LLNL’s “AI in Eight Pages” Team on Making Sense of Artificial Intelligence

Host Yolanda Fintschenko, executive director of Daybreak Labs and i-GATE Innovation Hub, home of the Startup Tri-Valley (STV) Initiative, sits down with Cindy Gonzales, Brian Giera, and Caspar Donnison of Lawrence Livermore National Laboratory to discuss AI in Eight Pages, a recently released report designed to make artificial intelligence more understandable for broader audiences.

In this wide-ranging conversation, the guests discuss AI literacy, model limitations, workforce impacts, policymaking, and the importance of human agency in shaping how AI is developed and used. They also explore why the conversation around AI should include not only technologists, but also communities, business leaders, and policymakers.

The episode highlights California’s role in the AI landscape and the Tri-Valley’s growing innovation ecosystem, while emphasizing a practical message: AI is not a magic wand, and understanding its risks is just as important as recognizing its promise. Tune in or watch on YouTube to hear how informed communities can help shape a future where AI serves the public good.

Transcript

Yolanda: We are here today with Cindy Gonzales, Brian Giera, and Caspar Donnison from Lawrence Livermore National Lab, the authors of the recently released report, AI in Eight Pages, and they are here to talk to us about it. Cindy Brian, and Caspar, welcome to the pod. Thank you so much for being here. 

Thank you for having us.  

Brian: Thank you. We're really excited about it. And just to launch right into it, AI in Eight Pages is a primer on the very hot topic of AI. We're totally aware that there's so much content around AI but what's different about this is the audience it's intended for, which is the legislature and their constituents of California, and we're talking about the important technical content in this very important technical conversation, and we're really happy to be part of the conversation. 

And we're on a road show. This is part of that. And one of the things we've felt helps our audience understand how we're coming into this conversation is I view myself as the AI optimist Cindy's the AI skeptic, and Caspar is the realist among us. And being an optimist is not being an AI hype man. 

Being a skeptic is not being an AI pessimist. And what do we mean by realist? It's really grounding the perspectives in this report, the Californian economic sectors and the benefits and risks associated with AI.  

Yolanda: Fantastic. Thank you so much, Brian, for that intro, and I love that you each have a role. 

It is like the Three Musketeers of AI. Maybe we can start with you again, Brian. I'd like to hear from each of you what originally drew you to working at the intersection of AI and policy and real-world application. As part of that answer, you can talk a little bit about your actual area of expertise. I'm sure you're all experts in AI and policy, and I'm very interested in your answer. 

You can talk about what drew you to a project like this that is not so technical, but much more focused on the intersection of something highly technical and informing policy. 

Brian: Okay, great. I recently got promoted into a position where I'm at the technical frontier of AI and manufacturing, and that has culminated from 12 years' worth of work in which, to be honest, I was doing AI before it was cool. Even then, I came into it from what I felt was an outsider's perspective because I have a strange background coming from the chemical engineering field into this. 

And I felt like I had to learn the important things that mattered to me and to my work. There were these new tools out there, and I needed to insert them into my problem set to solve. What do I need to learn to do that? How do I learn their limitations? And all while that was happening, we were deriving a lot of value from it. Then, over the past three years, the whole global conversation around AI changed. 

And as a person who's walked through the process of being foreign to a field and understanding how to process in the information, I realized there was a way in which we could have that conversation with a readership in a report just like this that said, "All right. Out of all the things you need to know, here's what matters to you." 

And that's the angle I came into my authorship for this work. Also, throughout my time as a scientist, we've always been trained to impact our community with publications and presentations and patents and proposals and partnerships, all the Ps, but what was missing was the policy element. 

And I felt like this was a really great entry point into that. What's been so nice is how receptive the Californian legislature has been to this. It seems that they were starving for something like it. At this point, it's so good to have that level of impact in a different way on these topics, and that's been my experience so far. It's only come out in January, so it's been a wild five months. 

Yolanda: I like that you were interested in taking the leap of using your actual technical expertise as a way of supporting an important policy conversation that, and I know I've heard Caspar talk about this a lot, and I'm sure he'll say it again, seems to be in a technical landscape that changes every single day. 

Yes. And you know what was really great, is  

Brian: to finally have something my mom could appreciate. Yes.  

Yolanda: Something  

Cindy: to share with mom. It's "Here, Mom, a Humber report you can actually read." 

Brian: For us, but for you, versus by experts for other experts. 

Yolanda: So, Cindy, why don't we pick up on that? 

I can give you the skeptic context for the skepticism. 

Cindy: the skeptic, the context for the skepticism. So give you the long story since we have some time. I actually started at the lab in 2016. I started as an admin at the lab. I heard about data science through a seminar that the proposal was, do you want to come learn about killer robots that take over the world? 

And I was like, "Why would I not want to go listen to that?" So I went into that trying to learn about machine learning, which was a hot topic at the time, using data to learn and predict the future, which I thought was really cool if we can actually do that. I decided to go back to school while working full time at the lab. 

Got my foundational knowledge in my bachelor's degree, majored in statistics, and then really wanted to make the leap from an administrative staff member to a technical staff member while at the lab. I just remember being so overwhelmed talking to research staff. These are PhDs, and they're so brilliant and smart, and I thought, how am I ever going to live up to that? And when we would talk about topics in statistics and data science, there was a lot of lingo that you had to learn to get up and running. 

And I remember there was a very dedicated mentor who took me under his wing, and we would have biweekly sessions, and I would be like, "What the heck does this mean?" And he'd be like, "In practical terms, all that's saying is this." And I'm like, "Really?" And he's like, "Yeah." I'm like, "Wow, that's great." The technical jargon people use can really inflate the topics, but if you break it down to these atomic-level things, it's really understandable if you take the time to actually learn about it. 

And I made the transition, went back and got my master's degree, but I just remember how impactful that was, right? Somebody taking the time to sit down with me, talk me through these topics, actually make me a productive member of the staff and able to contribute to all of these projects. Then I remember, before times, before AI was cool, trying to talk to my mom about what I did. 

I'm a computer vision researcher. I help robots see things. That's how I would try to explain it to her. She'd be like, "Can you come over and fix my TV then?" I'm like, "No, I work on computers, but not like that." And so when ChatGPT dropped in November 2022, she called me and she was like, "Is this what you do?" I was like, "You're getting it now. You understand." 

And I just remember how impactful that shift was, because now my mom can go onto chatgpt.com and use a tool and understand the impact that machine learning and AI is having on the world. Fast-forward, that was a pivotal moment in my career. 

The next April, Brian and I were working, Brian being the director of the Data Science Institute, me as the deputy director of the Data Science Institute at the lab. We had a fantastic researcher out. His name's Yoshua Bengio. He's basically the godfather of AI. He was talking about the risk of AI, and we brought together people from academia and the government and national labs. 

Brian: some policymakers. 

Cindy: Yes, and some policymakers. And we had this conversation of who's dealing with the risks of AI? And it became clear from that conversation, really nobody. And you look around at the lab, and it is a really unique place because we have researchers in biology, chemistry, nuclear weapons, right? 

We have all kinds of different researchers, and bringing their expertise to bear to try to assess these models and what capabilities they have and the impact that has on the national security domain was something I became passionate about. So I transitioned over to doing large language model assessments, seeing what they knew about the chem-bio area, leading a bunch of projects in that regard, and then more recently taking AI and applying it to operational challenges that we have, right? 

We're an institution that's been around for 70-plus years. The way that we do business is really interesting and complex, and so can we use AI to really transform the way that we're doing business? And I'm sure we'll talk more about that. But I think my skepticism is, yes, we can use AI. It's changing the world. 

But AI isn't this magic wand that people want it to be where they go, poof, now this is done. This thing that took you 50 years to figure out can now be done with the turn of a crank. And so for me, it's understanding the risks of the technology. Everyone in the world has access to this powerful tool. How are they going to use it? Bad actors can have access to these tools. How are they going to use it? And making sure that's a part of the national dialogue. 

And my involvement with this report and how we got involved is we went up, I was invited by our state representative, or what is his official title, Andrew? 

Brian: State representative. yes, our state- From our governmental affairs office. yes, our  

Cindy: Our California state representative for the lab invited me up to Sacramento to talk to people about the importance of open-weight models versus proprietary models. This probably sounds like jargon, which it is, but going to that, realizing that it was hard for me as a researcher to explain the difference between proprietary models and open-weight models to a general audience. 

And our sponsor for the report, Luis, came up to me after, and he was like, "You're able to break out these complex topics. We really need you guy- we really need people who can do that in the conversation." And I'm really grateful for that opportunity especially with this report, because putting together this dynamic team, like you said, the Three Musketeers, to really talk about AI has been a real pleasure so far. 

Yolanda: Thank you so much for that. I just want to pause and say that's an amazing story, Cindy. Cindy and I run together. I didn't even know this. The transition, the spark of your imagination while you were in an administrative role, the dedication that inspired in you to become an expert in something that captured your imagination over the period of years while you were working full time is really inspiring. 

It inspires me to pay attention to taking advantage of those sparks and having the discipline to see it through, and how it's led you to a place of having empathy for not just ignorance, but how people are willing to understand, and that the language itself is so difficult and far removed from what's actually happening, but it doesn't have to stay that way. It sounds like the lesson you learned from your mentor is it just takes somebody, like a team like this, being willing to break things down into practical, understandable chunks. 

it just takes somebody, like a team like this, being willing to break things down into practical, understandable chunks.  

Cindy: Yes. I think people are hungry for the information. They're hungry to be a part of the conversation.  

But sometimes that learning curve is steep. You go on and you Google AI, and it's like generative models. 

What does that mean? As a general person, I'm already like, "I don't wanna read this article." It can just be challenging, the language they use, the concepts. For all of you wondering, a generative model is just a model that can create human-like content. 

That's all that it means, right? But breaking that down into digestible stuff that my mom can understand, a 65-year-old woman who can barely use her smartphone, now she has to be able to understand what a generative model is, what model weights are. 

These things could maybe be on the ballot- ... later this year or next year. And it's- Voting on topics like this ... it's certainly  

Yolanda: in bills that are coming before- yes ... the State Assembly and the Senate and I... proprietary versus open source weighted models are... I think I've read that in somebody's proposed legislation. 

So as something that's defining some kind of outcome for industry or individuals, those definitions are incredibly important. So I think that's an amazing lens and path that got you to be a co-author on this. 

Cindy: And then we brought Caspar in. 

 connected via our state representative right after that, and they said, The sponsor approached me and said, "Can you write a report?" Brian was just coming back from paternity leave. The day he got back, I called him and I said- "I've got  

Brian: this thing. I've got this idea" ... "I have a crazy,  

Cindy: Thing I want to talk to you about." 

And he was like, "I'm all in." And then we were connected because they wanted a economic view from a California point of view. And I was like, "I don't know anything about economics. Does anyone at the lab?" And they said, "Yes, Caspar."  

Yolanda: So tell us more. Yes.  

Caspar: And my joke that Cindy and Brian and I have heard many times is that, as the only economist that I know of at the lab, I'm a scarce resource that gets handed out to different projects to lend my economic lens. 

But I actually found out recently that we have a colleague, Ben Baney, who's also an economist, but he leads the space program at the lab. So now my new line is that Ben Baney is the economist of space, and I'm the lab's economist of Earth. 

Multiple Speakers: We've  

Caspar: divided the- We divided things like that. 

Yolanda: That's hilarious. So tell us, Caspar, Earth economist, tell us a little bit about how you came to this project. In some ways it seems like a more natural fit in the sense that this is about policy and you have an economist background. I'm also a little bit curious what brought you to a national lab in California as an economist, because that seems not obvious. 

Caspar: Yeah. And my mentor at the lab, Roger Aines, in a generous term, calls me a unicorn for this unusual skill set. It has its challenges as well. To go to the very beginning, I've always been very interested in looking at technologies and problems across disciplines. 

My first degree was in philosophy, politics, and economics. I couldn't decide what to study, and I managed to find a degree where you can learn all three. I had a really strong interest in looking at problems from these multiple angles, and then I did a PhD looking at carbon dioxide management technologies and energy, and was still applying this lens, applying the economics, applying the political thinking to these technologies. Then I unexpectedly moved out to the US. I was doing my PhD in England very happily, and my supervisor called me in one Monday morning, I still remember the meeting, and she said, "I've got a job at UC Davis." 

I wasn't compelled to join her. I could have stayed, but I thought I would forever regret this opportunity, especially if someone else was doing all the paperwork for me for the visa to come out to the US. I've got to give this a shot and try. I've been here, this is now my sixth year, so I really am no longer here by any sort of compulsion. 

And the story behind the lab is I was at a conference and met my now mentors at the lab presenting some really cool systems analysis that the lab does, looking at these technologies from these multiple different angles. A really big team as well. Often when you're doing a PhD, you can feel like you're a bit on your own and you haven't got that group where there are people bringing other skills and other expertise. I saw that in this presentation that my now colleagues gave at the lab, and I thought that just looks excellent. 

I want to join. And I also really liked how the lab works more closely with policymakers, and because I had that background and an interest in politics, I liked the idea of working more closely- with policy design and with policymakers. So it's been really good. It's been excellent for... 

The lab is very good for allowing you to move around different divisions and- ... working with different scientists. And I have an interest in looking at the social acceptance of technology. And I've done some work on that looking at a number of different technologies at the lab, as well as the economics angle as well. 

So it's been great for my brain that likes to be diversified and sitting across different ideas.  

Yolanda: So you really came to this, I found myself listening to you talking about what you did in your education and coming at, for example, carbon technology from an economics angle. 

And I'm going to ask you to do something that I did not expect, but my background is actually as a scientist. We run an organization in economic development, but I'm wondering if you could explain what an economist does. 

Multiple Speakers: does.  

Caspar: What does an economist do? There are a number of different ways of defining what economics is. 

A classic one is it's understanding how we decide to allocate scarce resources. Other times it's about trying to understand how we measure the impacts of different decisions and actions that we take collectively and as societies. In relation to AI in 8, I think the way to define it is getting back to what Brian said. It's about trying to bring some realism to a topic, trying to ground it in a framework. 

And in the case of AI in 8, we're thinking if we just ask what are the economic impacts of AI, it's too broad a topic to think about how it impacts the economy and the workforce and things like productivity. So we took this framework of looking at the California economy. 

What does that look like at a high level? What are the sectors that are involved? What are the jobs? Where are they? And what proportion of California's GDP do they account for? And then when we've got that lay of the land at a high level, where is the evidence for how AI is changing each of those sectors, whether it's manufacturing or agriculture or education? We can then zoom in on each of those regions. AI has a different impact on each of those different sectors because they are different economic sectors, and there are different skills required and different technical means of doing those jobs. 

So that was how we tried to apply this broad term economics to think- ... about AI in 8.  

Yolanda: That is very helpful. And Brian, it sounds like you wanted to jump in.  

Brian: First of all, that's the first I heard Caspar's full entry point into this. It's hard to believe what we're able to achieve in what little time, but we haven't been working together for even a year. 

Cindy and I knew each other for four or five years now, but it's the right perspective that came in at the right time in this conversational piece. I thought that the scoping that we got from our sponsor was very helpful because if you were to condense a topic as big as AI, which has been around since the '50s, how do you paint it out clearly and meaningfully in eight pages? 

And it was through the Californian perspective.  

And I think that the readership is certainly valuable beyond just that audience-  

...  

Brian: Because of just how impactful AI is on a global standpoint.  

Yolanda: And  

Brian: as we say in the report, California's the hub of this.  

Yolanda: Yes.  

Brian: Absolutely. So the technology's emanating from here, the expertise is, the investment in this area, and it's really cool to be able to have the conversation with the leadership in that environment. 

I think when we were joking in the early days of before this was in the proposal stage, it's actually, with California being the fourth largest economy and all of these other factors I just mentioned, it is less valuable for us to write a similar report for the country of Switzerland than it would be Ca- California's the hub. 

This is it.  

And to have a willing set of readers who then also have the onus to create and craft policy, that's really incredible. And what is really unique in our position is that we legally cannot advocate for policy. We are not burdened by ad- advocacy.  

We are burdened by telling the te- technical truth of the matter. 

And as scientists, that's a very comfortable spot for us, but from our experience interacting with the legislature, they're not used to that. They're always used to a different angle. And those angles should exist. I think everyone needs to represent their economic interests and how it impacts business. 

Businesses are very part of our Californian fabric, of course. And so we just have a unique seat at the table in that regard.  

Yolanda: And it sounds like in that respect, what you're describing is that you are a fair broker of information. Because you don't have a particular policy perspective. 

Of, obviously you have some perspective, and I think Caspar just spoke a little bit to it. You outlined a framework of how you are taking what could be a very big, hard to digest problem and reduce it to an understandable framework that would explain some of the critical components of the technology and the impacts in a way that anyone could understand, whether they had an AI background or not. 

And this is something that I think is particularly useful because you have a perspective. It's an economic framework perspective, it's a technical perspective, but it's outlined and very clear, and it's not anchored to a particular policy outcome. 

Brian: Exactly.  

Yolanda: Yes.  

Brian: Yeah. And that, that was a well-desired piece in the conversation that we met. 

Cindy: Yes. 

Brian: And we- And- yes. Go ahead ... I would say  

Cindy: Also, everyone has very valuable input given their life experiences, what they do for a living, all of these different perspectives, right? At least for how we view this report, your perspective matters in the conversation about AI, which is still unfolding today. 

What you care about matters. How you use the tools matters. But what really is important right now, and we talk about it in the report, is get to know the technology. Do not be afraid of it. Go to chatgpt.com, go to claude.ai, go to any tool that you want. I'm not advocating for one or another, but figure out how these tools are good, how they're bad. A lot of people that we talk to aren't even that far along in the conversation. They've never used these tools. 

I think there's a lot of fear around these tools. In the headlines you see these claims that over-reliance on tools decreases critical thinking. I think people are really scared about the technology, and for us it's, I've used AI tools a lot, so check them out and see if they're exactly where you want to see them for what you're thinking. But giving them the scientific understanding, the technical jargon they need to know, and then inviting them, empowering them, your perspective is valuable in this conversation today. 

Brian: It's not a policy standpoint, but I think the thing we are advocating for is literacy. By the time you're done with the eighth page, we'll have accomplished that objective. And what we did is little fun things in the report like AI summaries and AI graphics presented. 

presented.  

Multiple Speakers: And  

Brian: We wanted to basically show you, on some level, this is a product of AI. Granted, we couldn't have AI write the report. 

Not if we wanted to. We're hoping that people can feel like they're using AI and understanding it and its impact on them, and I think we achieved that. I think that's the feedback we're getting so far. 

I think that's the feedback we're getting so far.  

Multiple Speakers: Yes.  

Caspar: It augmented our labor, to use an economic term. There's some debate. Will AI replace jobs, or will it help us enhance our ability to complete tasks? And here we found that, that ground where the AI was enhancing our capabilities. 

Brian: And what's so difficult to write about in this report is the pace of technology. So how do we write something that doesn't become outdated? 

Yolanda: Instantly.  

Brian: In some sense, this report came out in January. We were smart to put the number 2026 on there. A huge topic on AI agents became very publicly discussed, I would say, two months later. 

And while the report doesn't directly touch upon those, everything still applies in that new domain, and there's just a still requirement to continue this conversation. It doesn't stop at eight pages is what I'm trying to say.  

Yolanda: Yes, and it is interesting, too, because I I actually really appreciated the... 

There was a table, I think, pretty early on, table one, that sees recent and future applications, potential benefits, like you talked about, encouraging people to use AI, and potential risks or concerns. Looking at these topic areas, they're pretty broad, right? 

So it does seem like there's room for AI agents, for example, in all of these categories. One thing I realized, we've gotten pretty far down the road on this, but just so everyone knows what we're talking about, how do you define AI? 

Great question.  

Brian: We actually spent a solid week on that first question. 

Multiple Speakers: Yeah. Because we  

Brian: We wrote something, and then other people said, "That's not complete," and other people said, "But you left out this detail." I'm almost just tempted to read what we wrote. 

I was going to do that. Here we go. You know what? Caspar has the best accent here. Okay, Caspar, please read the opening. Give us your narrator voice. 

Multiple Speakers: voice. This is what a British voice  

Caspar: is for. So yes, right at the top, we say, "Artificial intelligence-" describes computational tools that process data to identify patterns, solve specific problems, and generate outputs that resemble human-created work, creating the impression of reasoning, perception, or language understanding. 

Yolanda: Ooh, I like it.  

Brian: I couldn't have read it as melodiously as Caspar. But I'll tell you some really important things to point out with that. We belabored a lot on every word of this report, but that was our opening foray, right? Resembling human-created work was really important to us. 

That kind of- It's  

Cindy: generative models. Exactly. Yes.  

Brian: Good hark back. And perception we wanted to include there, pattern recognition, we had to mention data as well. But what starts to emerge when you read that first paragraph actually bleeds into the second page of the four really innovative capabilities today that are leading to the discussion of even why we're having AI, which is large data sets, powerful computers, capable models, and then mechanisms to deploy them to a large group of people. 

And if you hear what I just said, data, compute, models, and deployment, you can hear that in the first opening sentence as well. And we just need these kinds of things to hinge our readers on. Okay, what are the important things? Here's four. What does it mean? Here's the first opening sentence. 

How does it matter to me? That's table one. And these are all of our perspectives in table one, right? There's the row-by-row economic breakdown from Caspar. There's what you get, what's future about it. That's me, optimistic. And then, oh, there are risks or concerns, and that's Cindy. 

And we wanted to have plain-language descriptions about why things mattered, and this started out too technical, and we needed to condense it to 10 words or less per entry. I think the things that stood out to us is the first draft we had looked good and we had to move on, and when we came back to it, some of the things that didn't come clear is, what's the particular example that I said about the driving example? Let me find it. It's to change a stop sign to be a speed sign. 

And that's Cindy's complete technical domain. It was, how does that matter to me? If I'm in a self-driving vehicle and there's some sort of malicious attack, what does that mean to you, the passenger of that car? If a stop sign is made into a speed limit according to the perception of that algorithm, what are the consequences of that? 

And what that immediately allows the reader to think about is these are not perfect God-like models that are true in all instances. They have limitations.  

And we as scientists must be aware of those limitations before we can deploy them in high-consequence application areas. 

Manufacturing, for instance, I don't want to waste all of this material. I don't want to make parts that don't perform. Even if AI helped me get faster to the wrong answer, is that valuable? Absolutely not. 

And we looked in table two at the different mechanisms for how that comes to be, different biases that could be baked into a model. 

Like for instance, data that isn't representative of a full group of people or population of scientific data. I talked about malicious attacks earlier, and everyone knows about AI hallucinations also.  

That's probably the most discussed failure mode, but there are many others that matter, and just because those exist doesn't mean we can't use AI. It's just that we need to understand what the constraints are when we apply them to areas that are important to us. 

It's just we need to understand what the constraints are when we apply them to areas that are important to us.  

Yolanda: And coming back to your role as skeptic understanding where our skepticism is deserved, that under- understanding the limits of the model is a really reasonable question for any user, not even a tech- not just a technical user to say if AI was used to predict this drug- how was it tested," right? W- And we have a process for that. The FDA's like you can come to a discovery any way. You can file your new drug as a discovery, but you're still going to have to test it for efficacy, right? And so there are some things where you have, you already do have policies in place that- are going to say "Yeah, if this went off the rails, it's going to stop before anyone gets harmed." It's just going to be a question of the company that invested in that, maybe they spent too much money on the wrong thing, right? So it's, there's an economic impact to the company, but no harm to the end user. 

And it sounds to me like what you've done is tease out like there, the, here are the places where it's reasonable to be skeptical and ask what is the limitation of the model? Or if you can't suss that out, how will you def- how will you find the limitation of the model- ... before harm is caused? 

Absolutely. Yes.  

Brian: yes. Or design its application space that is aware of those constraints- ... and still effectively use the tool. Absolutely right. Absolutely right. 

Yolanda: Where do you bring in the human, for example? Like up to this point, you can let the model go, and this is where you check in- with a human with enough information or knowledge to say, "Go, no go," right? 

Brian: And I think what we're talking about here- ... is understanding the constraints immediately helps you balance the risk reward  

In, in what dimension? It's not... Any AI model's not always good or not always bad, it just, it has a well-defined application path. 

And of course researchers and startups and every- are trying to broaden the usefulness of these, and they, and as they do, Californians benefit.  

Yolanda: Yeah. Absolutely. I am really... There are so many questions. This is such an exciting project. I'm very cur- you spoke a little bit of- about timeline and things like that, and I'm I'm a little bit interested in the project itself. 

I'm very curious. You spoke a little bit about timeline and things like that, and I'm a little bit interested in the project itself. Starting with you, Cindy, it sounds like it came from a policymaker pull, and then what happened? 

Cindy: Yes. So I, as I mentioned, gave a presentation at Open Source Day gosh, was that last year?  

Brian: It well, it was during my paternity leave. 

I, because I remember- That had  

Cindy: to be last year ... big time  

Brian: FOMO not being able to go. Okay. Yes.  

Cindy: So that was in May of last year. If I think about it, that timeline is incredible, like what we've accomplished. That's three lifetimes ago. But yes, it feels three lifetimes ago. 

So May of last year, I hooked up with Luis Quinonez, who's from the California Foundation for Commerce and Education. We really just started talking, and there was a clear need swirling around. Everyone is talking about AI, but a lot of people don't feel like they can enter the conversation, or they're hesitant to enter the conversation, or maybe the conversation's not rooted in the technical realities of today. 

As we think about AI, and as I was listening to your guys' conversation, I was thinking in my head there is an important factor here, which is to suss out frontier AI models, which are these large language models typically, or multimodal language models, so they do image generation and stuff like that, as one AI tool. 

AI as a field is a lot of different tools that you use to accomplish a task. It's in the definition, right? It's not just a one-size-fits-all magic AI model that can do all of the things. It can predict molecular structure of next-gen pharmaceuticals. No, it's a very customized AI workflow that we had to develop over years and years to be able to do something like that. 

And so like- Grounding the conversation, focusing on the facts, focusing on the reality, but doing it in a way where you're weighing both the pros, because people- we want people to be excited. 

Grounding the conversation, focusing on the facts, focusing on the reality, but doing it in a way where you're weighing both the pros, because we want people to be excited. I'm excited about AI and the realities, but also these are the things that we need to think about, and framing it in a way that's palatable. It invites people into the conversation. It doesn't lecture people on what it is. 

Brian: conversation here. I come back first day of paternity leave, I've got a 9:00 AM block with Cindy. 

I have spit-up on my shirt, I'm sure. And Cindy's like, "I got this crazy idea." "Okay, tell me more." "Okay, we have to write this report." "Wow, sounds cool. So I'm definitely down." And then she's like, "We need to put a proposal together later this week." I was like, "Oh, okay. All right. Roll up my sleeves. Let's do this." 

Let's do this." And I think in generally with AI and how fast things are moving, when the opportunity strikes, take it.  

Multiple Speakers: Yeah.  

Brian: And that was... That... W- we had to meet the moment and I think you had already been building out the third leg of our team with Caspar because I knew that we could handle the technical stuff. 

Multiple Speakers: Yes.  

Brian: And I, didn't really... I took intro to- I know ... macroeconomics in high school or something like that, right? I was going to say, you say  

Cindy: the word economics, and I start closing my eyes. Sometimes we- Not to be rude, but When we get on  

Caspar: the roadshow, we've given the presentation, and Cindy and Brian do this excellent job with these quiz questions for the audience. 

And then I come in afterwards. I'm like, "Okay, now it's the boring economist who's going to talk about some of the impacts on the labor force." 

Multiple Speakers: Yes. Yes. We're...  

Brian: yes. We- Caspar's the medicine. We put the sugar around it. It goes down nice with the audience.  

Multiple Speakers: And  

Brian: what Caspar referenced is as part of the dissemination campaign for this, we've had several different meetings. 

We met you under these Innovation Tri-Valley events, and with our audience we have this pop-quiz type of format. We don't go through any outline realistically of the paper. We just ask them what's the largest sector of the Californian economy? Can AI be used to generate pictures of cats at a drive-through? 

And other things. Then we start to get into heavier topics: can it be done for malicious attacks, and can it be done for scientific discovery? We really try to seed the conversation, and then, of course, we always let Caspar bring us down to reality. 

Caspar: over. 

On to  

Yolanda: the  

Caspar: boring topic. 

Yolanda: Yes, I don't know how boring that is, though- ... because the jobs impact is something that I think everyone's- Oh, it's generating a lot of interest. Yes. Big topic. Yes, it's top- It's- I have been to... I haven't been to a single AI event where even developers are talking about the impact to developer jobs. 

A l- the impact on entry-level jobs across all industries, I think is a discussion. It's not... It's in the media, but it's also in people's dinner tables and- Yes.  

Brian: And this brings up the point we were talking about earlier. It's just fear of the unknown. And that is the human condition, and we reveal aspects of the technical realities. 

We don't address all unknowns. We also don't know what is three months in the pipeline with AI. 

Multiple Speakers: We  

Brian: We know what we want to achieve. We know what is the opportunity. From a National Lab perspective, there's a global competition in and around AI, and that has implications for nations. 

And so we're talking about impact scales from at your dinner table to your state assemblies 

Brian: To your businesses and their multinational companies and other governments. I cannot think of an example of a technical concept that has had that many levels of discussion, and you cannot go a day without seeing AI in an article, on a podcast, in conversation. 

And that's exciting, and I could appreciate how it's overwhelming as well.  

Yolanda: Yes, and the rapidity of change. Caspar was our featured speaker at an event, AI at Scale. I think even that day some news had broken that completely changed the discussion. 

I think you mentioned it the day you guys presented at ITV. There was a new announcement of some technical capability, and it shifted the details. I don't think it actually shifts this framework very much, but it does shift the details, the specific examples, the what-can-AI-do-now conversation. It changes every single day conversation is, it changes every single day.  

Brian: And that's the real hard thing is how... there's such a dynamic landscape associated with AI. How do we put something that doesn't get dated-  

Multiple Speakers: ... In a year? 

Brian: And, this is only five months old, so we'll see how it can withstand- ... the test of time, I'll say. 

Caspar: And it's also the challenge for policymakers. They can't wait to get all of the information and the data in before they make decisions. They have to. So they're making decisions where they've got some information, but not as much as they'd like, and that's a really big challenge for them. 

You know- Yeah ...  

Brian: Sometimes their staffers and also elected officials are like, "What policy do you think would help with this?" And we say, "Actually, hands off. That's not our domain." 

Cindy: The conversations that we've had with politicians have ranged from- Oh, yes 

Should we stop AI development now? More towards, just give us a pause to figure out what we're doing, to this is going to change the way that democracy works if we can have an AI agent that sends, on your behalf, emails to your congressperson. How do we filter out the signal-to-noise ratio? 

Does democracy even work? And we're like, "Whoa, we're scientists." These aren't issues that I've thought about or want to consider, but as we start talking to different people, especially politicians and legislative staffers, I think, again, these are the most qualified people to have these conversations. 

All you need to know is how does AI fit into the reality of the world right now. And I don't know the answer to that question. Everybody has their own perspective, but making sure that everyone's educated on the technology and can come to the table and talk about that is really  

Brian: important. 

Another thing that reminded me with that topic is how we had to train ourselves to communicate through this document. Because it is nonpartisan. It is apolitical. It has to be, from a legal standpoint, as national lab employees we cannot advocate for particular policy. 

Another partner in this effort is our Office of Governmental and External Affairs, so our policy shop. They were strong partners and scrutinized every word on our behalf, and we were very grateful for that because just the word choices can impose, or it can cause a perception to get emitted. 

It can, yes, im-  

Yolanda: Imply maybe a perspective that isn't intended.  

Brian: And one, we don't necessarily want that. We want to stay credible as truth-sayers in this conversation. But also, I think introducing policy perspectives in a document like this would help it become dated, right? 

Yes. It would totally reduce the shelf life.  

So, we really benefited from multiple angles with all the partners that helped with this, and of course, we filter it with our scientific colleagues, people like, "What are you writing? Why does your audience care about this? Everyone knows that." No, not everyone knows that. We know this to be true. 

And being able to wade through everywhere from our PhD colleagues to policy experts to our sponsor, Luis, to people who have engaged more with the legislature than we have, it was a really incredible experience, and it remains to be, and we're looking forward to just continually trumpet technical realities into the mix. 

Yolanda: Yes. I'd be really interested, building on what we've just been discussing, what it's a, I'm curious about what surprised you the most, both as you did the work that went into the primer, and then as you've been briefing legis- the legislature and just making the public aware of it. So the I'm looking for two surprises here. 

I'll start with you, Caspar.  

Caspar: I would say the first surprise for me, taking on the mantle of looking at the economic analysis, was how much more complicated and nuanced it is. When I first picked up this, I was like, "Oh, I guess the economic impacts are just all negative." And I went to a couple of economic conferences and heard from people speak, like the lead economist at LinkedIn, and was getting a much more complex, nuanced picture, and that was very interesting, and made sense looking back that it wouldn't be a simple, straightforward thing to understand. 

And so that was a really fun thing to drill deeper into and understand the nuances and the complexities. And just to give one thing that I heard from this chief economist at LinkedIn, they have really good data on the job advertisements on their platform, and what skills are in demand, and what jobs people are applying for. 

And they were saying that in the last few years, yes, there's been an increase in demand for these AI literacy skills, but there's also been a big increase in demand for skills that have got nothing to do with AI. Those are more human-focused skills. Skills around critical thinking and mediating conflict and creativity. 

And the way that they made sense of that was that if you've got two candidates for a job and one is very good on the technical side, but not as good on some of these human skills, AI can lift the technical skills of a candidate so that there's a more equal footing between them and other candidates who are better technically. 

But it can't lift up social skills. Those are ones which humans have right now much more a stronger comparative advantage. So that was a really interesting thing to see on that end.  

Multiple Speakers: Interesting.  

Caspar: On the rollout of the report, I really like how the legislators and their staff really value in-person conversations and in-person meetings. 

The day we launched the report, w- we had a day just back-to-back meetings with different assembly members, state senators, staffers, up in the governor's office as well, and it was really refreshing to get out of the lab, get out from behind the computer screens, and just meet these people. And we've had several further meetings since then, either in person or over video, but they really appreciate hearing from you. 

There's not that many emails that go back and forth between us and them, which is a really nice surprise and just a really enjoyable part of rolling out the report. 

Yolanda: Oh, that's wonderful. So policymaking is an in-person job  

Brian: Yes. Yes. I think one element of surprise revolves around the word impact. 

I think that this has been an incredibly strange and impactful aspect to my career, and it's been so new, so I'm really viewing it from a recency bias. But people are really eager to have scientists at the table in discussions.  

And there's a broader audience out there for my work. 

And you know that, right? And I think when you look back into scientists in our society, it is very high-profile people that get into those discussions. Richard Feynman and Rob Oppenheimer and all these people. 

And we definitely didn't need that profile to be able to put this out there and have the conversation starters and the connections. 

And I also wanted to piggyback on your statement about this group of people. It's the dedicated staffers and the legislature and the elected officials, of course. I am a scientist, and I'm a very social human being. 

These people are on a whole other level with how they understand the dynamics of conversation and human connection and representing different interests. Just being around that is truly hard to appreciate until you see these folks in action. That was a surprise as well, because I'm so steeped in science as part of my everyday life. 

And, I am a technical expert. They are like human being experts on how to navigate those social dynamics. It's incredible. Yeah.  

Caspar: It's a great environment, isn't it, being in Sacramento? Oh, yes, it's electric. Yes. And our colleague, Andrew, who's our lab representative there-  

Who just hooves up information and sends it back to us and r- represents us he always chaperones us around. And I would say within three to five blocks of the Capitol, everyone on the street seems to know him. He's just high-fiving people and, 

Multiple Speakers: Yes. Smiles. Yes.  

Caspar: He cannot walk  

Brian: 10 feet without interfacing with someone, or three- Which is am-  

Cindy: amazing- 

by the way. The stark contrast. I'll hammer home some of the point that you were just making, is my day-to-day life on a normal day-to-day basis is I do technical work on a computer. I take a lot of meetings On the computer. I go home and check, my kids' homework on the computer, right? 

My life is literally a simulation at this point, right? I am talking to people who are using AI to interpret my emails, and then they're sending me emails, and I'm using AI to interpret their emails, and that's the reality of, I don't want to make it sound bleak, but in my opinion a lot of our lives are just so technologically driven. 

That when we go to a room like this, and I see Andrew in his domain, there is so much missing from the human connection point, right? It's like even going through this experience, something that resonated a lot with me is I should have more in-person meetings and get to know people and try to meet them in person because it's so much more meaningful to sit across from you right now. We're locking eyes. We're connecting on another level. I think that was a huge surprise for me, just how much they are so human-connected in that world, and bringing some of that home. 

Personally, as a researcher and a manager and a director and all of this stuff, it is so much on the computer, but human connection is so meaningful and missing, and these conversations have probably been the highlight of my year last year. 

It's just talking to these people, connecting with them in these quiet rooms, and just listening and talking. That was the coolest day ever. 

Yolanda: Yes. So I'm building on that. I'm curious, having had that, and y- you alluded to it a little bit, Cindy, where you said, "I should be having more in-person meetings," what are you doing differently as a result of this report? 

And it doesn't have to be something you learned here. It could be- Being on a podcast is one of the  

Brian: things I learned. What  

Cindy: Yes, I haven't done that before. I would say, going back to my skeptic hat, there is a lot of promise for AI technologies right now. 

Being a group leader for a long time of people leveraging and staying on the cutting edge of this technology has shown me that they are burning out in a way that's really interesting. We are at peak efficiency, peak performance. We can now use these tools to help us do different things that we've never been able to do before on shorter timeframes. 

But again, like- We're all burning out because we're just talking to this terminal all day- ... of talking to this AI agent, and we're missing human connection on a day-to-day basis. And this is documented. You can Google the term AI burnout. It's happening across organizations. 

And I don't... And I was reading... I'll have to send you the link. Yeah. But I was reading a position piece on this, and I thought their summary was very good. It's not happening because organizations are, doing this intentionally. I just think that people don't realize how difficult it can be sometimes to set up these AI workflows to meet what you're intending. 

There's a lot of intention- ... with the things that we do. Writing email's really good out of the box. But doing something like imitating a process.  

Multiple Speakers: And I'll give  

Cindy: you an example. A interview question that I heard recently is if you were to write an AI agentic workflow to replace you as a human on the planet- what would that look like? If you think about that question, it's really hard to answer 'cause what do I do? Yeah, I respond to emails, but I also hug my kids every night. And, like, how would I set this up? And how could I show up? I think we're putting a lot of emphasis on AI staff and people implementing AI. 

Like- ... do this thing. Just... The AI will do it. It's fine. But the effort needed to actually stand that thing up is really a lot that you're asking of your staff to do. And again, I think people are asking on shorter timeframes. And so I think people are working really hard. They're tr- they're being as efficient as possible, but I think the promise of AI is coming down a little bit 'cause you realize oh, it's going to take a lot of time to get something up and running. 

Yolanda: Yes. It's interesting that you say that. So one of the previous guests on the podcast Chin Jiang, who has a Monterey AI that's been acquired twice now, they're in Dublin actually said, she's an AI creator. She's a technical founder. And she said, "Anything I feel like I have to re- repeat I want-" I want AI to do. 

I want a c- I want a computer to do anything that I'm doing repetitively. And as someone who's deep in development, I think for her, this is actually relatively easy. But working with a- agentic tools, and even not agentic tools, just using, out of the box co-pilot to write something, it takes a long time to do something simple take my meeting notes and put it into my tasks- Oh 

into Asana. Yes. And it's that's not actually trivial, and by the time you've done that speaking as someone who just did that you're like, "Could I have just kept taking the extra like two minutes- Yes ... to copy- For real ... and paste? Yes. I feel like maybe. And  

Cindy: there's a conversation that happens all the time- 

because there's also like this nuance of like how you ask your agent or your model- Model. ... to do stuff, and we talk about this all the time. So Brian's super nice. He's like- I'm like very  

Brian: polite ... " 

Cindy: Please can you help me rewrite these meeting notes to be in this particular format?" And I'm like, "Bro," like- Do it. 

"AI agent, do this thing or I will not give you cookies at the end of the day," and it's just Yeah, if AI actually became sentient, I'm going straight to the severed floor- I- ... and you're like- Lauren's hedging a  

Brian: hedge on the revolution. Yes. I just wanna be polite to our robot overlords- Yes. 

if it comes down to it.  

Yolanda: I can track my... AI was helping me build this workflow also, like just helping me just figure out how to use the interface, and I can tell you like the first few minutes, like the first half hour was really polite. By the time I got to two, I'm just like, "This seems like it should be simpler." 

Why is this not done yet? Yes.  

Brian: My new thing is I use certain AI tools against each other, say, "Write me a really good prompt," and I'll take that really good prompt and put it into the other AI model, and okay, this is good. They don't know about each other for now. That's good. 

That's good. But it's just a way of extracting value as best as you can because it's just a lot. What we're talking about really is like even complaining about the effort of having to write a prompt, and I wrote a prompt to write a prompt. And so yeah, you can easily get into these, efficiency loops. But I also feel like  

Cindy: it's letting us do things at a timescale that we've never been able- Absolutely to do it before. I was dealing with some medical issue recently and got really deep into a hole, and I used an AI tool to basically look up if, this is going to be like Way in the weeds, but whatever. 

I was wondering if I'm, like, MRI safe because I have a metal plate in my jaw. And this is typically something you must call your doctor and request information. The surgery happened 12 years ago, so good luck getting medical records. But I went onto the website for my doctor, and they had actually documented the type of plate that they put into my jaw. 

And so I asked the tool "Is this MRI safe?" And it's "Probably, but if you wanted the actual answer to that question, here's the email address that you'd have to email." And I was like, emailing them. They said, "Yep, you're MRI safe." And I was like, that's probably a process that could've taken me months- waiting on end, requesting records doing this, doing that. And now I have an answer in less than two weeks' time, and I'm good to go get an mRI. And it was just like, yeah, I was thinking about even, in my parents' era. That probably would've been a never in your lifetime process to be able to get all that information. 

It could've easily- You probably just would've assumed things.  

Yolanda: you would've just... Someone would've just told you what the safest assumption to make is, and 

Cindy: You would've acted from there. Yeah. So it's, yeah, I think it's speeding up timelines, slowing down some timelines. The nuance is very difficult. 

Yeah. As Caspar said- Mm-hmm ... learning about how this is affecting things, I think people are like, "AI's here to take our jobs today." And the reality is probably not. We're not seeing that. I don't know. I'm looking at you, hoping that you'll- ... jump in here. Do it. Chime in with the data. Do it. The rule list here. 

Paging our local economist, our resident economist. Yeah, and there was that interesting Anthropic report that came out that you showed the radar plot on, if you wanted to talk through that. Yeah. It's just the promise versus the reality. And that was  

Caspar: an example of something coming out a couple of days before our next roadshow event, and- making a quick change to the slideshow presentation. But yeah, as we've said earlier on there's not that much data yet on how AI might be impacting the workforce. And there are a lot of other things that are impacting the workforce, so it's hard to tease out what's actually AI and what's other things. 

And that's- ... that gets talked about quite a lot when, say, there are tech layoffs going on right now. Are people just being lazy and blaming AI for some other problem that's affected their company? But the, Anthropic published some analysis where they had taken a assessment of all of the human tasks in all of these different sectors, and which tasks seemed most ready to be done by an AI tool. 

So that's the potential. And in some jobs, there are more tasks that can be done by AI tools. Other jobs- there are fewer tasks. Like a bike mechanic, there aren't many tasks could be done by an AI, but maybe a legal job, a lot of the tasks can be. And then they looked at the use data of their AI tool Claude. 

And they were able to understand what uses were aligning with those job tasks.  

And that then allowed them to overlay onto this radar plot of what the potential was, what the actual data was showing so far. And so it did show some job sectors where there was a relatively high level of AI use for those tasks, and other ones where it was quite low. 

But across all of the tasks, it was still relatively limited, which reflects that there was just a challenge and there's a lot of friction in integrating AI tools into a business workflow. It's not a simple task.  

Yolanda: No. Yes. I, and I think what I've been... What I think I'm hearing is it's probably, so Anthropic mapped what they're actually seeing versus what you would ex- expect to see based on the capacity of AI to- Yeah, the potential. 

The potential of AI to potentially replace certain tasks in different fields. And you are closed to something I think is really important. It's hard to replace business workflows. And I think the thing is there are a lot of places where you don't have a documented workflow. There is a workflow,  

Multiple Speakers: yes 

Yolanda: but it's not necessarily a documented workflow, which I think is really the first thing you have to do, and then make a decision is the best use of AI to replicate that workflow, or because you have AI, do you have a new workflow that you have to-  

Multiple Speakers: Exactly. And  

Yolanda: that, I think, that's actually a pretty big burden on a business to pause and do that if nothing's broken, right? 

Cindy: I think it's also a pivotal moment, too, and I'll say that just from my experience doing AI for operations, is, what I like to say is you can't throw AI at a bad process and expect that it's going to do it better. It's just going to make the mess move faster. We're actually using these tools and being like, "Why are we doing this thing?" 

Throwing AI at it can speed it up so that it used to take 10 minutes and now it takes a second, but should we even do this thing anymore? And now you can come to the table and say, "The ROI for us doing the, return on investment for us doing this thing is going to be" couple hours spent, but like really what takes the longest time for a lot of different things is the process. 

It's like Brian has to review something, I have to review something. Let's say you have to review the same thing. Why are all three of us reviewing the same thing? Can we talk about just like the process? And while AI isn't inherently in that yet I think it's allowing us to go, "Why do we do things the way that we do it?"  

Brian: So to add on to that in the to AI or not to AI conversation, should we even been doing this? And then again, back to thinking about the limitations of these tools, I like to say that AI is just as dumb as humans, but faster.  

Multiple Speakers: So if you have  

Brian: that framing, just because we can send an army of people at these particular tasks, should we? 

Brian: I think that this is true in business ops. This is absolutely true in my manufacturing domain and other scientific areas that my colleagues engage in. There is a series of low-hanging fruits, and you want to go after the lowest, juiciest hanging fruit. And I think that in and of itself is a skill that's very human-centric- Right 

now. Understanding what, what matters to this organization, and what is the next best thing to do? That's like a, almost like a strategic taste, if you will that I, just in my personal worldview. is definitely a human-led process. Or  

Cindy: just tell us what do you hate doing in your job? 

I hate calendaring, right? And maybe it's with, it's not in my job scope to have an executive assistant. How can I get rid of doing that thing by using tools? And is it worth my time to even invest in that? I'm calendaring all the time these days, so probably- yes- I should invest in like figuring out how to get some smart tool to do that for me, or some collection of tools to do that for me, but it's going to take some time to like actually get a system that works that I don't hate that takes all of my quirks and wants into effect and- And that you trust 

and that I trust- That you trust to do a good job. And  

Yolanda: It's funny because the thing that would, it's, it's- It doesn't make you more ready to trust a algorithm than it would take you to trust a person, right? That's what I'm saying.  

Cindy: I probably don't even trust a human- ... to get in my calendar or anything. 

Yeah. But at least with a person you can have  

Yolanda: you could have the real-time feedback. Yeah. "Oh, yeah, not this," or- Yeah ... "Why don't you sit with me for a..." And it's... And essentially you're doing the same thing with A- with AI, but faster, right? You're like, "Okay here's all the data of all the meetings I schedule, so you get my vibe, right?" 

And then it gives you something, and you're like, "Wrong!" That sounds right in theory, and then you just  

Cindy: spit that out, and I'm like, absolutely not that. So what did we do wrong here? And I  

Yolanda: think the only d- the difference is if you did that to a person routinely, they would hate working for you, whereas you can do it to AI ad nauseam, and it's still gonna do it. 

Yes. Feels the same.  

Cindy: So I don't know, making us more short and rude to our AI agents, probably not the best outlook for humans, but  

Yolanda: I wanna touch on one thing before we wrap up, which is how bring things back to the Tri-Valley. And so one thing I do wanna ask is how does being based in or connected to the Tri-Valley through Lawrence Livermore National Lab influence your perspective on innovation when this, this report, or does it? Do you s- do you feel like it, it has an impact and if so, how?  

Brian: Oh. We're passing the... All right, Caspar, drive. Caspar. Yeah. I'll, I'll kick this one off. Because-  

Caspar: Because Yolanda and I were at this event, Scaling AI for Startup Tri-Valley, a few months ago. Yeah. And preparing to speak at that event to the business community of this region- and engaging with them in, in, in the discussion that we had did make me think that there are, there's an important role for a startup community that is in the shadow of a larger one- ... Silicon Valley, but it is... But by being smaller, you have this ability to be nimble and to be flexible and to be strategic in- particular areas you wanna have a comparative advantage in.  

And in the scaling of AI, there's a lot of opportunities for that kind of startup community to support the innovation. And my particular interest at that event was how the Tri-Valley community can leverage its expertise in energy and sustainable technologies to think about scaling AI in, in a sustainable way. 

And there's a lot of discussion about the resource Costs of data centers and the energy, the water use, the materials And we had a good discussion at that event on where there are some ways to innovate to find new ways to cool data centers that aren't using water And to find new materials. 

There are some startups that are trialing data centers made from things like mass timber- ... replacing some of the concrete and the steel. Trying to find ways to use the heat that's generated by data centers. So I think there's a really interesting role for some of these startups to engage with the hyperscalers and those that are building the data centers, and to find ways to make these data centers more sustainable, less resource-intensive, which also fits into the angle of getting community support and public acceptance for the infrastructure as well. 

Yolanda: Yeah. Tha- thank you for that, 'cause it is a huge opportunity because we are y- many of the data centers are actually not h- I don't know, I don't think any data centers are here in the Tri-Valley, but they're in the Bay Area, and that gives us a unique ability for our founders and innovators to solve some of these problems in partnership and use some of the things that we have in the Tri-Valley to pilot these solutions like land. 

I was just talking to someone from Berkeley that was saying, we have UC Berkeley and that's spawning a lot of companies, but we don't have a lot of places for them to scale, and I envy that the Tri-Valley has space, right? And in the Bay Area, that's, that is a real limitation. 

Caspar: Brandon at Livermore City Council talks about this quite a lot. Yeah. The space, the resources that we have in Livermore for this.  

Yolanda: Yeah.  

Cindy: Yeah. I'd say the Tri-Valley is so cool. I'll start with that. I live in Livermore. I love living here. There's so much community here, but what's swirling around my mind is two things. 

So one, we're probably most forward-leaning as a lab in the national security enterprise. And I would say that's partly because we're so close to Silicon Valley- ... that it makes all of us have this almost entrepreneurial spirit where we're out there, trying to push our ideas forward. 

We're called the Big Ideas Lab for a reason, right? We're bringing people in who are innovating-  

... 

Cindy: Amazing, smart people that are innovating, at least at the lab. But in the Tri-Valley area, we have both this super rich history, but we're almost new at the same time. And I think bringing together those two worlds has been really awesome to see in the, both the local businesses here and the innovation happening here. 

But also understanding where our roots have come from. The lab is 70 years old. We've been growing grapes in this region for a long time. We have a really rich vintner culture here. But we also are, like, bringing in different new ideas all the time. The Cheese Parlor is one of my favorite places to go, right? 

I never thought I would like a cheese shop until you walk into The Cheese Parlor. It's like- ... God, that place is so cool. But it's also an innovation, if you think about it in some regards. It's like we've had wine here forever.  

Why did we never have cheese here- ... at the same time, right? 

I don't know. I just really like this region. I think we're well-poised because we have this rich history in our roots, but we're not afraid to lean forward and say "This is what's best for us." We need a cheese parlor. We should have a cheese parlor in every single town from here to, Georgia. 

That's my personal stance. But poor Brandon probably wouldn't wanna extend himself too far.  

Brian: Yeah, just for some perspective, Cindy brings up The Cheese Parlor once a week at the minimum. This is not a surprise to me, what's happening in this context whatsoever. So how how to compete with those answers, I'm not sure I can. 

But this report is a continuation of partnering mechanisms that the lab has set up in the Tri-Valley which is a huge I think a significant fraction of our employees live in that area, contribute to that area, send their kids to school in that area, buy wine from that area. And th- this report holds up to this particular group. 

We- we're touching upon all of these main economic sectors- ... in this Tri-Valley region, and we have such an important role to play in this global conversation. Proximity to the Silicon Valley is absolutely an important reason why, but also our perspective is valid, and we really have people in this area that have been driving technology for the country- And this is another part of that effort, that historical role that this, that's been played out in this region. 

Yolanda: I love that answer. I love that connection back to the, the people here m- make the product, right? And and then the fact that this is a microcosm of what you're seeing in California. I'm really curious how... So bringing it back to, you've talked a lot about Sacramento, but we have, the cities here and then Tri-Valley. 

There are a lot of cities actually who are writing right now some kind of AI policy document. It is it's come up recently even in the city of Livermore. I'm a planning commissioner. It's come up recently in the context of what belongs in the general plan and doesn't regarding AI, and I'm really curious, how can smaller regions, not the state, how can cities, counties use use AI in eight pages to in- inform their policy? 

I got this one ' 

Brian: cause we opened up with it. It's just at a minimum, increase your literacy on this very important topic.  

Multiple Speakers: Got it.  

Yeah. 

Brian: I- in some sense, maybe the next step is to harvest benefits from this technology for your area. And then maybe the our local skeptic will say, "And also guard yourself against the risks." 

And do so in a way that impacts your daily life and your economic wellbeing. So I think that's a nice full circle answer. I think the, that perspective applies also outside of California as well.  

Yolanda: Yeah, and I, but just building on what Brian said, so I'm curious, uh, Caspar touched on this for founders. 

So you know, policymakers, local and state, and maybe even nationally could be using this report to inform themselves. How about some of the founders here who are, maybe they're using, maybe they're actually making building AI companies. I just from my own observation at Daybreak Labs, I don't know a single founder working even in hard tech or life science who isn't using AI at some level. 

So I feel like it's safe to say if you are innovating in any meaningful way, you're probably using AI. So I'm curious, you're either making AI or using AI, how do you see the sort of opportunity space for the founders in the region and how they could use this report?  

Cindy: Yeah, I echo what Brian said is, becoming literate on this. 

A user of AI technology, you have a valuable perspective. Nobody's gonna know your target audience better- ... than you know your target audience. Using this technology to improve yourself maybe, understanding its limitations and where it falls short, and really walking that fine line like we talked about on this podcast is you're gonna have to make the business decisions that make sense for you. 

Going back to your question about, local region city councils and planning commissions- ... again, leaning in to learn about the technology, talking to your constituents. What are people worried about?  

And listening to that and really seeing how to address some of that. 

Is it unfounded? Is it a need for more AI literacy in the community? Is it, Yeah, what is it that people are worried about? And then hoping, trying to find areas where you can address that in some way. I would say if you're not using AI today, you should probably start. And this goes, I think we talk about it in the report, too, right? Having access to these tools everywhere, not just in, i- especially in the Tri-Valley region, right? Having broadband access and stuff like that- ... is probably important. If you're thinking that your kids aren't using AI, they absolutely are. 

I have a second grader who's issued a laptop. He's at school here in town, and he's issued a laptop as of first grade. He can get up and running on the internet on that thing. And I'm like, "No, you will not use AI." I'm, like, so adamant. You have to understand the re- you know- ... the risks and stuff of it. 

He's a second grader, though, right? He can't even put his pants on right. But the other day he pulled open his laptop and opened Google, asked a question, typed it in. 

Multiple Speakers: And  

Cindy: he says, "Mommy, the internet says this is the answer." And I'm like, "What are you talking about?" Oh, AI summaries at the top of Google's page. 

My son's interpreting that to be the lay of the land. "The internet has spoken, Mom- ... and it says that it's this." And so I have to have that conversation with him- ... of "You have to think about that stuff, dude. Does that make sense to you?" 

These tools aren't perfect. Maybe someday they will be, but not right now. And again, that's entering his life as an eight-year-old. So we're talking about AI literacy from school-age children all the way up to, my mom's age and older, right? Such a hard thing to dial in on.  

Brian: Yeah. I don't know if this is too far off topic from your question, but we have- Some AI usage at the home. 

I have a five-year-old daughter, and there's two things that she really likes with AI. One is nighttime stories tailored around a particular topic- ... and, get a quick couple paragraph story, and then maybe even generate an image for it. Actually, she likes my off the top of the head stories better, so I'm still doing all right in that department. 

But then another thing is we have a smart remote to control our TV, and she can prompt in there, "Generate AI art of a unicorn with glitter," or something like that. And she actually has a whole little portfolio. We let her run with it. I think that's kind of cool. But then, I don't know, she was just holding the button and asked a question of us or whatever, and all of a sudden the TV starts to answer her question. 

Oh y- and then it... Now she's "Oh my God, I'm interacting with this television." And she starts talking to this TV w- through the remote, obviously AI powered. I'm like, "Whoa, okay," the, for me, this is a boundary now that we're crossing- ... 'cause now she's interfacing with this digitized expert on who knows what she's ra- raising. 

I- in my own household, my own perspective, there's already good and bad things that can happen with AI, and that really surprised me because yeah, they're absolutely using it, and sometimes it's just cool and entertaining, and other times the, the- You have to  

Cindy: decide, yeah- Yeah 

what you're comfortable with, right? Yeah, that's right. So again, even if you're not using it in your workplace for accelerating your tasks, like if you have little kids at home, you have to get to know it- Yeah ... so that you know how to block that stuff as they start coming up. I think  

Brian: the... yes. The ostrich head in the sand strategy is not, I think, the correct play for this m- very momentous technology. 

You have to engage with it. That doesn't mean you have to like it. That doesn't e- mean you have to automate every aspect of your life with AI. I'm not advocating for that. I'm just saying be conscious of it.  

And actually, again, we use AI in the report to generate little summaries of different sections, and  

Cindy: this amazing artwork.  

Brian: Yeah. No doubt. Which, like- ...  

Cindy: don't zoom in too close- ... 'cause it's nightmare fuel to some extent too, looks good- yes ... but zooming in, it's bad. yes, this person has six fingers or something. yes.  

Brian: Yeah. Yeah, this person's face is not a face- It's bad ... when you look close enough. 

But we actually kind of like that detail- ... 'cause we had this discussion of does this even make sense? Is there such an office meeting room that looks upon the Cap- absolutely not. But, it, it- But there should be. It's a conversation starter. Absolutely right. All right. And those are not palm trees, which would be in front of the Capitol. 

Hang on. We digress. So it- We digress. We're way off  

Yolanda: topic. So I think maybe just this kind of leads me to, I think, a great place to close, which is just one thing. What is one thing you wish more people understood about AI right now? And I'll start with you, Caspar.  

Caspar: For me, it is that the future is not determined with AI, and how it impacts us both in society and in the economy is going to be determined by the decisions that we take. 

So we have a lot of agency, and I think that sometimes has been lost in this discussion. Sometimes maybe people think that the horse is out of the stable and it's on this runaway train. I don't think that's the case. I think we collectively as citizens and as all the other roles that we play in society, in the household, in our homes, companies, we have a lot of ability to shape these tools and design this technology in a way that can be really for social good. 

So I think agency.  

Yolanda: Agency. I love it. Not agents, agency. Yeah.  

Brian: That's a really good tagline.  

Multiple Speakers: Human  

Brian: agency. Yeah. It, to be honest, I have a similar answer. You have a voice in this conversation. It as large of a volume there is about AI, your perspectives and how it impacts you matters. 

So you have a voice, and nothing's foregone.  

Yolanda: I like that.  

Cindy: Yes. I feel like you guys said everything. Yes. I just think, yeah, don't feel overwhelmed by the technology. Lean in. There are many things that you can read about online that can help make it digestible. We're happy to engage with anyone who has questions- from a technical standpoint. We got a lot of really great technical staff members who this is such an exciting time and they're happy to weigh in on those things. So feel free to reach out to me, Brian, Caspar, whoever. Absolutely. Love it. Yes ... and we're happy to have those conversations with  

Yolanda: you. 

So I cannot thank you all enough. I think this is a great way to end, which is, first of all, human agency ... is the most important thing, I think, that I heard that people don't understand about AI, that this is a tool that we are actually shaping. And in addition to AI in Eight Pages, in addition to this report, you have Huge resources available to you in Cindy and Brian and Caspar, who, um, I'm happy to put your, whatever contact information you wanna share in the show notes so that people can reach out to you. 

And I cannot thank y- the three of you enough for responding to that initial ask from your partner in the legislature at pulling this team together, writing the report, and now, doing the roadshow about it. So fun. Thank you so very much.  

Brian: Thanks for having us. 

Thank you.