to do, tell your people you want to do it, and you’ll be amazed that people will actually try to do this.

So what I’m asking this community – this community in artificial intelligence, this community in the Department of Defense – is to understand that speed is the critical piece of this entire process.  We have to figure out how to move fast in this area, and putting together teams of hundreds of people writing code all at the same time will never allow us to move fast.  We have to put together small teams, small teams that are integrated at the – at a broad strategic level that we move quickly with.  We develop products.  If they work, we use them quickly.  If they don’t work, we throw them away and we move on.

It’s a completely different way of doing business, and we have to be able to do that because if we can’t right-code that way we’ll never fully embrace what artificial intelligence is doing.  And the key is you have to understand what artificial intelligence is, in China, in Russia.  But China in particular.  They have identified that as a critical issue, along with quantum, along with space and cyber and a number of different areas.  But artificial intelligence is a key piece, and they’re moving incredibly quickly.  They’re educating themselves in China, they’re educating themselves in the United States, and they’re moving unbelievably fast.  And if doesn’t matter how far ahead you are; if somebody’s running faster than you are, they’re eventually going to pass you in the race.  And you have to consider that we are in a competition right now, and the goal of a competition is to win – not to lose, to win.  Which mean we have to move fast and we have to move fast again.  But we are challenged right now in figuring out how to move fast.

So everybody in this crowd should be excited about working in the AI world.  They should be excited about the potential of machine learning.  They should be excited about the ability to integrate computing and data and algorithms and software to develop capabilities that can provide the United States enormous advantages as we move into the future.  But if we continue to do business the way we’ve been doing, we will not be successful.  We have to do business differently.  So I ask for your help in figuring out how to do that.

Then the fifth element.  The fifth element is ethics, because there’s a lot of challenges about what do we do with artificial intelligence in warfare.  And one of the first times I spoke in public about artificial intelligence it was at Halifax, at the defense conference in Nova Scotia, and I was on an evening panel – not really a panel, but just two of us, Eric Schmidt of Alphabet/Google and me, standing up, talking about the ethics of artificial intelligence in warfare.  And if you’re sitting next to or standing next to Eric Schmidt talking about that topic, you have to realize that you are by far not the smartest person in the room; he is.  How, then, do I make the point?

And so I decided to start with the John Kennedy quote I started with earlier because I think that describes exactly what we’re talking about.  And what it describes is it’s choice.  Just like every other technology this country has developed, it all goes back to – every technology that’s ever been applied to warfare, from the rifle to the tank to the airplane to the satellite, whether it’s a force of good or a force of ill depends on man.  It depends on how we use it.  And it will be the same way for artificial intelligence.

So my broad ethical look at artificial intelligence is that artificial intelligence should never be used to start a war, ever.  Those decisions can only be made by humans.  So the men and women that lead our country – the men and women that lead the country, they have to make the decision whether the country will use armed force to achieve our political objectives.  And if we’re attacked, I guarantee you that we will use armed force to stand up for our – for our nation.  We have always done that and we always will.  So the decision to go into a conflict cannot be based on artificial intelligence; it has to be based on human intelligence and the human decision process.

But once we’ve made the decision, then we have to be all in to achieve our objectives, all in to achieve victory in that conflict.  And that comes from, in many ways, applying all technologies to our advantage, which means applying artificial intelligence as well.  And we should not shy away from that.  We should embrace that, understand it.  And if it gives us an advantage we need to figure out how to use it, use it in the right way, make sure we use it within the elements of the law of armed conflict, which should not be difficult to do.  We’ve figured it out with every other technology we’ve ever built, and we’ll figure it out with artificial intelligence as well.  It is just another technology.

And then we have to make sure we engage with industry, and it’s a different industry in many cases.  I talked about that earlier when I talked about the research-and-development challenges, but the industry here is fundamentally different than the industry that we’ve dealt with forever.  It’s not just the Lockheeds, the Northrops, and Boeing.  They’ll be involved in this, don’t get me wrong, but it’s different companies.  And you have to reach out to those companies that actually do this and actually understand it, and you have to partner with them.  You have to figure out how to contract with them, how to build those capabilities, how to bring them in.  And many of them don’t like doing business with the Department of Defense because we are difficult to deal with.  The paperwork that we make a contractor fill out in order to do business with this department is just outrageous in many cases.  But we have to figure out how to do that.

So you have to figure out how to, one, cut through the red tape, but if there is red tape that has to be done figure out how to do it, because we need the full breadth of industry.  Because I remember five years ago people said by 2020 China would be ahead of us in artificial intelligence.  I don’t think they are.  I don’t think they’re ahead of us here in 2020.  But it’s not necessarily because of the work that’s going on in the JAIC.  That’s part of it, but it’s the work that’s going on across this entire country.  All of our industry is involved in that, and that’s what’s keeping our advantage.  So we, the Department of Defense, have to figure out how to leverage all of industry.

And then we have to figure out how to do this with our allies and our partners.  And our allies and partners have to be with us every step of the way because whatever conflict we get into in the future, wherever it happens, whenever it happens, the one – the biggest advantage the United States has is that we go in with friends, we go in with our allies and partners.  And so we can’t close the door right from the beginning to our allies and partners.  We have to bring them in, make sure they understand exactly what they’re doing.  They’re also investing large amounts in artificial intelligence.  We have partnerships with many of them already developed and we’re – and we’re moving ahead with pace, which is good.  But we have to figure out how to take advantage of that friendship, that alliance, those partners, in order to make sure that when we move forward in the future we’re ready to do it together.

Right now we’re about to deliver to the secretary of defense a new Joint Warfighting Concept.  That Joint Warfighting Concept will be delivered by the end of the year.  And so I know that the – many of the folks here in the Pentagon, folks on the Joint Staff get nervous when I say we’re about ready to deliver because it’s not done yet and we’re still going through the experimentation phase.  But if we’re going to deliver in December, which we’ve promised the secretary we’d do, we have to actually deliver a product and we have to make decisions on what that product’s going to be.  We have to experiment, understand, and learn, but we’re going to deliver that.  And December is only about four months away, so we’re about there.  And we’re starting to understand what that really is.

And the amazing thing about that Joint Warfighting Concept is it – the fundamentals of it is it eliminates all the lines on the battlefield that we’ve dealt with our entire lives.  So for almost 40 years in the military, I’ve dealt with fire-support coordination lines and forward edge of the battle area and all of these lines.  I’ve dealt with killboxes.  I’ve dealt with all these structures where we say this is the area that the Navy can operate in, this is the area that the Air Force can operate in, this is the area the Army can operate in.  And those lines are going to go away because we’re going to be able to bring fires from all domains, including space and cyber, kinetic and non-kinetic.  We’ll be able to bring fires from all domains seamlessly.

And the speed which we do that will overwhelm and adversary – overwhelm an adversary and hopefully create the environment where we no longer have to worry about fighting that war because an adversary will look at us and say I never want to enter into war with the United States.  That is deterrence, having a capability that prevents the war from happening.  And goodness knows we never want to have a war with China or a war with Russia or a war with any nuclear-armed adversary, and the only way to avoid that is to have such strength that is demonstrated to our adversaries that they will not challenge us.  We can do that.

And you can see that if we go into a future where there are no lines on the battlefield, and we have ubiquitous all-domain command and control and logistics that go seamlessly from place to place, from service to service, and it all happens in enormous speed, holy cow, that is the world where an adversary will not challenge us.  Therefore, we have to figure out how are we going to do that.  And I’d tell you I can’t figure out in my mind how to do that without artificial intelligence.  And when I talk about that, the Joint Warfighting Concept, I realize we left off a critical element because it actually should be a Combined Joint Warfighting Concept, because we have to bring on our allies and partners.

So it has to be enabled by artificial intelligence.  We have to be able to use machine learning to create that environment.  An all-domain command-and-control concept has to have all those pieces together, and we have to have our allies and partners with us.  That’s the key to moving forward.

And how many times in this remark – and you can go back and look at the tape if you want, since it’s taped – how many times have I said, that’s the key?  At least five times I’ve said that’s the key because all of these things have to be linked together.  All of this has to come together.  But when it does, that’s when we have a concept that will deter our adversaries and prevent major-power conflict.

So I’m going to read that quote to you one more time but I’m going to change one word.  So:  “We set sail on this new sea because there is new knowledge to be gained, and new rights to be won, and they must be won and used for the progress of all people.  For artificial intelligence, like nuclear science and technology, has no conscience of its own.  Whether it will become a force for good or ill depends on man, and only if the United States occupies a position of preeminence can we help decide whether this new ocean will be a sea of peace or a new terrifying theater of war.”

We want peace on this planet.  We want peace to be the world that our children live in.  That’s the world we want.  That’s what we need to use all of our capabilities for, and artificial intelligence is one of those.

So I look forward to working together with you to achieve this kind of vision.  This is a complicated process.  Hopefully, today I’ve given you a few things to think about as you begin this conference or as you continue this conference.  And again, as you see me in the hallways, or feel free to send me an email, because this shouldn’t be just a speech; this should be a discussion.  And I look forward to that discussion continuing.  Thank you very much.

No comments:

Post a Comment