AI, Energy and Climate: Nuclear security in Iran, energy innovation and AI: Ernest Moniz

right here on the DSR network.

“And there's no better way to enjoy the DSR network than by becoming a member.”

Members enjoying ad-free listening experience, access to our Discord community, exclusive content, early episode access, and more. Use code at DSR26 for a 25% off discount on sign up at the DSR network.com. That's code at DSR26 at the DSR network.com/buy. Thank you, and enjoy the show.

To the AI, energy, and climate podcast, a special series from the DSR network hosted by David Sandalo, and inaugural fellow at Columbia University Center on Global Energy Policy.

Join us as we talk with leading experts to explore the intersection between these critical

issues that will impact the future of each and every one of us. Hi, I'm David Sandalo, this is the AI, energy, and climate podcast. Ernie Moniz has had a remarkable career. He served as U.S. Secretary of Energy from 2013 to 2017 during the second Obama term. Before that, he had a distinguished academic career, serving as chair of the MIT Physics

Department and founding director of the MIT Energy Initiative. He's also the founder and CEO of the Energy Futures Initiative, and serves as board chair of the nuclear threat initiative. In my experience, there are very few people who combine the sophisticated understanding of policy diplomacy and technology that Ernie Moniz brings to energy issues.

So when the zero week organizers asked if we would be open to recording an episode of the AI energy and climate podcast on the ground in Houston, I was delighted when schedules were blind and we were able to find time on Ernie's very busy schedule. He and I had a chance to discuss the war in Iran, focusing in particular on implications of the war for a nuclear security, as well as a wide range of topics related to AI.

I hope you enjoy our conversation. Ernie Moniz, thanks for joining us, and welcome to the show. Legislative David. So you serve as board chair of the nuclear threat initiative. We're recording on March 26, 26. As of today, how do you assess the impact of the war and the Gulf on nuclear security?

“Well, I think it's important first to go back and establish the baseline.”

And this is not new today. There are certain new realities of stemming from last June, and from the years before that. Because frankly, Iran started in 2021 to build up its 60% enriched uranium stores. And by June of 2025, they had over 440 kilograms, roughly half a ton of 60% enriched uranium. Unfortunately, and I've said this directly to the media, I think the media has done a disservice in their discussions about this,

because if one listens to the media or reads the media without knowing a fair amount, I think most people would take away the impression that you need weapons grade uranium to make a weapon. That is a false proposition. 60% enriched uranium can be directly used in a nuclear explosive. Consequently, Iran had, in my view, already crossed a red line in terms of being able to make a weapon. They didn't need to enrich further. Enriching further to 90% of which is the usual weapons grade threshold is better.

You can make a more efficient weapon, a smaller weapon, a lighter weapon.

“If you want to put that weapon on the top of a missile, it's very useful to have a small compact weapon.”

But as a reminder and/or new information, as the case may be, the bomb that the United States used in Hiroshima was not weapons grade uranium by that definition. It was not delivered on a missile. I mean, there wasn't a missile at that time, but it was a big lumbering claim with a big lumbering bomb.

And it had significant destructive power. And by the way, it was never tested.

So I think it's really important. It changes one's view when one realizes that they had already crossed that red line.

“So in that context, I'll be honest, I have never either condemned or endorsed the military action that we've seen against Iran last year and this year.”

In the sense that, again, I think they crossed the line, they rolled the dice, the dice came up, snake eyes and let's just move on and figure out what to do from now. Now, from now, from now going forward, the reality is, I have to say, I don't think we are in a great position. Obviously, militarily, there's no doubt about who the superior actor is, but frankly, we have through action and through talk, given Iran the situation where their objectives are pretty clear.

And in some sense, minimal, a survival, b, their objectives are not to win this war. It's to prevent a third military strike.

“And so I believe the actions they take now mostly against energy infrastructure, and of course, the straight-up war moves, that is their deterrence strategy against a third attack.”

And certainly opening up fully opening up the straight-up war moves is not an easy proposition for the long-term.

We don't want a long-term presence there for sure, even then it might not be so simple.

And secondly, while the original actions would have certainly continued the disruption of the energy markets for a while, the kinetic activity against some of that infrastructure, such as the LNG export capacity of gutter with the destruction of two large trains, they're talking up to five years to fully recover. So it's going to be a very complex time. You know, in a narrow sense for U.S. LNG exporters, there may be additional market share to be had there, a kind of an irony of what Iran is doing. But overall, it's a very bad situation. Oil prices, natural gas prices are high.

But the lesson that should be learned, of course, is that as the G7 and the EU wrote after the 2014 Russian invasion of Ukraine, energy security explicitly involves going to lower carbon sources. We've had that lesson put it in front of us many times before without actually being picked up in a serious way. I'm not saying that we haven't made a lot of progress in deploying low carbon technologies, but I don't think it's been associated in the long-term with the energy security lessons that we should be learning.

So we'll see, it's all, it is true that today we have techno-economic possibilities in going to low carbon that we didn't have in some of the previous oil disruption. So we can, we can be optimistic that this lesson will be learned with a bit more staying power, but I'm for Missouri, show me. Well, that's a key topic. I want to come back to that, but just to stick with the nuclear security question for a moment, President Trump has threatened to obliterate, I think that was his verb Iran's power plans, if the Iranian government doesn't fully open the streets of Hormuz.

“And I wonder if that happens what threats do you see from a nuclear security standpoint, particularly with respect to the light water reactor at Bashir?”

Well, first of all, I would hope that Bashir would certainly be off limits just as the power plants in Ukraine have been an area of major concern, because nuclear power plants and war zones are not comfortable, a bedfellows. Now, it would be, in my view, also a mistake to obliterate the more conventional thermal plants, and of course in Iran, they do use their domestic gas or resources in the power sector.

And that's not, I think, in my view, that's not a great idea. It would obviously lead to great suffering of the population. Now, one can always say, and some would say, well, that's the step we need to have, you know, regime change. I don't think regime change is, if it happens, it happens, but, and it has happened in some other situations in other countries, but I don't think it is a reliable objective, and certainly in the Iranian situation, we all know that the Revolutionary Guard has a very, very strong position.

When negotiated the Iran agreement, the nuclear agreement in 2015, our estimate was that the Revolutionary Guard controlled somewhere between 20% and 30% of the Iranian economy, given the sanctions, the since then, the extreme pressure approach, we actually think that that's going up, maybe it's between 30% and 40%.

“So the issue is that not only are they, you know, a military organization that reports not to the elected government, but to the Supreme Leader, but in addition to be perfectly blunt, they have a major economic interest in maintaining the status quo.”

And that has played out in a very unfortunate way to an opposition that we all know does not have an apparent organization, it does not have weapons, it does not have a charismatic leader that can organize them. So, it's a very difficult situation, and I think that we need to find the diplomatic offer amp, I don't think it's easy, because again, I think Iran has in itself interest the idea that it wants to royal the energy markets for a while as a deterrent to further military action. Yeah, and of course, you were a key player in negotiating the joint comprehensive plan of action in 2015, so you bring a lot of knowledge to this topic and Secretary Moniz, just one final question here.

Let's say there's an attack which wipes out a fair amount of Iran's civil power infrastructure, but doesn't directly hit the plant it by share with what type of risks would that create with at the civil plant it by share.

“Well, I think it would certainly put at risk things like as we've seen in Ukraine, like back up power that is typically needed to maintain, you know,”

cooling of the reactor, cooling of spent fuel storage ponds, and in Ukraine, there have been periodic problems with getting back up power to the plant.

The plants typically, I don't know, frankly, directly at Bashir, but I'm assuming that similar to the plants in Ukraine, they do have back up on site power like diesel, but that's equally vulnerable to kinetic attack.

The storage of the fuel for those backup generators is exposed. These are not exactly hardened facilities because they're not built for war zones.

“So I think there is, there certainly is risk. We need to maintain enough backup power back up energy to keep that plant operating without creating a major risk, which, of course, would be very bad for Iran and the people, but we all know.”

The nuclear power is pretty connected internationally. We've seen how Chernobyl and Fukushima, for example, had enormous consequences for the regulatory approach, the costs at American plants, let alone Japanese plants, which have since Fukushima, had the down for a long time, there, you know, a fair number of them are now finally back up, but that was quite a while ago when Fukushima happened. A decade of difficulty and also bad for carbon. So anyway, I just think it's playing with fire to start fooling around with the electricity system broadly, not just that Bushir, because it is part and parcel of the security and safety of that facility.

Okay, well, this is the AI energy and climate podcasts. So let's pivot from Iran to talk about AI, but perhaps stick with the top AI two vowels in the word Iran.

Well, on the risk side, I would say the risk that is most often identified for AI and the nuclear weapons systems revolve around the command and control system and AI, of course, AI enhanced cyber attacks are a concern.

It may not be talked about a lot, but it's happening today. Our major infrastructure is under pretty intense cyber attacks. That's true without AI, it's even more true with AI.

The other side of the story, however, whether it's nuclear weapons or civilian infrastructure, including energy infrastructure, I at least, and I think many others, but I'll just speak for myself, believe that in the end AI is going to have to be part of the cyber solution.

It's hard to see how, to me, it's hard to see how AI and neighbor cyber attack of vector is going to be defeated by anything other than an AI system.

It's our bots against their box, and I think there's progress on this. There are, I mean, part of AI solutions and also synergistic other ways of deflecting the cyber attack vectors, but this is probably the number one opportunity and risk in my view.

In terms of security, now, there are other issues of AI integrated with the energy system, for example, let me give three examples. I can count to three.

One is certainly AI, machine learning, and possibly other emerging technologies like robotics, etc., will be important for improving the reliability and resilience of the grid. If we've been looking for a long time at the kind of sensor network, etc., that can give early warning of problems on the grid, can distribute loads in ways that enhance reliability, etc.,

“with AI, I think we will finally reach this stage where all that entire system becomes a decision-making tool.”

For better or worse, I think we will see a lot of autonomous decision-making in the grid with AI. The second example that's fairly obvious, again, and the theme of autonomy is never far away in my view from the AI discussion. Obviously, in the transportation sector, it's AI that enables this whole move towards autonomous vehicles, and way beyond waymo, and I think it's important to recognize that in a way that, in my view,

“it goes beyond even the opportunities in the grid. I think that the AI enabled autonomous mobility systems will actually provide qualitative changes in the way we organize our lives.”

Things we even can't imagine right now will eventually be doable, because right now we're still in the baby stages. You can buy a fully autonomous Tesla, and fully autonomous means it's got cameras making sure you're looking at the road, and you're not looking at your iPhone, etc., etc. When we reach the confidence level of being really autonomous and opening up transportation for those who currently do not have the agency, very old people, very young people could be in that group.

“The ability to do entirely different things while you're in the mobility stage. I think we've not yet seen the creative business models that will emerge from that AI enabled mobility.”

Then I'll give one more example, which is probably less visible to the average person who just turns on the weather channel to find out what's happening outside, or what's going to happen in the next day or two days or three days.

Building out from, you know, the laws of physics, how the weather was going to go, we are way into the new age where, as a physicist, I can say, "Ah, screw the physics, we'll just do data, we'll do pattern recognition, we'll predict whether on the basis of the patterns that have previously led to snow or cold or rain or whatever extreme weather events.

“Some weather events are still a little bit difficult for either system, but we are seeing already today, in many cases, this kind of AI approach to weather prediction being superior to the physics-based approach.”

And of course, if we can enhance weather prediction capabilities, enormous implications for the energy systems, obviously wind farms would be an obvious example, but early warning systems, all kinds of implications for energy and for other aspects of social organization.

You know, that's kind of an almost unseen revolution that's already happened in the of direct energy relevance.

This podcast is underwritten in part by the U.S. Embassy of the United Arab Emirates.

Its editorial content is completely independent, and the views expressed are exclusively those of participating experts.

It is presented live without editing. For further information about the UAE's efforts in the areas of artificial intelligence and technology, go to the website of the embassy at www.ue-emBC.org and search for UAE-US tech cooperation.

“Thank them for their support. We thank everybody who is supporting this podcast for their support, and we look forward to it developing and growing over time because the issue is so important.”

Oh, or any, I have a lot of follow-up questions. This is making me wish that this was one of these three-hour Joe Rogan podcasts. There's so much in which you just said, but let me let me just start with one. You talked about autonomous decision-making on the grid and question is, how do we get to the point where AI is reliable enough for that type of autonomous decision-making on systems like the greater say industrial processes.

If an AI system makes a mistake and recommending a Netflix movie to me, that's pretty low consequence. I just don't pay attention.

But if an AI system hallucinates or makes a mistake and operating the grid or chemicals plant, for example, that could cause serious problems.

“One, are we at the point today where AI systems are sufficiently reliable for autonomous decision-making and those contexts, and if not, what will it take?”

Because at the next level, because the answer to the question is, are we ready, is different from the question, is the system already more reliable than human judgment? So, if I go back to the mobility case, we know we've had some bad accidents from autonomous vehicles, but certainly I think one can argue, and many do argue, that even with those, it's actually a much lower accident rate than you would get with human drivers. So, I think the standard is very high, and it should be very high, I think we're probably not at the stage, by definition, with a public, is ready to accept that pure autonomous decision-making in either of these contexts, even if it is overall safer.

So, I think, you know, it's a combination of technical, social, and emotional issues for independent public perception, do you have a view as to whether AI tools are now at the point where they're sufficiently reliable to be doing decision making, say on the greater industrial facilities?

“I think right now, we're at the stage where, operationally, again, given all the factors, including the non-technical factors, we still need some human backup.”

That's true in all the domains, including what we've started with, by the way, nuclear weapons decision-making. Let me answer it in that context. You often hear, and I'm not arguing against the issue of, there will always be a human in the decision loop.

You can extrapolate this to the energy situations, but let me focus on the nuclear weapons.

Okay, great. A nuclear weapon will not be fired unless a human, in our case, the president of the United States, makes that decision. But who prepares his briefing, his or her briefing paper in one millisecond? It's called AI. And you can argue as to whether that AI conditioned environment gives the human more or less latitude in making the decision. What do you think? I actually think that, again, in the nuclear weapons context, I think the president, he or she may be a bit more boxed in by providing, by having a very detailed memo tailored to the situation, whatever it has been detected incoming fire, for example.

“I think it may almost make it more difficult for the president to exercise kind of the gut feeling judgment, such as we've seen before, in both the United States and in the former Soviet Union, we've seen misinformation fed into the system.”

And the human quotes in the loop actually saved us from catastrophe, including by the way, the famous story in the Soviet Union, where a fire officer was given the order to fire and didn't do it because he correctly surmised that there's something fishy here.

It doesn't make any sense. Why would the United States, in a period of peace in his thinking, send one missile? If it's 200 missiles, well, that's a whole myth. That's a different decision.

“But the question is, will AI have the sophistication to not box in the decision-making process, and I think we need to be cautious on that, and have every step in that intellectual supply chain be checked, at least for now.”

Well, a similar set of issues arises with autonomous vehicles, although obviously the impacts are very serious, but less, less huge, if in the context of what times vehicles may are in nuclear weapons.

But you talked about AI and autonomous vehicles, and I wonder two questions. So first, how quickly do you see autonomous vehicles deployed in the years ahead in the United States and around the world?

And then second, what do you think are the sustainability impacts of autonomous vehicles? Not obvious that ABCs will lead to lower greenhouse gas emissions, for example, they could lower barriers to driving lead a lot more individual mobility and increase pollution and greenhouse gas emissions. First question is, what do you see as the trajectory or second was the sustainability impact? Well, I think, first of all, it's a fact that they exist, and then they're being tested, we say, in various cities in the United States, China is obviously taking, having big strides in this, and how fast will it be to scale?

I think that the human track record has always been to underestimate how fast these technologies would develop. I frankly believe we are going to see, we're going to be surprised at how rapid I think this is a curse.

Clearly, if we continue to have a significant emissions from the grid, well, then the autonomous vehicles, if it increases miles driven, etc., etc., can certainly have some negative impacts.

But clearly, in my view, we need to create markets in parallel. We can't do things in series. So even as we are moving towards decarbonizing the grid, I believe we have to create the markets for autonomous vehicles, for hydrogen, etc., for example, on hydrogen as an aside here,

“I think that the tax credits that were put forward following the IRA of so-called 45V tax credits, I think were misplaced.”

They, in some sense, said, we have got to start immediately with, you know, additionality in terms of carbon-free power, etc., and I'm all for, you know, that transition, but my view was the way it was written, it has, it may be history already now, but the way it was written, it impeded the growth of the market. And I'd rather go for the home run, get that market developing even as we decarbonize the grid, and they will eventually come together, we'll have an established market, and we'll have carbon-free electricity, let's say.

So whether it's charging vehicles or hydrogen, etc., I believe that we need to accelerate the time it takes us to hit the grand slam, and to have the right markets, the new technologies, and the clean grid, all come together, sooner rather than later.

“Well, I think first of all, it's very important to realize that progress towards a lower carbon economy has been very asymmetric in the power sector, for electrons rather than molecules.”

The reality is, we have a ways to go, but we have a lot of options that have been developed, that have had cost reduction in the power sector, you know, renewables, plus battery storage, etc., we have a big push now on nuclear energy, we'll see whether that gets across the finish line, we have possible big breakthroughs like engineered geothermal becomes a big deal.

Nuclear fusion becomes a reality, etc., we have a lot of options that look to be in a very, very sensible, you know, economic range to do things.

I would say if we go back to molecules, it's been less encouraging. For example, a key example would be advanced biofuels. Here we are, 45 years after starting to try to develop advanced biofuels, we've gotten, you know, roughly speaking nowhere in that, in terms of affordable.

Sure, we can make them, but not at any cost that society will bear. So that's where hydrogen comes in now as another potential pathway, roughly speaking to molecules.

And right now, green hydrogen, you know, by splitting water, with ideally, with maybe renewables, they have a carbon-free supply chain for that hydrogen. It's just too expensive, I mean, we're in the same spot.

“There's no reason to think that those costs can't come down, which goes back to my point, that's why I say we should be focusing on developing the hydrogen market.”

That's how you get cost down. It's by getting a big enough market. It's by working down the learning curve, getting the manufacturers innovating, not in giant breakthroughs, but in incremental improvements that keep driving cost down. And that's why I say, I'd rather gamble on that and get my carbon-free grid to catch up with it. On the other hand, as we know, there are many colors in the rainbow for hydrogen.

That is much closer to economic viability. There are other approaches, such as turcoise hydrogen, which is pyrolysis, high temperature pyrolysis of methane, producing a solid carbon plus hydrogen.

“And a solid carbon, of course, means disposing of it is much easier and maybe using it in economic applications is much easier than with carbon dioxide gas.”

There are breakthrough potentials. We really don't know yet whether school is in or out on geological hydrogen, either mind geological hydrogen or produced subsurface produced hydrogen, running water over iron bearing minerals, et cetera. We have that breakthrough, every indication is, it'll probably be in a very nice cost zone as well. So I think we don't really know, but I do feel that in a broad-based approach to developing hydrogen markets, we should also be thinking that we need those markets, especially to bring down the green hydrogen costs.

We will now be able to provide more than 50% of the population with less population population in the population or even more in the area of the population.

When we are still there, we are able to provide the Dine 9 Cope for a meeting with the population in the population of the population, based on the population of the population from the year 2025 in Rosbrotanian and EU. A double-chiech-barger is a promise and a co-operative of 99, although I don't think so.

“Once again, already if so many follow-ups to which you just said, and so let's start with geologic hydrogen, because I think there received wisdom for many years, so that was way too expensive, would never be possible.”

The oil and gas industry certainly amusing AI tools for the subsurface, very successfully, in recent years, do you think that AI tools can help bring down the costs for geologic hydrogen and make the reality in the next decade? Let me first broaden the question beyond geological hydrogen. In fact, at the nonprofit that we created in 2017, the energy futures initiative, we are advocating a more holistic look at the subsurface of the earth as a new energy frontier.

Geological hydrogen is whether natural or produced is certainly a subsurface phenomenon, but so is engineered geothermal, so is critical minerals, so is carbon sequestration,

“so our deep bore holes for nuclear waste disposal, you know, we could go on and on. I think people haven't stepped back and said, the subsurface is a new energy frontier, and think of it in that more integrated way.”

In that context, AI is certainly a very, I mean, AI plus the appropriate sensors, et cetera, again the theme, usually it's about integrating these technologies, so but so that kind of AI grounded system has a much bigger role to play than just geological hydrogen, I believe in understanding and mapping the subsurface if for very different geologies, we will be looking to design a above-surface systems that serve the subsurface in terms of providing energy because there are huge implications for the above-surface, how one does regulations, how the states approach regulation.

When you go below the surface, it's really hard to see those state boundaries, but then when you go to the surface, boy are they real, and so that's an example of how we need to integrate our thinking below and above the surface, but the subsurface is an incredible opportunity for new clean energy. You referred to integrating these technologies, what did you mean by that, which was technologies? Well, well, we listed all of those subsurface opportunities, many of them will involve very similar geologies, and yet they're all studied in their own, I guess, stove pipes are like wells to have, I guess a bit of a pun, so I think that the question is where are the synergies, things like on the basic science side?

We go to the other end in a certain sense, like the above surface regulatory issues, we ran it at EFI, we ran a workshop on geological hydrogen. We had a bunch of state regulators there, a major focus was on the regulatory issues, and it was stunning.

“How first of all, how very different states approached anything in the subsurface, many states had no classification that was relevant to geological hydrogen.”

I've never been thought about regulations were put in place for natural gas, and well, what's this hydrogen stuff? How do you regulate it? How do you get any kind of uniformity? How do you cross the state line? Something that's already clear in the, let's say natural gas and oil exploration, but it's also clear for sequestration, a very small number of states. Have so called unitization in terms of how a subsurface reservoir is to be managed, suppose two thirds of the landowners on the surface say, it's fine, go ahead, give me my royalty and one third of the surface landowners say, no, no, you can't put that, you can't touch that reservoir under my land.

Well then, what do you do? Some states have thresholds, if a certain number of landowners say yes, then you can go ahead, you can put you, you can do your CO2 reservoir for sequestration under there, some places you can't. So there's a lot of variability, and I think getting a more coherent unified view of that, we're talking about, ultimately we're always talking about some fluid under the surface, and how do you manage that fluid, how do you regulate it, how do you do the science, et cetera, is I think a missing part of the clean energy puzzle.

Let's step up even higher level of generality, you've discussed a number of different clean energy technologies and approaches here, including better grid management, autonomous vehicles, what's happening in the subsurface and more.

“How do you think we can reduce the barriers to AI helping advance energy innovation in all those spaces? What does it take for AI to make a big difference in accelerating AI innovation there?”

I think the, frankly, the main thing is just doing it. You know, I think we've, we've got a lot to learn, and we're not going to learn by, in any way, other than doing it.

So I think the question is, how much can, how much do we learn by applying AI and then refining it in the way that we need fit to purpose? There's no doubt about AI itself continuing to to improve to grow at breathtaking pace.

“The, I think the whole question is, what do we learn in terms of what it does do, what its limits are at any given time and how we address that and go forward. So I'm just in physics, I'm a theoretical physicist, but I believe in doing experiments.”

Well, we're, we're starting to run out of time, and I want to turn to another topic, which is AI in education, because you spent the fair amount of your career as an educator, you were chair of the MIT physics department. And no number of people have been thinking a lot about the role of AI in education. It's very challenging. I know in my course at Columbia, we've had to change the way I teach because AI tools today can do a lot of the assignments that I used to give over many years.

And there's a core question about what should we be teaching in an era in which AI tools can write graduate level term papers.

Do you have any thoughts on that topic? How should academia, how should teaching more broadly adjust to the new world of AI? Well, you know, you're making a mistake of asking an old person about this as opposed to a young person, because the answer might be quite different. There's great wisdom in it. I believe that more and more every year. Well, yeah, you have to, but let's face it, but in a serious level there is a divide between different different ends of the longevity spectrum.

And, you know, just to go back to personal experience, this is not education per se, but, you know, I mean, I'm using AI every day, not consciously, but just the way I use a search engine, you know, Google today is so different from what it was several years back.

“It wasn't a conscious decision. It's just that it's become, I can now ask relatively unstructured questions of a Google search. You couldn't do that 10 years ago, right?”

And, and get some pretty sophisticated answers. As an example, by the way, going back to earlier in the discussion, I just did a test a few days ago, I asked Google to tell me about the evolution of whether prediction. And again, we exactly the lecture about physics-based models versus AI-based models, et cetera, et cetera. You know, I knew something about it and it was accurate, you know, so I think we have to learn and, and, and I don't know what the answers are, but we have to learn how to do a combination of structured classroom approaches in which AI is not accessed.

To, you know, kind of test where students are in terms of a deeper understanding on their own. But I think that deeper understanding, in some sense, in the end, is about being able to tell fact from fiction in AI.

Because you can't, you mentioned earlier hallucination. You can get misled. And you can go down the wrong fork in the road and never get over to the other road with AI.

“And that's where I think you need to know enough. And that is, I think, a major role of education. You need to know enough to know and to sense when this doesn't sound right.”

Let me try this again. Let me refresh things a bit, et cetera. But the access to the data that one has through AI, it's just breathtaking. It's not fully formed. It would be nice if one didn't have only language models, but maybe mathematical models, or, for example, it's still got something to learn, but it is just breathtaking. And I think it's foolish. It would be foolish to educate students not to understand how to use those tools in a sophisticated way.

I think this is the key point. And I think it's very challenging. I know when I read answers that I get from ChetGPT or Gemini or Claude that are in my area of expertise.

It's easy for me to identify what the impressions are, what the errors are. But when I read answers, in areas that I don't know a lot about, I can't do that.

“And I think it's a big challenge for the young generation of students right now to develop the expertise to be able to discern what's right and what's not right in AI answers.”

It's a core challenge that I think educators have to figure it out. Well, yeah. And it's also the issue. And it's a broader, a broader problem we have with social media is that what a hope that this doesn't end up channeling people to go with they wanted to go anyway, as opposed to challenging them to think more about maybe places they didn't want to go. And you think the AI answers will do that. I think sophisticated use. So again, I will at least today where no one near the stage where I would trust AI fully in that way.

As you say, like, especially in areas that you don't have quite the depth of knowledge. But again, I think a big part of education today and going forward is going to be to be able to have a sense of what when when you need to question a bit more deeply. Well, we need to wrap up. And I always close this show with two questions in the back. You've already partially answered the first question. I always close with, which is how are you using AI these days, but maybe just say I say another word about that.

Yeah, I would say the most important thing for me is the way that search engines can manage relatively unstructured questions. And you know, that's a fancy way of saying what I do every day.

I want to, I want to learn something. Google, you know, and ask a question questions again, as I said earlier, the questions that you never could have asked Google or some other search engine, but I just say Google because it's like, it's like Kleenex, you know, it's it's become a brand. And I think it's just remarkable. And it helps me every day.

I highly recommend it to everybody, but but our final question is, could you recommend three books reports or articles to our listeners could be anything older, new. Well, I'm not sure it's recommend, but I will mix observations based upon books.

“One actually series of books that I think has an interesting lesson for us, there are also very good books, but is the series of thriller novels written by David Ignatius.”

You know, the well-known Washington Post opinion writer, but he's also, you know, he's a fantastic novelist. His novels aren't exactly fiction, they're pretty heavily researched and based on what's happening. But you know, I think the point is, and David's a good friend, but I started getting interested in his books at the recommendation of a very good friend of mine, who was a CIA case officer.

And he recommended David Ignatius first book called Agents of Innocence, written way back in the 80s.

And my friend said that it was the book that most accurately reflected what a case agent, what a CIA case agent actually does. And it was quite opening, and the setting was Lebanon involving Palestinians, etc. I'm not going to go further into your book, but it is a wonderful book.

“But what I want to do is skip ahead to what Ignatius, let's say I think is the most recent book, and not just skipped from his first book to his last book.”

It was called Phantom Orbit. It was not AI, but it was about emerging technologies, in this case Space Technologies, and the whole issue was the competition with China. And by the way, earlier, maybe 10 years ago he wrote a book "Quantum Spy" based on quantum computing. So it's interesting this arc going from kind of traditional intelligence and Lebanon, the Middle East, to emerging technology in China. And I don't know to what extent that was conscious, but it shows how history has evolved over these 40 years.

Amen, and the director is also a great book. Yeah, that's also one of the famous books.

“Another book I would note, very topical, I think it's about a year old, maybe a bit more, by my MIT colleague Susan Solomon, called Solvable.”

And that book basically drew upon her considerable experience in looking at earlier major environmental problems.

Like the ozone hole, you mentioned some of that earlier, in fact, and in that book, she tries to draw the lessons for how we would approach climate. And emphasizing, you know, pragmatism, but also the fact that, and I think we have failed at this narrative, making the impacts felt at a personal level. And very importantly, in my view, and this is something, by the way, to go back to our earlier discussion, is also true in the nuclear weapons area. People have to understand what a problem is to scale up the problem, and that they have agency, even as individual citizens, to do something about it.

How can an individual citizen have agency on nuclear weapons issues?

The reality is that if you look at major initiatives in the non-proliferation space, for example, many of them have started in the private sector, non-profit sector, not in the government.

And as an example, if I mentioned energy features in the initiative, I've also been involved, and I'm still, you mentioned earlier, I'm still the co-chair of the board of the nuclear threat initiative. As an example, go, and this goes, this, this will resonate with your DOE checkered history. If you go back to post 9/11, it was in the private sector that suddenly there was a focus on, oh my god, there's, you know, 50, 60 kilograms of HU sitting in Serbia, kind of in a like in a war zone or a former war zone, not with great security.

It was NTI, frankly, using funding from Ted Turner, okay, he's not a typical average citizen.

“He's got a little bit of deeper pockets, the most people do, but still, it was, it was seating what became a US Russia joint operation to put those, put that HU out of danger.”

And as you know, that was fundamentally the beginning of what has become billions of dollars spent by the Department of Energy to rescue, especially HU from a whole bunch of countries. So anyway, going back to solvable, it's about what we can do in the sense of having agency, having a personal narrative, actually, okay, I'll, I'll make another aside on this, that at EFI, we are in the early stages of taking that advice and trying to look at how we can change the narrative on a climate.

“And I'm saying something that's not quite public, but okay, now it will be that, that in that we are collaborating, we will collaborate with a, perhaps the world's largest public relations firm.”

Amanda Adelman, among other things, runs something called the Gen Z Labs with a large cadre of Gen Zers across the world who are available to be poll, to be tested, to be, you know, market tested, etc. So there's an example of, you know, this doesn't require governments. We're going to try to see if we can't reach a new demographic cadre in this, in this, in a changed climate narrative.

So that's the second one. And then I'm going to end with, okay, a third one.

One published book because I'm still writing it, but tune in, I can't say when for, do you have a title? The title is, I think, the publisher has agreed, in principle, that the title would be herding dragons, and the dragons refer to Enrico Fermi's reference to the dragons inside atomic nuclei, and that gets back more to the security questions.

“And there's three examples, I'm not saying any of them are my favorite books, but, but I think they all have, they're very, I think they're good books, and they have very, very important messages.”

Secretary Ernie Monies, I always learned so much talking with you and listen to you to you. It's been fun. Thank you so much.

This has been the AI Energy and Climate Podcast, a special production of the DSR network. [Music]