新泽西的 AI 登月计划：州长 Phil Murphy 谈合作、进展与准备

Today my guest is Phil Murphy, governor of New Jersey, and, when it comes to AI, one of the most forward-thinking elected officials in America today. Spending as much time as I do in the AI bubble, I often find myself confused and, at times, frustrated by how few leaders in business and government seem willing to truly grapple with the magnitude of change that AI technology could bring in just the next few years.

So I was very pleasantly surprised when Governor Murphy’s team reached out to us to suggest this episode. As I dug in, including by using OpenAI’s Deep Research to explore his record, I came to really appreciate how much time, energy, and desire to learn he’s bringing to AI issues. Especially considering that he’s in the last year of his term-limited tenure as governor, he could easily have coasted to the finish, but in fact he’s done just the opposite.

Concretely, under his leadership, New Jersey has trained over 13,000 state employees on AI tools, begun to use AI systems to help reduce wait times for government services—in one case, from an average of more than 40 minutes to just 90 seconds—and established an AI hub in partnership with Princeton University, with Microsoft and CoreWeave as corporate partners. The hub is meant to promote AI R&D in the state, train workers on new AI tools, and ensure that AI is deployed in ways that benefit all New Jersey residents.

In addition to reviewing his accomplishments to date, in this conversation we discuss how he came to recognize AI’s transformative potential; his approach to balancing efficiency with equity as AI capabilities grow and begin to affect jobs; his strategy for meeting AI’s energy requirements while still achieving clean-energy goals; his thoughts on whether states should regulate AI in the absence of federal action; how to maintain American values amid intensifying international competition; and whether there’s any hope for U.S.-China cooperation despite the deep differences in approaches to privacy, intellectual property, and individual rights across the 2 countries.

Overall, while Governor Murphy maintains some perhaps appropriate skepticism around the most extreme AI possibilities, I was really impressed by his humility and open-mindedness. At one point, you’ll hear him acknowledge that the pace of progress has repeatedly surprised him, and toward the end he commented that our conversation will likely feel dated in just a few months—a truly admirable level of situational awareness from someone with so many other responsibilities on his plate.

My hope is that you enjoy my conversation about New Jersey’s AI moonshot and much more with New Jersey Governor Phil Murphy.

Governor of New Jersey Phil Murphy, welcome to The Cognitive Revolution.

Thanks for having me, Nathan. It’s Jersey to Detroit, Michigan, today. I love it.

In preparation for this conversation, one of the things that jumped out at me most is that you seem to be among a pretty small percentage of our elected officials who are really grappling with just how big of a change AI might be about to usher into all aspects of life and society. I saw in one interview you said, quote, “Everything is on the table.” I’d love to hear how you came to that realization, how wide-open your view is of what the future might look like, and how many of our other elected officials you think are there with you in understanding that.

It’s hard for me to answer that last one. I know we’re all in, and I’ll give you a little bit more color on that. It doesn’t come up a lot in discourse with other electeds. I mostly traffic in New Jersey, obviously, with electeds of all shapes and sizes, from our congressional delegation to county and local officials and our state legislature. Outside of Jersey, I mostly traffic with other governors, and it’s not a hot topic, I’ll tell you that.

By the way, it’s good to be on with you. We’re fans of yours from afar, and thank you for having me. What really clicked with us about 3 years ago was a seminal dinner. My wife, who’s a full-time volunteer for the state, several of our colleagues, and I had a dinner in the Bay Area with a bunch of very significant AI thought leaders, most of them running businesses. Mustafa Suleyman was one of them, and I’ve kept in touch with some folks who are in the thick of things.

They were commiserating about the business environment that they had to deal with—the cost of living, regulations, et cetera. I said, “Let me ask you a question. After hearing that litany, why are you all here?” They said, “We’re here because of talent.” We then had a long talent discussion, and we walked away from that saying, “You know what? We’re a high-talent state. We have the number 1 public education system in America. We have Princeton University and Rutgers University, and the highest concentration of scientists and research professionals per square mile in the world. We have a huge innovation legacy: Einstein, Edison, Bell Labs.”

There’s no reason why New Jersey can’t stake a claim in this emerging realm. I call it a realm, not a technology, because, as you rightfully point out—and let me stipulate, you’ve forgotten more about all this than I know, so I can’t get into the deep end of the pool with you—this is one of those things where you know it when you see it. It just feels like a natural for New Jersey.

I’ll stop there for a second. I won’t get into all the things we’re doing to underpin that, but even if that had not happened, it’s quite clear this is going to touch all walks of life. We’re a big pharmaceutical and life sciences state. The whole notion of drug discovery is going to be transformed. This is not a narrow technology lane. The mistake I think a lot of people make is thinking it’s in a narrow technology lane. This is going to change all of our lives.

Even if I hadn’t had that “Wait a minute, why can’t we be a player in this industry?” moment, I think we’d be aggressively pursuing all of it—delivering state services more efficiently and worrying about ethics, equity, and access. We’re now becoming a major hub within the generative AI space.

You’ve teed up about 5 different questions that I want to ask, but maybe just 1 follow-up on this dinner and what made it click for you. I think a lot of people who are in the AI space are bewildered and frustrated at times by the fact that they’re trying to get people in positions of power and influence to realize that this is going to be a bigger deal than previous technology waves we’ve lived through, and it’s often a struggle.

Was there a demo or some moment that made it click for you? Would you give any advice to people trying to communicate more effectively to leadership just how big of a deal this could be?

No demo then, but we’ve had demos that have blown us away since then—just staggering. You’re right about the frustration, and I’ll tell you what poisoned the well. I’m not a “hell-no crypto” guy, but I’m not a crypto guy. I just don’t—the so-called use case does not jump out at me.

That’s a completely different realm on 1 level, but a lot of frustration—I sat down with some early-stage companies a couple years ago. A buddy of mine in New York gathered them around me. There were enormous levels of frustration. That’s 1 example of—I think you could probably pick several—where you now get to generative AI, which is the real deal. The use cases are overwhelming, and the impact it’s going to have is overwhelmingly obvious.

You say, “Wait a minute, how come we’re not getting the traction that we believe this deserves?” I think it’s partly due to history, as I say. I didn’t get any demonstrations then, but we’ve seen a lot of crazy stuff since then. We’re using it in state government. We’ve trained, I think, now up to 13,000 of our employees who are using our version of Microsoft’s Copilot, and it’s making a huge difference.

We’ve established what I think will be a game-changing venture with Princeton University, the NJ AI Hub. We announced that just over a year ago, and only a week or 2 ago we announced that Microsoft and CoreWeave—which is a hyperscale player, New Jersey-born and bred—are coming in as 2 key corporate, not the only 2 but 2 key founding corporate pillars.

One last thought, Nathan: I think it’s a lead-lag. If you take a snapshot of the emerging curve of artificial intelligence, you might be disappointed with where the discourse is at that moment relative to where it should be, or the potential of where it should be. But if you instead film a video, I think you’re going to see the curves start to converge as more and more people have personal experience and see it with their own eyes.

That seems true, although it may not be a great answer if we take seriously the possibility we’ve heard recently from folks like Sam Altman and Dario Amodei from Anthropic, who are saying pretty clearly and directly at this point that we are likely to get superhuman AI that can do literally—Dario is saying—everything better than humans can do in the 2027 timeframe.

You make a fair point. I may be right that the lines are converging, but maybe not at the pace they should be converging.

So, do you take that—I mean, that’s a radical possibility, right? It’s the sort of thing where I think the crypto callback is informative, because leadership has been sold that this is the next big thing time and time again, and there’s fatigue about it. At the same time, as you said, this seems like it really could be it. How much does the possibility of genuinely superhuman AI weigh on your mind as you go about your business right now, knowing that the people building it are saying it’s only maybe 2 years out?

It’s certainly there, and I’m not going to argue with the geniuses who are at the forefront here. I spend more of my time on the pragmatic implications. I mentioned training our state’s workforce, which we offer for free by the way. We offer our AI assistant package to other governments, and other states and cities use it. We’re all in this together in that respect.

Let me hit this now, if I could. We’ve established what I think will be a game-changing venture with Princeton University, the NJ AI Hub. We announced that just over a year ago, and only a week or 2 ago we announced that Microsoft and CoreWeave, which is a hyperscale player in New Jersey, born and bred, are coming in as 2 key corporate—not the only 2, but 2 key founding corporate pillars.

Why do I say all that? I’m not going to be the guy who’s going to take us to superhuman intelligence, but I’m assuming it’s going to happen. I want to make sure that New Jersey has the very best catcher’s mitt out there when all of that comes our way. I want to make sure that we put things in place that allow us to gain an advantage.

If you’re a resident who needs a government service, if you’re a startup and you want to be part of that Princeton venture, or if you’re a thought leader and you’re trying to think through the ethical or other regulatory implications, I want folks on the margin to lean toward, “You know what? I can find a lot of that inside the 4 walls of New Jersey.”

So tell us about this New Jersey AI assistant. I thought that was really quite interesting and had heard of it, actually, before going deep on your recent record in AI. Are you using it? Who's using it? What are they using it for? Who built it? I think you said Microsoft was the provider—the Microsoft Copilot?

I'm the least qualified guy to be using it, but we've got about 13,000 of our approximately 60,000 state employees using it regularly. We train everybody for free. The people who deserve the most credit here are our AI chief strategist, Beth Noveck, who's been with me since day 1. She had been the head of our Innovation Office, and she deserves a gold star. We've got AI guy Dave Cole, who succeeded her in the innovation space, Chris Ryan, who's our chief technology officer, and then all the various departments, because these folks are spread across government.

There's a natural question that I think needs to be asked. It's one that we ask, for sure, and that is, "What's this going to do to my job? What's this going to do to my lifeblood?" We've seen that in surveys that we've done, both in the public and private sector, I might add. There's a good amount of enthusiasm. Folks want to learn more; they want to get what we're now calling the AI assistant knowledge and understanding. But there's also rightful concern about, "Am I going to still be in this job X years from now?"

I see this every day. I'm now in year 8 as governor, and I've seen it throughout the time: all of us, yours truly included, are stuck too often in mundane weeds that prevent us from being able to soar and get to parts of our job. Either we don't get to it at all, or we get to it for 20% of our life, as opposed to what should be 80% of our life.

That is not just our hope; that's what we're seeing actually becoming—again, early stage, I'm not going to overstate it—but early-stage becoming a reality. This whole notion of upskilling: is it just a nice fancy word from a consultant, or something abstract? It's actually happening.

At the same time, residents are happier because a call-center setup, for instance—which is an obvious one—they're getting their answers 50% faster than they were a year ago. The phones are getting answered, and we have a lot of retail parts of state government. Think about unemployment insurance, which during the pandemic was insane, or Motor Vehicles, which is a hugely retail operation. This stuff is real.

So far, at least, it's in the win-win category: the resident is happier because service is more efficient; the employee is happier because they're able to focus on stuff that they've been, frankly, prevented from being able to focus on because of the minutiae. I hope that trend—not just in New Jersey state government, but more broadly—I hope that continues to evolve.

Yeah, that's cool. Are there any—I mean, you mentioned a couple, and I don't know if those were literal metrics in terms of the phones being answered 50% faster—but how much have you been able to quantify that? Is there any more you can share about the before and after, and a quantitative understanding of how this is making residents' lives better?

I took some notes. One of my colleagues said we've seen a 50% boost in call-resolution rates through things like self-service options. For one of our programs, we reduced phone wait times from 40-plus minutes to an average of 90 seconds.

I do a program once a month on television and once a month on radio called Ask the Governor. In fact, Natalie and I are doing the radio version of it tonight, and it's a great reminder that government service is incredibly central to the 9.5 million people who live in this state—and I'm sure it's true of states around America. They want, "You know what? I'm paying my taxes. I rightfully am able to get X or Y service."

You'd think people would call up and say, "What do you think of immigration in America?" or something. You get some of that, but you mostly get, "I've been on the phone for X hours trying to get my unemployment benefit resolved," or, "I've got a property-tax question and no one's answering my phone."

It's a great reminder for me, at least, that it's a version of Tip O'Neill's "all politics is local": all government is basic—one foot in front of the other, make it work. This is helping us—early days, but helping us a ton.

I love that. You're ahead of many companies in that respect right now, which is a pretty impressive thing to have achieved at the state level. I think there are a lot of companies that are thinking about that, and not that many yet that have actually shifted wait times, for example, in a meaningful way. So that is definitely worthy.

As I said, I hope that continues in that direction, and I hope the upskilling and focusing on big-picture stuff that we haven't been able to get to—I hope that continues as well.

Do you have other ideas for AI extensions to, let's say, what citizens can rightfully expect that you're maybe not yet implementing, but kind of thinking about? Things that come to mind for me, for example—and these are just prompts to get you riffing—are, if you're entitled to a lawyer in the criminal justice system, maybe you should also be entitled to access to AI to help think through your case and advocate for yourself.

Or maybe we should have more self-driving cars on the road, and maybe we should be thinking, "How do we actually prepare our local jurisdictions to lead in that," as opposed to just wait for it to happen to us? What's on your mind in that respect?

I was asking myself a similar question: where else are we touching and maybe transforming a certain skill or a certain service? I wrote down a couple here. Our Office of the Public Defender is creating a brief bank, which is a pretty cool thing—an AI-powered tool designed to help the defenders more efficiently draft legal briefs. Remember, public defenders are not doing it for the money, so anything that can make that process more efficient is useful.

Other areas include business certifications, permitting, and notary public services. You asked a good question about criminal justice and access. You're entitled to a lawyer through a public defender appointment; should you also be entitled to some AI assistant or some AI tool? It's a good, fair question. I hadn't thought about it.

I think self-driving is here to stay. Like a lot of things, you want to do it in a way that is equitable and fair to the folks who make their life and their living from driving. How that shakes out is going to be important: we've got one of the largest ports in the world between New York and New Jersey, and probably 90 or 91% of the port traffic is on the Jersey side—enormous trucking and other ancillary-related livelihoods. How that looks, and how you could do that fairly, is obviously going to be a big issue.

We're the most densely populated state in America, so getting self-driving right seems to me—if you're going to test self-driving, they ought to be testing it in Jersey, given the density. I know that every time I go to a state like Arizona, I see Waymo everywhere, and I've always thought—and I've said this to the Waymo guys—"You've got to, if you really want to make sure this is right, test this in Jersey City or Hoboken or someplace like that, where you've got real traffic."

Again, I refer to this as the AI realm, Nathan, and I think it will touch literally everything. I would bet the biggest impacts in Jersey, probably around the world, will be in the private sector, just because of the scale. I'm not sure what our GDP today is—probably toggles between $700 billion and $800 billion. Our state budget, which I'll unveil week after next, is high $50 billion, so you've got a huge amount at stake.

I don't have to tell you this because you live in the space in the private sector, but I think this has endless implications, including some lousy ones that we've got to be very mindful of. That big Paris gathering was enlightening, but on some level concerning, over the past few days. How this plays out at the national, global, geopolitical level is going to be incredibly—not just important—but, yeah, it probably dwarfs everything else we're talking about.

Yeah, it's fraught at every level. Let's take them one at a time. You mentioned fairness. This strikes me as an unbelievably big challenge that we are not ready for, and so I don't expect you to have all the answers because I don't think anybody has great answers right now.

But I am really curious as to how you think about what the role of government should be or could be when it comes to helping smooth these transitions. Maybe even just in state government, if somebody came to you and said—and you've been an admirable early adopter of this technology so far—but it seems like we could imagine, not too far into the future, it might get to the point where it's like, "Actually, AI can answer all these calls now, and we could provide better service than we used to at some significantly reduced headcount."

Obviously, businesses are going to be facing this as well. How do you think the government should approach those sorts of trade-offs? Is the primary commitment to deliver efficiency for citizens, or do you try to retain those jobs even if you don't really have to? I don't really know where government leaders' heads are at on that today, and it seems like it's coming at us potentially pretty quickly.

I know there's a very legitimate debate on regulation overall, and I think there's a sense that Europe may have gone too far too early without having all the answers—and getting that right. In the United States, should that be at a state level, federal level, et cetera? I think it should be at the federal level, but I thought that a month ago. I'm not sure what that looks like now, so we'll leave that for another conversation.

As it relates to fairness and equity, states have to play a big role. There's just no question about it. Is this going to be a tool that is valuable for very few people, or is it going to be something that's valuable more broadly? That's certainly one of the big areas of thought that will go on in this NJ AI Hub with Princeton University and their great minds. Certainly, states have to play that role.

I don't think we're anywhere close to the trade-off you described, and that's not to say someday we won't get there. But I don't think we're anywhere close to that because I see how much headroom there is in that category I mentioned earlier of stuff we're not getting to, stuff that we haven't been able yet to figure out and/or spend time on.

For the moment, at least, I believe that is beyond the powers of generative AI. That may not last if you believe some of these thought leaders you quoted; that may be sooner than we all think. But for the time being, at least, there is a gap between more efficiency here, more time to think, and ultimately doing more over there. So I don't see that trade-off needing to be made.

It's possible, if not likely, that this is gradual—maybe it's via attrition over a period of time that you're able to perform the overall governmental tasks with fewer bodies. You don't have to make a decision one day where you say, "We don't need these X number of people," but that over time there is a reality.

You could ask the follow-on question, Nathan, which is, "Okay, even in that more benign scenario, where are they going to turn to—the folks who you didn't hire?" Again, this is not getting rid of people, but not hiring people, and what do their options look like? That, to me, will be very tricky.

We're not there yet, but that will be very, very tricky. We now in New Jersey are employing more people than ever before; more businesses are operational than ever before. Our unemployment rate is in a decent—not yet where I want it to be—decent place, so we're not at that moment today.

But there will be some of those inflection points over time. Given the unmet needs in government right now, that's not tomorrow or even 2027 or 2028. As I sit here at this moment, I think that can be done gradually, largely through attrition: "I'm not going to hire that person. I just don't need the next body to fill that space." We shall see.

One of the big drivers of jobs that might make a little dent in the unemployment rate is the massive build-out that seems to be just ramping up in terms of data centers, and then energy, of course, to power those.

How are you thinking about what New Jersey should do, or how it should position itself to capture a share of that? I know there's probably some tension in this because, of course, there are environmental concerns. There's local NIMBYism, which might be a little bit too harsh of a term, but nevertheless, people don't always want these things where they live.

You've also got a 2035 100% clean-energy target, which is potentially in some tension with this. What's your outlook on how New Jersey should capture its share?

Great question. This is an artificial break between types of jobs. There's the sort of CoreWeave AI hyperscaler—big compute power, big energy needs—thank God headquartered in the great state of New Jersey. But if you look at those data centers, that's not a big employment spinner, and my guess is you wouldn't disagree with that. So that's largely an energy-space question. I'll come back to that in a second.

But then there's the—you know, New Jersey sells very well internationally, so we do economic missions around the world and quite successfully. We're an incredibly international state. We treat California as a nation-state; it's the one state that we treat as a nation-state.

We go in the South to L.A., largely for film and television, where we're a big player, and then we go up North for technology, mostly these days for AI. I mentioned that seminal dinner, so we also visit not only CoreWeave in Roseland, New Jersey, but we'll visit Anthropic, or you name it.

There's a different type of employment buzz going on there. It reminds me of the bullpen days back at Goldman Sachs: everybody's in an open architecture, everyone's all in there together. In this case, they're all certified geniuses, and that's very much talent-driven, very much cutting-edge, at the leading edge of the envelope, et cetera.

I view this—I may be simplifying too much here—but there's kind of a barbell. You're in that latter category, Nathan, so you're a huge player in the industry, but you're not sitting inside of a big data center. I think we've got the ability to attract both. That group I sat around that seminal table with was all in that category, and I think personally that Princeton will be a huge magnet for that type of person, at least on the East Coast, if not more broadly.

On the energy side, you raise a very fair point, and it is something that we're working our tails off on. I regret, at the moment as we sit here, that we've not been able to find a compromise with the Trump team on offshore wind, which we are perfectly positioned for. We had the largest program in America, and we've had to sort of hit the pause button on that to hopefully try to find common ground.

We have significant nuclear power still that's been licensed through, in some cases, to the middle of the century. We have, I think, one of the largest campuses left in America in South Jersey, and I view that as carbon-neutral. I know that not the entirety of the environmental community would agree with me on that, but a lot of them do.

We're looking at whether or not that's an angle for us on some of the untapped power we have. We probably have the strongest environmental track record over the past 7-plus years that I've been governor, with the most densely populated state, as I mentioned when I was talking about Waymo. I don't want to blow that in the last year. So doing this and doing it right is incredibly important.

CoreWeave, which is building out around the world, is building out in New Jersey, and so far we've been able to get that balance right. We've not tripped a wire yet in terms of dirty energy, et cetera. You rightfully point out that we've got a 2035 objective of 100% clean energy. Again, wind was going to be—and I hope will still be—a big piece of that.

That's sort of how we see the landscape. So far, we've been able to have our cake and eat it, too. I hope it continues that way. And by the way, CoreWeave is not the only one looking at data centers and compute power in New Jersey. They are our local champion and growing like a weed, but there are others who are kicking the tires as well.

Aside from hopefully patching things up with the Trump administration to figure out how to get the windmill project going again, are there other to-dos that you feel are going to be required to make this happen? Permitting streamlining, or opening up land that maybe isn't currently designated in the right way to be used? Or do you think it's kind of okay, and it's just a matter of following the process?

We're not a large state, so in terms of our landmass—we're the ninth- or tenth- or eleventh-largest population-wise. I mentioned the most densely populated, and energy is going to be at the center of a lot of these decisions, so I think these are going to be one-off situations.

Having said that, to your very good question, I'm proud of the fact that we've been creative with some of the land and structures that we have. I'll give you one example: there's a former pharmaceutical campus in Kenilworth, New Jersey, which is right in Central Jersey, and that pharmaceutical company moved to another location, happily in New Jersey, a number of years ago.

A developer came in, bought that, and has been trying to figure out how to populate it, freshened it up, updated it—and lo and behold, they've struck a very good arrangement with a big AI player. I'm not sure it's public knowledge yet, so I'll leave out the name, but it's a classic example of a creative solution.

We were right behind that, supportive of it. We had to make sure, by the way, to your prior question, that the energy situation was sufficient, and it was.

Plopping a data center on or near a nuclear campus seems to be another potential—I emphasize potential—creative solution. I think, as a general matter, Nathan, if this keeps going the direction it's going—and I'm assuming DeepSeek is an interesting development but does not materially change the trajectory of the industry, and if I'm wrong about that, then probably all bets are off—I think this will lead to, I'm not sure in New Jersey or not, a renaissance of nuclear technology and nuclear power. That's something I think all of us need to keep a close eye on as well.

Yeah, my cards are on the table: I hope so. What about thinking about things like balancing—I mean, you've kind of mentioned international relationships that you have developed over time—and obviously a big boogeyman in a lot of AI conversations is sort of what's China going to do?

You alluded to DeepSeek, and obviously they're developing their stuff. I think you're right, by the way, that it doesn't signal a trajectory change for the AI wave as a whole, although it is a notable data point. It's more on trend than off trend.

But one of the big questions that people always have is, "What happens if China gets there first?" Then Chinese values are the values in the AI, versus American values. Obviously, we want to have our own values as primary in the AI.

How are you thinking about maintaining our values—forget about China for a moment—in the presence of AI? What about, for example, policing? You could imagine policing could get a lot more efficient and maybe effective, but you could also imagine it might have China-like consequences, where all of a sudden you've got cameras everywhere and maybe life doesn't feel as free as it used to. Where do you think we should go on that, and what kind of moves are you making?

Nathan, I'm not sure when this is going to air, but I'll say, as of the moment we're taping it, American values have not—I’ll go out on a limb here a little bit; don't get mad at me—had a good 3.5 weeks. Let me just say that, unfortunately, and I fear where we're headed in our own—the stuff that we can control.

More broadly, you could not be more right. I've lived in China, I'm a former U.S. ambassador, and I've dealt with their missions overseas and seen how they approach things. I've seen how they approach intellectual property, even right here in New Jersey.

We are a massive per-capita research state when you look at Princeton, Rutgers, the other higher eds, plus the big corporates. Per capita, I think we're at the head of the class in terms of research dollars spent. They approach security, you're absolutely right, in a very different way than we do.

Privacy is not an issue; the collective matters at the expense of the individual. That's always been their MO and always will be, and AI will only further all of the above.

As a nation, we should be incredibly proud of the technology reality in our country—this sector, our leadership. That's a badge of honor across the board, not just in generative AI. But there's no question in my mind we're in a race. No one's going to hand us the title, and there's no entity that's more compelling in that race than the People's Republic.

We've got to have our eyes open, and God willing, we find our footing on our own values and make sure that the technology that we develop and the norms that ultimately exist are ones that we feel proud of as Americans. But we can't assume that's going to happen.

Is there any hope, or do you have any suggestions, for people who—and this is me, by the way, asking for myself—what can I do to build bridges with China and try to promote whatever cooperation or collaboration is possible, realizing it might be limited, as opposed to what I see so many conversations doing: just seemingly ratcheting up the tension and increasing the intensity of the race?

My big worry there is that the race ends badly for everybody—not that one team wins or loses, but that we all end up in a dystopian outcome, a la the nuclear arms race, where we've still got tens of thousands of nuclear weapons in the world. I would hate to see us layer on an AI version of that.

So what, if anything, can people do if they're not in super-powerful positions to try to steer that in the right direction?

I agree with your premise, by the way. I don't think this is going to be Eagles 34–0 in the 3rd quarter, where there's going to be a 1st-place winner and a 2nd-place winner. But I don't think it's going to be a shutout, no matter who's on top.

If it is dealing with China generally or Chinese entities, I'm not sure I've got good advice. I admit this—and again, I lived in Hong Kong, I did business in China, made some good friends and good contacts there over the years—I'm a big skeptic on intellectual property and intellectual-property theft. Institutionally, I fear there's little constructive way forward for guys like you and me.

That's a sobering perspective. It sounded very negative, so I appreciate you gave it to me as you see it.

You mentioned—and this kind of ties together, I think, the sort of international competition but also American values—you mentioned earlier that you ideally would prefer to see AI regulation be federal. It seems like conventional wisdom right now is that's probably not going to happen in a meaningful way in the short term.

In the absence of that, are you interested in putting things into law in New Jersey that would try to encode some values, or put some guardrails on AI development?

I've got to give you a TBD on that one. I probably would have had a different answer before the election than I do now. I'm not saying this was black-and-white, good versus bad. I did share worries with others that the Biden team had gone too far in terms of restricting access.

One thing we will clearly do here is make sure that, as AI becomes more pervasive, its use is consistent with laws that we already have on the books—antidiscrimination, et cetera. These are basic, fundamental pillars of value that we have in New Jersey already. As this rolls out, those are respected and adhered to.

Whether or not there's more of an AI-specific regulatory playbook at the state level, I would prefer, without question, a federal solution here. You probably saw California came close to doing something a couple of months ago and pulled back.

I'm not sure where that stands today. I was following it very closely then; I'm following it less so today. But if you listen to folks on both sides of that, it had a big potential fallout. So, again, to be determined.

We'll make sure that, in whatever form it takes, it's adhering to our current laws. What it ultimately looks like in terms of a state-level regulatory regime for AI: strong preference for the federal side.

I know we're coming up on time. Anything we didn't get to that you want to mention, or any problems you think more people should be working on? Any closing thoughts?

You're the best. I've really enjoyed this. I think Natalie is with me, and I think it makes sense at some point, whether it's in a podcast or privately, that you and our chief AI strategist, Beth Noveck, connect.

She's really special. She advised David Cameron, Barack Obama, and Angela Merkel, and now me. I'm not sure how she landed on me, but she did. I've really enjoyed this, and I'm not sure there's anything else that we haven't said.

Although I'm going to say something which my gut tells me you may agree with: 3 months from now, maybe 6, this conversation will look stale. So one thought is maybe we do this again in 3 or 6 months.

Sounds great. I would be happy to connect with Beth and welcome you back anytime. I would just say to you that I appreciate the fact that you are as engaged in this as you are. You almost certainly could have gotten away without doing all this, and I think it is really critical that more and more leaders start to engage in a substantive way.

If there's anything I would leave you with, it would just be: please keep wrestling with that possibility that things could go really fast.

I think the DeepSeek moment in particular—it was on trend, but one of the things that really became clear there is how simple a lot of the techniques are that are driving the advances. My expectation in 2025 and 2026 is that, as much as our minds have been blown the last couple of years, we're in for at least as much additional mind-blowing stuff in the next couple of years.

I don't know where that leaves us, but I do think it's really important that people in positions of leadership and power ask the hardest questions.

It's a great point to end on. Said maybe differently, but with a similar sentiment, the distance that we've traveled from ChatGPT to today will be dwarfed by the distance we travel from today, say, 2 years from now.

I'll be frustrated because I won't be governor—I've only got another 11 months—and I'll want to stay in this space somehow. That's on my back to figure out. But this feels like it's accelerating, and that's the one takeaway I've gotten from talking to you today: you can't let any grass grow under your feet in this space.

Absolutely. Well, Governor of New Jersey Phil Murphy, thank you for being part of The Cognitive Revolution.

Thank you, Nathan. Talk to you soon.