Transcript: How AI Affects the Environment
CLARICE: Hello, everybody. And welcome to our 100th episode of Environmentally Speaking.
MARISA: Hi, everyone. I’m Marisa Desautel, an environmental attorney.
CLARICE: And I’m Clarice, celebrating with a very tiny, very shiny party hat. I know this is an audio platform, but I needed to celebrate.
MARISA: I hand it to you that you always manage to accessorize appropriately. Happy 100.
CLARICE: Happy 100. And today I thought — I got to take this off. I can’t take myself seriously.
MARISA: Okay.
CLARICE: There we go.
MARISA: There.
CLARICE: It’s so tiny.
MARISA: It is tiny.
CLARICE: Folks, just for reference this little cone party hat fits in the palm of my hand. We’re talking like it’s — I think it’s a pet hat. I’m not even going to lie to you.
MARISA: Are we dressing up for the Halloween episode? We did last year.
CLARICE: Yes.
MARISA: Okay. All right. Good to know.
CLARICE: I actually have an accessory under my desk.
MARISA: I’m sure you do.
CLARICE: We’re ready to roll.
MARISA: I’m sure you do. What are we talking about today?
CLARICE: So we’re talking about something pretty — I think pretty timely. It’s been coming up a lot in my work. I don’t know if it’s been coming up in your work at all, but we’re talking about AI and the environmental effects. So I don’t know if that’s something that you’ve given any thought to or if it’s something that’s sort of come across for you, but because we’re — for folks who don’t know, I work in the alternative dispute resolution field primarily and we’ve spent a lot of time in my office talking about AI, what effects it has, how we can potentially integrate it. Is it going to be useful. How can we use it. And the fact that we’re constantly talking about what is this tool and what are its benefits to us, if any, I was thinking about are there any environmental impacts. It’s not a tangible thing. It doesn’t have a physical footprint, but are there any impacts. So I started looking into that and I was wondering has the office looked into it? What are your thoughts on it?
MARISA: The only examples of artificial intelligence that I’ve come across are, number one, an attorney said that there was some kind of free seminar on the topic and how attorneys might be able to use it. And then I had an expert bring it up and she indicated that through ChatGBT —
CLARICE: GPT, yeah.
MARISA: GPT. Excuse me. That she plugged in the topic she was going to be testifying about and it came up with a pretty solid document that she wouldn’t have relied upon, but just for purposes of trying it out apparently it’s pretty reliable because it’s just gathering information from the internet.
[0:03:15] CLARICE: Yeah.
MARISA: Right?
CLARICE: Yeah. So one thing to understand, at its core foundation what ChatGPT and other AI processing tools at that level are, it’s basically a great predictive word generator. So what it’s going to do is you give it a topic and it’s going to make a best guess on what the next logical word is in that sentence.
MARISA: Really?
CLARICE: Yeah. That‘s truly at its very —
MARISA: I had no idea. I thought it just pulled information.
CLARICE: But that’s how it’s doing it. So really, really watered down it’s saying, I’m looking for — if you’re saying, I’m looking for a recipe on bread it’s going to scour the internet, research everything it can find about bread, narrow it down to say bread and recipe and figure out what are the things that are coming up over and over and over again. Well, I’m seeing 99 percent of times water, flour, salt, yeast. Now let me go a little bit deeper. I’m seeing more often than not, maybe 80 percent of the time, these ratios.
MARISA: Okay.
CLARICE: Now I’m seeing 70 percent of the time these temperatures.
MARISA: Okay.
CLARICE: I’m seeing 75 percent of the time these times. That’s how it’s functioning and this is like a rudimentary version of it. There’s so much more math involved. There’s so much more thought involved than that, but that’s sort of what it’s doing.
MARISA: Okay.
CLARICE: And it’s constantly learning from the feedback we’re giving it because at the end — for folks who haven’t used it, at the end of every time you’re using it you have the option of saying, can I reask it the question, was this answer helpful, how is this feedback. So every time we interact with it it’s taking our feedback of that interaction and learning, okay, I went and did that search, it wasn’t great, it was great, I can adjust my predictive word for the next go. So you searched bread. You baked the bread. It was burnt. You went back to it and said, not great. I go and search bread. It gives me something slightly different. I tell it, great job. The next third person is now going to get the benefits of my search and not your search.
MARISA: Oh. Listen to me, like I understand. Oh. Geez. All right. That’s a good explanation, though. I had not spent that much time trying to figure out what it was, so I appreciate the description.
CLARICE: Yeah. I’ve been listening to a lot of podcasts about how it works and then, like I said, we’ve been trying to figure out ways to implement it and if it’s beneficial in our work.
[0:06:15] MARISA: How would you use it in your legal field?
CLARICE: Yeah. So some options for us in the ADR field — we actually just did last month a kind of lunch and learn about how mediators specifically — primarily my work is with mediation — if it’s beneficial for mediators. Are there tools that it can provide us. There are actually simulations out there where mediators can practice their skills. It can pretend to be two parties in conflict and you can have a chat conversation to practice —
MARISA: Oh.
CLARICE: — being in a mediation with them.
MARISA: Okay.
CLARICE: Granted, most of these are behind a paywall, but there are some free demos. And I will say after practicing with that they’re pretty realistic. You can type in different types of legal scenarios whether it’s small claims, a divorce proceeding, child custody, things like that, to businesses looking to dissolve, things like that.
MARISA: Okay.
CLARICE: We asked it some basic questions about what advice would you give a mediator to stay neutral. What advice would you give a mediator in dealing with a difficult party, in overcoming bias, in practicing active listening and it gave some really thoughtful answers. And you could see where it was pulling those answers from. It was coming from some sources that we had seen before and I was pretty shocked by it. So it’s really coming from thoughtful places and in seeing that — and, again, going back to this idea of it’s giving us useful valuable content, but it doesn’t have any physical effects. Does it have an imprint. Does it have any sort of effect on the environment. And I think of the environment so often as a physical thing. The environment is like your trees, your earth, your air, things like that.
MARISA: Right.
CLARICE: And when I started researching this, I was seeing that researchers are starting to see an effect on the environment from using AI.
MARISA: And let me guess. It’s a negative effect?
CLARICE: Yes.
MARISA: Of course. There are also some positive effects of AI on the environment on environmental quality in the sense that AI can assist with predicting how to be more energy efficient and using predictive models. But the issue is the amount of power and energy needed to —
CLARICE: Bingo.
MARISA: — drive AI itself is detrimental to the environment.
[0:09:15] CLARICE: That’s exactly it. That is exactly it.
MARISA: It takes a ton of energy, right?
CLARICE: Yeah. They’re saying it’s something — oh, I have to find it. I’m not finding the actual number right now, but it’s something like 100 times more than your average Google search. And a lot of AI programs are now integrating their AI feature into your search feature. So Bing has one. Google has one. So what will happen is if you go to Google something you can Google and ask AI at the same time, so you’re doing both which is using sort of both sources of energy when you may not necessarily need to. And we’re not thinking about it.
MARISA: Wait a minute. I just made that noise like, ugh, because human tendency — I just did it internally when you were saying you can now Google two ways. I thought, oh, I want to try it, you know. Like, oh, I got to try this out. Do I? No, I don’t. I’m perfectly fine with just Googling something, but I want to try it now. And once it becomes more mainstream, I assume most people are like me, they want to try it, but where is that power going to come from?
CLARICE: Well, here’s the thing is I didn’t think about that. I will confess right now Google sent me an e-mail and said, hey, do you want to be a beta tester for the AI search add-on and I said, yeah, yeah, let me go be a tester so I can AI search before everybody else. Do I get to try something new and get to try something new before everybody else? Yes, I want that. Did I think about the strain on the grid? No, I didn’t.
MARISA: No.
CLARICE: Am I at fault? Yes.
MARISA: Well, I disagree with that. You’re at fault now because —
CLARICE: Yeah. Now that I know.
MARISA: — now you know and if you choose to do it then, yeah, you’re at fault, but part of the issue with environmental matters generally is education. And I’m not saying people are doing anything wrong. Life is hard. You get up every day and you do your best, but it’s not like people are, oh, sitting down and saying, oh, I got to educate myself on any environmental matter. AI — I’m an environmental person. This is what I have chosen to do with my life. Didn’t occur to me that AI would be a strain on the grid, so education is a huge component of being aware about what’s happening with the environment and most people just don’t know.
[0:12:20] CLARICE: They have no idea. Yeah. So they’re saying that not only is this a substantial strain versus your average just sort of search —
MARISA: Internet search, yeah.
CLARICE: — I mean, it’s really exponentially larger. What’s happening as a result are companies like Google, like Bing — do people still Yahoo? If there’s like three people who still Yahoo.
MARISA: Excuse me.
CLARICE: You do not.
MARISA: I love Yahoo and I still have my Yahoo e-mail address.
CLARICE: Well, our friendship is also strained.
MARISA: You had on a tiny hat earlier.
CLARICE: I’m festive.
MARISA: But is Yahoo not a thing anymore?
CLARICE: I don’t think so. Listeners, write in. So these search engines are now building larger facilities to house sort of power grids and just places to power this. They can’t keep up.
MARISA: But where’s that power coming from? Is Google developing wind farms on land and solar farms to power this stuff?
CLARICE: From what I read I’m not hearing what type of power. I’m not hearing where. I think the other thing that I’m seeing, too, is they’re now — and this is going to show how little I fully understand this field is that they’re now creating more data centers. So if you’re thinking of that early 2000s hacker movie where they need to go to the place where all of the —
MARISA: Yes. Where the computers.
CLARICE: — like where the mainframe is stored.
MARISA: Yeah. Yeah. Yeah.
CLARICE: You know, like I’m using air quotes again. You guys can’t see that, but they’re building more of those facilities where they actually need to have these process servers which, again, is taking more power and energy.
MARISA: Yeah. And where’s the power going to come from?
CLARICE: I don’t know. They must know. I don’t know.
MARISA: Okay.
CLARICE: They need some facility and structure and infrastructure to keep up with all of these searchings. What was interesting is the University of Pennsylvania Environmental Innovations Initiative, they had a wonderful interview with [inaudible] who is from the School of Engineering and Applied Science. I’m going to link it in our show notes. This looks like something that he has been studying and he put together a — there was a great interview talking about what the strain is, what this looks like, what the effects are.
And ultimately the big takeaway is just to be a little bit more mindful of is this something where ChatGPT or an AI search is actually going to be beneficial? Is this something where you need an interactive answer, or is this something where Google or Bing or Yahoo is going to be helpful? And it’s just sort of making that decision in the long run of how much power do you need behind this.
[0:15:52] MARISA: Well, and it’s going to come down to — to people making a choice that’s pro-environment.
CLARICE: Yeah.
MARISA: That ought to go well.
CLARICE: Well, I mean, I wonder if now that folks — I wonder if it’s a thing that people are now thinking about. Does this information now affect you, listener? Now that I know I might — I’ve turned — I mean, this morning I turned off that feature, so it’s now as needed. If I need it I will use it, but I don’t need to ask AI everything where that was what the feature was doing before.
MARISA: Yeah.
CLARICE: If I was Googling directions, it would also give me an AI summary of the place that I had Googled directions to.
MARISA: Yeah. I mean, that’s not necessary, right?
CLARICE: No. Something I didn’t think about.
MARISA: Yeah.
CLARICE: I had no idea it existed.
MARISA: And the power’s got to come from somewhere, so that’s the environmental impact, not necessarily that it’s using the power, but that power’s got to be generated somehow. And we’ve already got, I think, a couple of — at least two instances in the past few years where the strain on the grid which we’ve talked about —
CLARICE: Yes.
MARISA: — created a blackout or a brownout, so how are these searches going to increase those instances of blackouts. Will they be more frequent? Will they last longer?
CLARICE: And interestingly I wonder if they’re going to be localized or if they’re going to — where these blackouts, brownouts — where are they going to start happening.
MARISA: Yeah. Yeah.
CLARICE: If it’s just in the area where these process servers and graphic process units are, then will the rest of us care? Let’s be really dark about it.
MARISA: Yeah.
CLARICE: That’s happening just out in California.
MARISA: Yeah. And we’re not feeling the effects of it, then who cares. It would be hilarious if you, Clarice, are at your computer doing work and like all of a sudden the neighborhood just shuts down.
CLARICE: Then I’m going to Google —
MARISA: Clarice, get off the AI.
CLARICE: — way less. If that’s what starts happening, I will feel the effects.
MARISA: I mean, that would certainly operate like a scarlet letter type of branding. You’re on AI and your house just powers down.
CLARICE: Oh, God.
MARISA: Yeah. I don’t know enough about that localization.
[0:18:50] CLARICE: Considering that my neighborhood is largely 65 plus it will just be my house.
MARISA: Yeah. Yeah. You’re right.
CLARICE: So, yeah. We’ll put this interview and another article discussing the impact in the show notes. I’d love to hear if this changes what you think, if this affects how you plan to search or if you’ve even kind of messed around with these tools in the past if they’ve had any effect on your work or if you’ve found them interesting at all. So reach out to us on the socials. It’s Desautel Browning Law, Instagram, Facebook. You can look at my funny tiny hat on YouTube. I thought it was very cute.
MARISA: It was very cute, yeah. And if anyone has questions or wants to follow up with some additional communication, send me an e-mail. It’s info@DesautelBrowning.com.
CLARICE: All right. Thank you, everybody.
MARISA: Thanks. Happy 100.
CLARICE: Happy 100.
