How the Pentagon Went West

Earlier this year, 3,000 Google employees urged CEO Sundar Pichai to cease the visitor'southward involvement in Projection Maven, a Pentagon project that used AI to analyze drone footage. "Google should not be in the business of war," they wrote.

Google complied and vowed to never develop AI technologies that can be used for state of war or surveillance, though Pichai said the company volition "continue our work with governments and the armed services."

To those who work for the authorities or armed forces, the uproar might be perplexing given that much of Silicon Valley'south most popular innovations are built on engineering that originated at defence force-funded academic institutions and were developed past military-affiliated contractors.

As Maynard Ansley Holliday, a senior engineer at the RAND Corporation, tweeted, "The unproblematic fact that Google engineers can have this contend in an open up and costless society, where their Chinese, Russian, and Northward Korean counterparts cannot [proves] freedom is not free."

Holliday, a former senior technical counselor to the Office of the Undersecretary of Defence force for Acquisition Applied science and Logistics, was instrumental in establishing what was then the Defense Innovation Unit of measurement Experimental (DIUx), which sought to give the Pentagon a foothold in Silicon Valley. Earlier this month, it dropped the "experimental" from its moniker to become the Defense Innovation Unit (DIU).

We spoke on the telephone to Holliday recently to hear more about his former life at the Pentagon, the early days of the DIU, and what he believes is next for defense tech in his office at RAND. Here are edited and condensed excerpts from our conversation.

With a upkeep of $450 billion, the Defense Department was a big sandbox to play in.
They had all these engineering science and inquiry programs to support, through diverse channels. What virtually people [in the tech manufacture] don't realize is what the government does is purchase downwardly the technical risk. Traditionally, the DOD was always the primeval investor in all emerging engineering science: the net, AI, GPS, autonomous vehicles, and they let information technology all play out. Much of the research fails. Then Facebook and Google and others get to come along [years later] and scarlet pick what's survived and works.

You didn't have a direct path from schoolhouse to the Pentagon though, right?
No, and it was while I was at the National Labs working in the Nuclear Test Engineering Division that I realized all the interesting avant-garde assignments went to those with advanced degrees so I took a career break, won a scholarship to Stanford and took every robotics grade they had. And so, when I returned to work at the National Labs, I was designing robotics systems for hazardous environments, working on the Lawrence Livermore laser plan. Until recently they had the globe'due south near powerful laser, the National Ignition Facility, and I designed its predecessor'southward target positioning systems.

National Ignition Facility at Lawrence Livermore (Photo by Jason Laurea/LLNL)

Not to date y'all, merely it must have been an interesting time, as that was during the autumn of the Soviet Matrimony.
[Laughs] It does date me, but yeah, it was. In fact I got to go to the Ukraine [in 94] after the establishment of the Nunn Lugar Cooperative Threat Reduction Program [which secured and dismantled weapons of mass destruction and their associated infrastructure in the one-time states of the Soviet Matrimony]. I was working there with the Plant of Special Mechanical Programs, [which formerly] developed guidance and control systems for nuclear missiles, including gyroscopes, and circuitous mechanical systems. During that fourth dimension it was widely reported that a lot of weapons-class textile and know-how was being sold on the black market [later the Soviet Matrimony collapsed].

Was it your job to get the gear dorsum?
No [Laughs]. Nosotros went to mitigate the threat and stalk the flow of technology transfer through cooperation with our counterparts in the former Soviet Marriage, working on projects of common interest. For case, our opposite numbers in the Ukraine said: "Nosotros need help with the clean-upward at Chernobyl. We demand a radiation hardened vehicle to categorize and record what transpired at unit four." So our articulation US/Ukraine team designed a robot chosen Pioneer nether my leadership. We adult a mission plan, configured information technology, and utilized NASA JPL'south Vision System, which enabled the robot to record and infer material properties inside Chernobyl.

Early prototyping for what became adjacent-generation Mars Pathfinder tech?
In a way, yes. NASA JPL wanted a terrestrial analog for interplanetary robot exploration for robot systems, which they were already working on.

In the spirit of cooperation, did you lot become to utilize the data from inside Chernobyl?
Pioneer robot at Chernobyl We delivered the robot in 1999. It worked. But when we asked if they could share the information they said "No, that's a country hugger-mugger."

That is tricky.
[Laughs] No annotate.

So what happened adjacent? Can yous say?
My lips are sealed.

Understood. Did you return to the U.s.?
Yes, I returned to the National Labs and worked on national security systems studies for DOD and DOE. Simply I wasn't getting to do robotics, and I really wanted to.

And so you skipped town and headed to Silicon Valley to build robots.
I spent over a decade there working at semiconductor companies, on micromanipulators that repaired integrated circuits. Then on an early on robotic vision systems that used novel algorithms to recognize patterns and place objects. Those were afterwards used in Sony's Aibo. I as well worked on high-definition cameras for surgical robots and on health tech that later evolved into what's inside the Apple Lookout today.

When did yous get the "tap on the shoulder" from the Pentagon?
It was after I had returned to another nuclear weapon lab, Sandia National Laboratory, that I heard from the White Firm. They'd started looking for people with hybrid backgrounds, like me. Funny enough, it was during the period my daughter was doing college tours and then I tagged along.

While your daughter was checking out Georgetown, you dropped into the Pentagon for your interview?
Something like that, yeah. [Laughs].

Permit'south cutting to your get-go day at the Pentagon [in September 2022]. What was that similar?
An experience of animate really rarified air. I had all these pinch me moments. I met all these luminaries in the field.

Fast forward to Apr 2022 and [so-Defence force Secretary] Ash Carter's Drell Lecture at Stanford [on charting a new path on innovation and cybersecurity]. That's when we all knew the script has been flipped. The Defense Department was no longer the No. ane client due to globalization. The commercial sector was moving much faster in areas like, robotics, autonomy, micro satellites, cyber security, synthetic biological science, and we needed to switch it up. Nosotros knew nosotros needed a foothold in Silicon Valley. Simply we didn't want to build "Pentagon West" for what became DIUx.

DIUx needed to take the ethos of both authorities and geek? But those two worlds don't share a language or much else.
That'south why, when we came to staff information technology, we were really careful to pick people with hybrid backgrounds, possessing the ability to role in both realms.

People in compatible who could talk to people who don't own a tie, and vice versa?
Right.

How did you lot start working out what Silicon Valley needed, or would have, from defence funders?
I started talking to all my quondam colleagues and contacts: "What would bring you in to work with us?" I asked. "Non-dilutive capital investment," they replied.

Where they don't lose equity and you don't spend every weekend reading board memos?
Yes. Which is very dissimilar to the traditional VC model, even those used elsewhere in the services, including IQT, the CIA'southward venture arm.

And then how did it piece of work?
DIUx was prepare to provide non-dilutive capital in the grade of pilot contracts for commercial innovation that solves DOD problems rapidly, with contracts being signed ordinarily in under 90 days. DIUx could as well, with understanding from the Under Secretary's office or above, fast track "patent awarding approving" to jump other companies [in the line] due to security need and move apace downwardly the development process. We looked at diverse and flexible exit strategies—monetizing by licensing the tech to authorities, or selling to a Tier-1 defence contractor. Nosotros also gave access to incredible test environments, of course.

Such as on military bases?
Every bit well equally in cyberspace in the Armed Forces cyber exam ranges.

You were able to institute a rapid contracts-to-prototyping process.
Through its Commercial Solutions Opening (CSO), DIUx can execute awards in as little as 31 days, assuasive Silicon Valley to tap into a $100 billion market swiftly.

Allow'south talk location.
Initially, we looked at existent estate in Silicon Valley but said: "This is going to bust budgets." I took the team to Google, Facebook and said "Await, yous can't build a Dilbert cube farm and have any street cred upwards hither." People on the DOD side needed to understand DIUx was a different organization. Gensler designed Facebook's offices and they were already under contract with us, and then nosotros did a crash program with them when we constitute a California National Army Reserve Center that was just five years old at Moffett Field in Mount View. DoD already endemic that space, of form, being Army Reserve. We made a program, using the best of what nosotros'd seen elsewhere in Silicon Valley.

Open plan spaces with lots of snack-and-chat areas?
Exactly. We learned from Google near the perfect ratio of standard micro kitchens per employee, per foursquare foot to maximize colleagues "crashing" into each other, having conversations to drive innovation. The Army Reserve Heart as well has a tremendous auditorium—all wood paneling—and a huge assembly hall, which is a great space to host meetings and nowadays pitches.

It all went live, officially, in September 2022, but yous had to get out less than a year after due to the the election?
Sadly, yes. Equally a presidential appointee, I had to resign as the new assistants came in. I left the Pentagon on Jan nineteen, 2022. DIUx went through a reboot after that and I was offered a job at RAND a calendar month afterwards. Initially I said: "I tin't have another remote task." And they said: "You're in luck, we want to open up an office in San Francisco." I thought: "What'southward not to like?"

What are you doing at present at RAND?
I go to play in the autonomous vehicle space. I'm advising the regime now on beingness a fast follower to Silicon Valley. And so much of autonomy is useful within defense. You can take soldiers out of damage's way by getting them out of convoys swiftly, for case. Government entities are reaching out to understand how these vehicles are testing, and how to build confidence in these systems. It's a great function [for me].

Terminal question, I have to ask: Did yous get to keep your top security clearance?
Oh, yes [Laughs]. I retained my Active TS/SCI Clearance with Total Scope Polygraph. Because once you become that your aperture is open and you lot get to run across a lot of absurd stuff.