Episode 14 - AI & the Environment from an Economic Perspective

00:00:17:11 - 00:00:43:15
Jie Ren
Hi everyone. Welcome to my podcast, When Tech Meets Ed, this is Professor Jie Ren. Let's continue the discussion on the topic of AI and the environment. Today I have a friend and colleague, Doctor Marc Conte, who is an Econ professor at Fordham University and also an expert in the domain of environmental economics. So, let's talk about numbers. So, Marc, welcome to my podcast.

00:00:43:16 - 00:00:45:01
Marc Conte
Thank you so much for having me.

00:00:45:03 - 00:00:47:08
Jie Ren
Okay. So, could you please introduce yourself?

00:00:47:10 - 00:01:15:18
Marc Conte
Sure. As you said, I'm an environmental economist, so I do work around actually market failures that relate to the environment. And we'll talk about some of the terms later. And I've been at Fordham for about 12 years, in addition to faculty member here, I'm on the New York City panel for Climate Change, which is an advisory board that, like, helps the city think about its plans for dealing— adapting to climate change.

00:01:15:20 - 00:01:21:05
Jie Ren
That’s nice. I know that you have done— you have published many papers along these lines.

00:01:21:06 - 00:01:23:04
Marc Conte
Yes. A few, not many.

00:01:23:09 - 00:01:25:10
Jie Ren
Could you please introduce your research.

00:01:25:15 - 00:01:53:00
Marc Conte
Sure. So, I'm interested in the area of climate change. I'm interested in a couple of different aspects. One is natural disasters. So, thinking about hurricanes, tropical cyclones and how they affect insurance rates, which can maybe be a signal for the changes in risk exposure that come with a changing climate. So that's something I work with a colleague at the University of Miami.

00:01:53:02 - 00:02:17:14
Marc Conte
Along those lines, I'm also interested in kind of climate policy. So, William Nordhaus, who won the Co—, was a corecipient of the 2018 Nobel Prize in Economics. He developed these climate integrated assessment models. And I'm doing some work extending those to think about damages, not just to market activity, but also to aspects of nature that contribute to human well-being.

00:02:17:16 - 00:02:43:23
Jie Ren
Okay. Those are great. And then definitely very relevant to today's topic. Yeah, right. I relate it to AI. So, before I go there, I want to— I want you to kind of summarize like one message that you could deliver to all the students that are interested in this domain, right. The protecting environment and everything, is not only about the impact of technology on environment, just overall.

00:02:44:01 - 00:02:49:05
Jie Ren
Any overall takeaway to the students to set the stage for today's discussion?

00:02:49:06 - 00:03:22:05
Marc Conte
For sure. I think it's a very exciting time to be working in this area. Certainly, we face a lot of challenges now, but also, we have computing power and data that make answering a lot of the pressing questions much easier than they were in the past. And so, yes, we've kind of not done a good job of managing these valuable resources to this point, but I think there's an opportunity with good, thoughtful effort and impactful research to get us back on track.

00:03:22:08 - 00:03:48:12
Jie Ren
Okay, nice. So, let's start with some jargons in economics, right? I'm a rookie in this domain, so correct me if I'm wrong. So, there is jargon that's called externalities, right? It seems to be one of the central concepts in econ. And then could you please explain it as that's a very relevant topic related to this thing?

00:03:48:12 - 00:04:16:02
Marc Conte
Certainly, it's relevant for the environment in general and certainly with AI. So, if we think about the price you pay for a product like for a gallon of milk, we think that the price reflects both demand and supply factors. So, something to do with the cost of production. So, we need the cost of keeping cows for the dairy manager.

00:04:16:06 - 00:04:47:04
Marc Conte
And then also how much people like milk. But oftentimes there are costs to the production of goods and services that are not reflected in the private costs of production. So, when thinking about milk, sticking with that example, if it takes land for the cows to grow on, that can have impacts to society of using that land in certain ways that generate other costs.

00:04:47:06 - 00:05:10:17
Marc Conte
Okay. And so those costs that accrue to society but don't affect producers, those are known as external costs. And there's also external benefits. So recently with the pandemic we learned about the value of immunization. When you get a vaccine, you're doing it for your own personal benefit. But you being vaccinated means the disease is less likely to spread.

00:05:10:17 - 00:05:21:13
Marc Conte
So, there's an external benefit to that decision. So, externalities talk about kind of differences between private and social benefits and costs. So, there can be both sides.

00:05:21:15 - 00:05:32:15
Jie Ren
So great. Could you please give an example of that related to any technologies, not just a necessarily AI. Let's start with a broader concept and narrow down to.

00:05:32:16 - 00:06:01:05
Marc Conte
Sure. So, the classic example in the environment would be thinking about fossil fuel extraction and then the use by consumers. So, when you think about the cost of taking all or gas out of the ground, you're thinking about the pumping costs and the transportation costs, but also extracting those resources can impose impacts to society based on some spills that happen.

00:06:01:07 - 00:06:30:17
Marc Conte
Environmental impacts. There's been some articles recently about testing for deposits in the Gulf of Mexico. When you test that, you're sending sonar pings to the bottom of the ocean. That is very loud. So, it affects the ability of whales to communicate and may have impacts there. So those are kind of broader environmental examples. And really there are very few industries where we don't have impacts on the environment that are outside of the prices we pay for goods.

00:06:30:20 - 00:06:50:10
Jie Ren
Okay. So, let's think about how to, given our understanding of the external cost and how to implement that. Is it reflected in the increased price of the product, or is it reflected in the tax amount that the particular company is paying to the government?

00:06:50:11 - 00:07:15:14
Marc Conte
Yeah. That's right. So, the classic way to address an externality, to get the market back to an efficient outcome is to impose a tax. And so, in the context of climate change, you've maybe heard of carbon tax. We want to put a price on emissions. Right. That's a classic example. Because then what you've done is you've made those social impacts material to the firms.

00:07:15:14 - 00:07:45:04
Marc Conte
And so, if they have to pay those costs, even if they pass those costs are into the consumer, it changes their production decisions. Right. And so that's an example of kind of an incentive-based policy mechanism. And while there is some interest by certain firms to kind of think about changing behavior to reduce greenhouse gas emissions, generally, we think that it is going to take regulatory action to help address those external costs in the context of climate change and most other externalities.

00:07:45:06 - 00:07:50:02
Jie Ren
Okay. So, I did have one episode recording with Adam. Right?

00:07:50:04 - 00:07:51:05
Marc Conte
Yeah, right.

00:07:51:07 - 00:08:19:08
Jie Ren
Talking about this from a legal perspective, we talked about this particular phenomenon from a regulatory perspective. Right. So, texts like that term has been discussed a lot. And then, now let's shift our discussion to AI. Right. Could you please explain the externalities related to artificial intelligence? And we know that it definitely has pros and cons.

00:08:19:12 - 00:08:46:09
Jie Ren
It has been developed so well. Right. And then many individuals are benefiting from the use of, let's say ChatGPT. At the same time, it is also, for example, affecting the consumption of electricity and water. Right. Because a lot of data centers are being built. Right. And then at such a high speed that the supply is not catching up with, and then that could cause burdens, right?

00:08:46:11 - 00:09:00:15
Jie Ren
Or pressure, for example, the market pricing of these, these resources. Right. So, from your perspective, I mean, how would the externalities connect to artificial intelligence?

00:09:00:17 - 00:09:24:15
Marc Conte
I think there are several ways. And the first is just in the development of these, these large language models or the different models that are being used. They were trained using basically all the information that was available on the internet. And so, Google and other companies, they scanned textbooks, they scan novels. Right. And the people who wrote those novels did not get paid.

00:09:24:17 - 00:09:27:04
Jie Ren
Copyright Law being violated.

00:09:27:05 - 00:09:59:05
Marc Conte
Exactly. And so that's a classic example of it should have been costly to acquire that information. They didn't pay those costs. So that means those costs are external to their decision making. Right. So, if they had had to pay those costs, then the cost to users of using those models would go up. So just from the beginning, these models were developed using a staggering array of information that they were able to get access to because they didn't have to pay the full price for that information.

00:09:59:07 - 00:10:22:03
Marc Conte
And that also builds a little bit on the tech sector business model of it's really an ad revenue-based model, right? So, the user typically or the consumer is buying a product from the producer, but in this case the consumer or the user is the product. In some ways, because having lots of users makes the value of their— [ Jie Ren: Retention economy.]

00:10:22:04 - 00:10:53:11
Marc Conte
Yeah, exactly. And so those are two examples. Before we even get to what you brought up, that these GPUs, graphics processing units are incredibly energy hungry. Yeah, right. We're doing maybe even tens of billions of calculations, sorry, tens of billions of queries and queries every day. Yeah, right. So, it's just a staggering amount of activity which requires the processors to be working almost constantly.

00:10:53:11 - 00:11:13:07
Marc Conte
I think 99.97% of the time they're active. And that is energy intensive because they need energy to run. But it's also energy intensive because when the chips are running so constantly, they generate a lot of heat. And so, then we have to do something to keep them from overheating the space where they're located.

00:11:13:08 - 00:11:13:22
Jie Ren
Like water.

00:11:14:00 - 00:11:36:16
Marc Conte
That's right. So, we could either cool with air which they used to do. But those fans are very loud and people complained about living near them. So, the alternative is water. And there are a couple of different approaches. Open loop systems basically just pull water from a water source. So just like a house, you have your faucet on, they just have the faucet on and it pulls water in.

00:11:36:19 - 00:12:01:02
Marc Conte
The water goes through the computing arrays to cool them off, and then it gets dumped out when it's warm. That's called an open loop system, but it's very water intensive. And there have been some examples, for example, a meta data center in Georgia. Once it came in, it led to some drought conditions in the area and water bills went up by about 30%.

00:12:01:04 - 00:12:21:03
Marc Conte
So, firms are thinking about how can we reduce our demand for water. And one way to do that is to create a closed loop system. So, you take a fixed amount of water and then you just run it through the tubes by the GPUs repeatedly. But if you're going to do that, that means you have cold water going to the chips.

00:12:21:03 - 00:12:57:17
Marc Conte
But when it goes through the chips, it's hot and somehow you have to cool it down again. And to do that, you need more electricity. So, there's a tension here about these resources: energy demand, electricity demand and water demand. And both of these things factor in to additional costs of using the lens that consumers don't pay. So, I don't I don’t know if you use the available chatbots or generative AI much, but a simple subscription of like $20 a month can give you access to around 12,000 queries.

00:12:57:19 - 00:13:25:17
Marc Conte
The ability to write 12,000 questions to the model. So, the price that you are paying does not reflect at all either the cost of the information required to train the models or the environmental impacts. And so that's a problem. And that's where we see an opportunity. I'm doing some research with collaborators at IBM and at University of Colorado Boulder, where we're looking at the greenhouse gas emissions associated with different query types.

00:13:25:19 - 00:13:49:02
Marc Conte
And the hope is it's not to say, oh, we shouldn't use AI. It's to say, can we design these models and place data centers in ways to take advantage of cleaner sources of energy? Right. So maybe put them in parts of the country where our fuel mix is heavily tilted toward renewables, or to handle the queries in ways that reduces the energy needs of the GPU.

00:13:49:04 - 00:14:14:12
Marc Conte
So maybe we have some of these big models, like Claude and Chat that have maybe trillions of parameters, but there's other small models that have maybe 1 billion or 10 billion parameters. And maybe for some of our queries, we don't need to use the huge models. And so, I think there's a recent acknowledgment within the tech sector and the AI sector about these impacts.

00:14:14:14 - 00:14:37:03
Marc Conte
And certainly communities are aware of this. Right. You've maybe seen there have been some protests and some states having rules preventing data centers from coming in. So, I think the public is becoming more aware of this, which I think is a good thing. We want to know what the tradeoffs are of our actions. So, then we can decide is it worth it or not, and how can we change its design?

00:14:37:05 - 00:15:09:12
Jie Ren
I like the ending remark that you said about the public is becoming more aware of this, but overall, the public is not so aware of this compared to the good benefits of just asking for, you know, anything. So, I like the research that you are working on, right. In terms of hopefully to disclose the association between the carbon emission and also the use and development of the AI model.

00:15:09:12 - 00:15:33:02
Jie Ren
So, with that information being more and more transparent and also been disclosed in the media and everything. So, we know as users that when you are interacting with ChatGPT kind of like when you are using the running water and you know, like how much environmental impact that is having on the society here, on the environment. Right. So, yeah.

00:15:33:02 - 00:15:55:20
Jie Ren
So, and then from the economic perspective, any economic tools that you can use to regulate this situation. Should the companies face, for example, carbon pricing, something like this, to regulate, to try to mitigate the impact, negative impact.

00:15:55:22 - 00:16:18:19
Marc Conte
Yeah. I'm a fan of carbon pricing in general. And we've seen it implemented in different parts of the world. We have a you know, there's a small programing in California, but China is doing a lot of work here. The European Union has an emissions trading scheme that they've had for a long time. And I think if we get the price rate, it can be a very effective mechanism.

00:16:18:21 - 00:16:44:11
Marc Conte
I think the other things we could do is think about the placement of the data centers in areas where the fuel mix is tilted toward fossil fuels, or in areas where we're already thinking about water scarcity being an issue, maybe because of climate change, maybe those are things that we'd want to restrict some of those. The use when the AI company is given access to the land.

00:16:44:14 - 00:17:06:14
Marc Conte
But to me, and as an economist, I think the price signal is a very effective way to change behavior. Because once you have to pay these costs, then you're going to do your best to reduce the cost that you have to pay. Right. And that's how we can get the markets back on track to these efficient outcomes.

00:17:06:14 - 00:17:17:17
Marc Conte
And I will say just anecdotally, I've read a couple of articles and heard some conversation about trying to put data centers into space, right? Like this—

00:17:17:19 - 00:17:17:23
Jie Ren
We’ve talked about this.

00:17:18:05 - 00:17:40:12
Marc Conte
This impact could be sufficiently great. You know, a lot of these, tech companies have already walked back their goals for carbon neutrality because of AI, because it's so energy intensive. And maybe it's just like, okay, it's pretty cold in space. Yeah, not much energy needed to cool it. Maybe we can do solar up there to get the energy more easily.

00:17:40:12 - 00:17:55:22
Marc Conte
So, I don't know enough about the technology, but I mean, there are very thoughtful people thinking about this in these companies, and it's just a matter of how quickly can we make these major changes, and when will the technology be ready? Yeah.

00:17:56:00 - 00:18:07:20
Jie Ren
Let's continue to talk about sending data centers to the space. Maybe we are not annoying, you know, the residents in the particular area. We are annoying the aliens, right?

00:18:07:22 - 00:18:08:14
Marc Conte
Right, right.

00:18:08:15 - 00:18:38:12
Jie Ren
Yeah. Right. So, yeah. So, at this time of the episode recording for this episode and then, so we just had the IPO of Space X, right? And then which is the largest IPO ever, making Elon Musk the first trailblazer, in the history. So yeah. So, do you think sending data centers to the space is feasible and also sustainable?

00:18:38:12 - 00:19:03:06
Jie Ren
And of course we see the upfront high cost, the launch cost. Right. Because you have to use the rocket to send the data centers there. And how about the maintenance and if something goes wrong. And of course, you have because you are sending it to the orbit in satellites. Right? You have endless right solar energy. And how do you cool it?

00:19:03:11 - 00:19:18:16
Jie Ren
I mean, it's a great idea theoretically, but how feasible that is, how sustainable that is, right? And of course, sending the data centers there, in theory you are getting more data transfer. Right? So yeah.

00:19:18:18 - 00:19:43:00
Marc Conte
It's a big question and I'm not trained to really have the answer. But if we think about what might make it appealing either there are a lot of costs. So, meta had announced that they were going to build a Hyperion data center in northern Louisiana, and it was the facility was scheduled to be the size of the island of Manhattan.

00:19:43:02 - 00:20:17:22
Marc Conte
Wow. Right. So that's a huge amount of land. It's a huge a lot of materials and construction expertise and expertise in these new emerging fields about cooling and designing the arrays to get the GPUs located, to do the work without overheating. And so, there's been some work thinking about all of those costs that they impose on society. And I was quoted in a, an article last year thinking about the middle class bearing a lot because a lot of this need for this massive construction.

00:20:18:02 - 00:20:45:16
Marc Conte
Right. That's inflationary. Right. And so, I think in space, I don't know, it seems like there would be some similar inflationary pressure, but maybe it would be a sufficiently specialized workforce and maybe the materials needed and the size of the centers or the arrays could be modified in some way where if overheating is not a problem, maybe you don't need these massive things.

00:20:45:16 - 00:20:56:02
Marc Conte
Maybe you could have many small things that communicate with each other. So, I'm not the right person to offer an informed or insightful opinion, but I certainly have an opinion.

00:20:56:03 - 00:21:19:18
Jie Ren
No, it is. It is insightful. I was thinking that, I don't know, we'd be jokingly like sending the data center to the space. So, you are not actually belonging any part of the earth. So that could be a way of avoiding certain types of tax and other certain types of carbon pricing, right? Because you're not like affecting any residence on any territory.

00:21:19:19 - 00:21:21:15
Jie Ren
Right. On Earth.

00:21:21:17 - 00:21:43:15
Marc Conte
So yeah, right. I mean, it is an important, you know, economists care about pricing and tradeoffs, and they also care about property rights. And so, understanding what the implications are and how you would write, we probably don't want these huge companies that have trillion heirs as CEOs not to be paying their taxes the way, you know, people, individuals.

00:21:43:18 - 00:21:47:17
Marc Conte
Yeah, get in trouble. Households get in trouble for not paying their taxes. Exactly. Yeah.

00:21:47:19 - 00:22:12:10
Jie Ren
So now so far, we have discussed two possible solutions. One is the possibility of imposing carbon pricing to this companies. Second, strategically planning the place the location of the data centers. And then how about tax? I know that definitely it is being considered right. How about the carbon tax?

00:22:12:11 - 00:22:36:22
Marc Conte
Yeah, I don't know if it's being considered in the US in this context, but certainly as I said, we have examples elsewhere. Different industries are taxed for their emissions. And I think this could be something that could be feasible here. I don't think under this administration that's likely. Right. There's been a lot of kind of quid pro quo behavior here with money going in both ways for favors.

00:22:37:00 - 00:23:12:18
Marc Conte
But the other thing that I think that has emerged in some other environmental areas is, you know, like labeling organic goods where people are willing to pay a premium. I have heard about some AI companies thinking about being we’re the green provider. Right. So it may be that as consumers get more information about these impacts, the impacts of AI use on greenhouse gas emissions and water consumption, people might choose to do work with companies that have designed their facilities to reduce those impacts.

00:23:13:00 - 00:23:41:18
Marc Conte
So, there is some chance that the market could provide signals through a higher willingness to pay for subscriptions for those products or those models or other aspects. I'm not, as you said earlier, some people are knowledgeable about it. The general public probably is still not sufficiently knowledgeable about it. But, you know, we're living in an age certainly of misinformation, but it's easy to spread information as well.

00:23:41:18 - 00:23:46:15
Marc Conte
So, people could find out about this pretty quickly. So, I'm not sure. Maybe there's a possibility they're.

00:23:46:17 - 00:23:51:16
Jie Ren
Taking advantage of a social media to diffuse positive and truthful information.

00:23:51:17 - 00:23:52:04
Marc Conte
Yes.

00:23:52:05 - 00:24:15:07
Jie Ren
Right. For sure. So, could we also go to the global market to talk about like in different regions and then so in your opinion, right. So, what is the status quo of these different regions in terms of using the tools to regulate AI related markets?

00:24:15:09 - 00:24:51:07
Marc Conte
I probably don't know enough about the approach to AI in different countries, but in thinking about climate policy, certainly different countries have different tools that are available. [Jie Ren: Yeah.} So, the China's governmental structure, it's pretty easy for them to impose some type of regulatory protocol. And what that means is if they impose a regulation that is pro-environment, the industry responds pretty quickly.

00:24:51:09 - 00:25:32:15
Marc Conte
That, of course, is more challenging in the US, where you have more choice in your elected officials, you can impact policy as a stakeholder. And so, yeah, I think I would be surprised if something like this happened in the US in the next couple of years. I do think that there has been bipartisan proposal of restrictions in the US, and so we'll see if that comes to bear any fruit as we become, you know, more as we integrate AI more into our lives, because I don't think anything is going to be sufficiently impactful to stop, you know, maybe we can slow down adoption and incorporation.

00:25:32:15 - 00:26:08:10
Marc Conte
But I think there are many benefits, as you noted. Now, those benefits may not accrue to everybody we're thinking about beyond the environment, massive job loss things of those natures. So, some people will certainly benefit. It's not clear that everyone is going to benefit. And there you have an opportunity to think about distributional consequences. And should we allow the winners to keep winning, or should we sort of have a framework that allows some transfer and have some of the people who have been left behind a little bit previously be given more access to the benefits?

00:26:08:12 - 00:26:45:17
Jie Ren
Yeah, I agree with you. The ship has already sailed, but we need to regulate the direction for the ship to go to. So, while the government is thinking about the different ways it could be the legal framework in terms of regulation, it could be the economic tools to regulate this market. We also need the voices from the public, because at the end of day, those are the people, the users, the stakeholders that are taking the external costs, right, to let them have their voice heard through the use of social media, for example.

00:26:45:19 - 00:27:27:18
Jie Ren
Right. Okay. So, let's talk, let's switch our conversation to a different direction. So yesterday when I was like having a recording with Adam, we talk about the negative impact of AI on the environment, right. In terms of affecting the supply of the resources, including electricity and water. Now, maybe let's reverse logically. Yeah. So, do you think the availability of the electricity supply and also all the needed resources can limit the development and usage of AI?

00:27:27:20 - 00:27:28:21
Jie Ren

00:27:28:23 - 00:27:56:06
Marc Conte
I think certainly there is an urgency that we're seeing from these companies. And the way that manifests is just like a house or a factory that needs to be you want to be connected to the electrical grid because that's how you get electricity. The country has a huge interconnection queue. So, of several years wait to be connected for some of these bigger projects.

00:27:56:08 - 00:28:22:07
Marc Conte
And what the AI companies are doing in response is they're just building their own energy production facilities. Right. And to this point, it's typically been coal fired turbines or natural gas or maybe even diesel generators. So, there is a problem there where there's a limited access to the grid. It's forcing them to choose energy sources that are relatively dirty.

00:28:22:09 - 00:28:35:08
Marc Conte
Could they choose energy sources that are relatively clean? Could they use solar and wind? Well, one of the problems here is those renewable energy sources are intermittent in some way, right.

00:28:35:11 - 00:28:36:21
Jie Ren
The sun.

00:28:36:23 - 00:29:01:10
Marc Conte
Half of the day, the chips are running basically 24 over seven. So, what's needed is really innovation in storage, in storage and energy storage. And I think a lot of companies are doing this. But you mentioned a limited access to electricity. Could it slow it down? I think it has probably already slowed it down a little bit in terms of having to build these things.

00:29:01:12 - 00:29:28:21
Marc Conte
But the other thing that's happening is the kind of concern about the reliability of the grid and all that stuff. We haven't talked about nuclear as a way forward. So, a lot of companies are talking about small modular reactors. And in the wake of Fukushima and, you know, memories of Chernobyl, yeah, it may be hard to have communities be like that.

00:29:28:23 - 00:29:30:02
Jie Ren
How do we safeguard it?

00:29:30:03 - 00:29:57:23
Marc Conte
Yeah. Yeah, exactly. And so, you know, it's a big problem. And I think we are starting to take all aspects of the problem a little more seriously. And so hopefully we're moving toward a world where we balance the benefits and costs better than we have to this point. I don't think, you know, I'm not interested in taking access to AI away from people.

00:29:58:01 - 00:30:28:09
Marc Conte
But I do think that if you had to pay the true price for your use of AI, you may not spend as much time, you know, asking it to make a birthday card for you or to tell you how to plan your road trip right. You might be willing to do some stuff on your own. [Jie Ren: Yeah.] And so really, what my research is trying to do here is not to throw the brakes on it, but just to say, hey, this is going to be hugely beneficial to humanity.

00:30:28:11 - 00:30:49:05
Marc Conte
Can we do it in a way that doesn't impose these unnecessary costs, these external costs? And I think there are ways, ways forward here, but it's just a matter of can we get the companies who have their own incentives and their own objectives to do things that would be beneficial to society? That's where I think regulation will need to come in.

00:30:49:07 - 00:31:17:02
Jie Ren
Yeah, I like the phrase that you use that it will affect humanity. And then right now what we see is that truly is a struggle between two things. One is the short-term monetary gain; the other one is the long-term economic impact [Correction: long-term environmental impact]. And then not everyone is thinking about this in the long run, right? So, we definitely need to advocate for this one more as educators.

00:31:17:02 - 00:31:43:21
Jie Ren
And we need to get this message across. So, let's talk about education as a final remark. Right. Both you and I are educator's right. So, here's a question for you and a final remark to any students that are definitely using AI models day to day in terms of being mindful of the environmental impact of AI.

00:31:43:23 - 00:32:14:02
Marc Conte
So, I say as an economist, I think that discipline is incredibly useful today. It teaches us about the importance of tradeoffs and incentives. And so, if you can understand those concepts and apply the approach of the simple, setting the marginal benefits equal to marginal cost to get the optimal outcome, you can really do a lot to help design systems to be more effective.

00:32:14:02 - 00:32:41:10
Marc Conte
And so, students who have training in this discipline, especially, you know, with an environmental perspective from the economics training, you might be more likely to think about your behaviors. And for those of you who you know, for students who are great incorporating this into everyday life, I think there's probably a curiosity as a student, right? That's why people are students.

00:32:41:10 - 00:33:03:00
Marc Conte
They want to learn about something. I think we are developing information that is not quite keeping pace with technology, but hopefully will be disseminating this knowledge so that they will be able to access it and make more informed decisions. So again, I think great opportunities, but just want to be aware of the tradeoffs.

00:33:03:02 - 00:33:07:03
Jie Ren
Thank you so much for the wonderful conversation. I truly learned a lot, Marc.

00:33:07:03 - 00:33:07:18
Marc Conte
Thank you.