Ep 47: The Environmental Effects of Generative AI

Generative AI programs such as Chat GPT are proving disastrous to the environment through carbon emissions and fresh water usage. Let's take a deep dive into how generative AI works, how it differs from other types of AI, and how it is affecting our environment.

MIT Open Letter
Sources

Search for affordable flights with Skyscanner
Claim your free Audiobook with Audible
Protect your next trip with World Nomads travel insurance
Get $100 off your first Loop Abroad internship with code Humane Nature

Transcript

Stacia (00:01.39)
Hello everyone and welcome back to Humane Nature. I'm your host Stacia. And today we are going to be talking about AI, artificial intelligence. This has been a massive topic across the board, across like different things. There is currently, if you weren't aware, an AI race going on. All these different tech companies trying to create and release the most high powered

an intelligent AI program and beat out all of their other competitors. But generative AI has had a real toll on the environment. Like it's, it's bad. It is a lot worse than I originally thought when I was doing the research for this episode. So yeah, AI has definitely taken off since generative AI has become more popular with programs and

Generative AI can produce complex outputs including full written articles and papers images and even short videos that are So realistic that it like it sounds like a human it looks sounds like a human made it a lot of the time and It's really scary. It's been really scary for a lot of writers like myself artists videographers

like just content creators in general, because some companies are turning to AI now to write their scripts, write their articles online, like all this other stuff, and it's putting people out of jobs. We've been using AI for a long time, so AI is not like a dirty word. It's not necessarily a bad thing.

But only recently has it gained the ability to generate human -like language and realistic photos and videos. So generative AI is a relatively new thing and it is rapidly progressing and at a rate that isn't sustainable and it's having like huge impacts across the board on the environment as well as on people.

Stacia (02:22.67)
So computing power takes energy and this has always been true. Think of it like your brain. Your brain is like a natural computer. It has all this computing power. Like it requires power in order to function and that's why we need to eat, especially a lot of sugars and carbohydrates. Like that is what fuels our brain. And that's why we feel very sluggish while we have like

brain fog and stuff if we haven't been eating properly or if we'd become really dehydrated.

Computers are the same way. So every time you turn on your computer or type in a Google search, energy is being spent in order to do that computing power.

AI powered search engines require a lot more energy to work than just like your average Google search. Meaning they also admit more carbon than previous search queries. This makes a lot of sense. The processing platform is much, much larger and therefore takes more energy in order to work. The problem is

Generative AI requires long -term training of the computer system in order for it to work, and millions of people are beginning to use it for, like, no real reason.

Stacia (03:55.118)
Neither OpenAI or Google, which are like two of the biggest platforms, have stated that like what the computing cost of their AI products are. They're keeping that very close to the chest and probably for very good reason. A third party analysis by a researcher on a platform similar to GPT -3, which is what chat GPT is based on, which is like a really big generative AI.

thing led to more than 550 tons of carbon dioxide. So I don't know if this is like per search. They didn't say very well in the article if this was per search per day, just in total. But a lot of this carbon that's being released, it's not just when we are actively searching and creating things with generative AI.

It is also for the long -term training. So this carbon output is, and water usage, and we will get into that later, is massive even before, like while they're developing these. So even before people are able to use it.

So this 550 tons of carbon dioxide that was released by GBT3 is roughly the same amount as a single person taking 550 round trips by car between New York City and San Francisco. I know a lot of my listeners are not in the US. I see a lot of you are in Germany. Hi, hi, hello, thank you for listening.

but the U S is massive. The country of the U S of the United States is roughly the size of the continent of Europe and San Francisco, California and New York city, New York are on opposite ends of the country. So this distance is massive. It's nearly 3000 miles or just over 4 ,700 kilometers.

Stacia (06:12.398)
for a single trip. That's one way. So double that for a round trip and then multiply that by 550 and that is the amount of carbon dioxide that is being released by chat GPT, roughly. That's a lot. That is a lot. Another study suggested that chat GPT is consuming the energy of 33 ,000 homes.

more than five times the energy of a conventional web search. So I think that's per search here. A lot of these articles that I was reading for this, which I will, the sources will be linked in the description as always, weren't super clear on if this was per search, per day, per user, whatever, but this sounds like it's per search.

It also suggested that within a few years, large AI systems are likely to need as much energy as small countries in order to run, which is absolutely insane.

Not only does generative AI use huge amounts of energy, but it also requires large amounts of water to cool the systems. And these companies are using fresh water because salt water corrodes the electrical equipment. And that would just cost them more money to manage. So they are using fresh water. The country of West Des Moines, Iowa sued GBT -4

After it was revealed that they used over 6 % of the county's freshwater supply in the month before its official launch. So this was before people were even using it. This was before the millions of people were plugging things in and using this generative AI. This was just during the training of this AI system. It used 6 % of the county's freshwater supply. That is insane.

Stacia (08:12.942)
This amount will only increase when more people are using this. And I think GBT4 has been released by now. It's really hard to get accurate and complete data on the environmental impacts of generative AI because these companies are keeping them very closely guarded. And that means you know it's bad.

Stacia (08:36.686)
companies are again not using salt water, which would be better for the people because it corrodes the equipment. It would cost them a lot more money in the long run to use the salt water and because they would have to continuously maintain their equipment. So of course they don't want to do that. It's going to cost them more money. It would be one thing if generative AI was only used for certain areas of research or medicine.

because it is very useful for some people. It is an incredible invention. It is absolutely insane the things that generative AI can do, but millions of people use it every day and they really don't need to. Like we're using chat GPT to, people are using chat GPT just to make fake art.

They plug in random things into it and see what crazy image that it creates, which is all fun, but like it's using so much energy and releasing so much carbon in order to create your random photo of a horse flying over a rainbow with an octopus on its back. I don't know. I just made that up. But that's the kind of things that people plug into.

these generative AI and then it comes up with like this really obscure image and they think it's funny and a lot of it is really funny but yeah, I don't think most people realize how much of an impact this has on the environment. And we're all using chat GBT whether we want to or not. Even if you're not directly going to the website, you're using it. Companies like Google, Bing and Facebook, which I guess is meta now.

have implemented chat GPT into their search functions. So if you search for something on Google now, there is an AI generated, like suggested answers at the top, and that is chat GPT. Facebook has it too, where at the top it says like, ask meta anything, that is generative AI. So it is being used everywhere, and we're using it.

Stacia (10:56.302)
even if we're trying to avoid it.

Stacia (11:01.262)
Martin Bouchard, I think I'm saying that right, co -founder of Canadian data center company QScale, has estimated that adding generative AI to the search process of like general Google searches will require quote, at least four or five times more computing per search at a minimum, end quote. So it could be much, much larger than this, especially as they continue to grow.

and increase these things. And it's, yeah.

Before generative AI, computing data centers already accounted for around 1 % of the world's greenhouse gas emissions. Now that number is going to skyrocket. They say at least four or five times. And it's probably much, much more than that. He was being very, he later says that he was being very conservative with that estimate.

OpenAI chief executive Sam Altman admitted earlier this year in 2024 that the AI industry is heading towards an energy crisis and that it will consume vastly more power than they originally expected. And his suggestion for this? Nuclear fusion. Which it turns out he has money in. He's got his hands in the pocket of

nuclear fusion, which honestly does not, nothing surprises me anymore. Experts say that nuclear fusion will not contribute significantly enough to decarbonizing just in general to combat the current climate crisis. So it is definitely not a solution for making generative AI more environmentally friendly.

Stacia (12:59.63)
An open letter written by an interdisciplinary team of MIT researchers explores some of the key drives of AI's explosive growth over the last year and why sustainability considerations are currently lacking. They are calling very loudly for more responsible gen AI development using a comparative benefit

cost evaluation framework. I will link this open letter in the description. It is titled the climate and sustainability implications of generative AI. It is hard to read.

Stacia (13:43.758)
if you're not like well versed in a lot of the terminology, but it is, it'll be good to peruse because the numbers are very frightening.

So the big question is, can AI become cleaner? Is it possible? Theoretically, the environmental footprint and energy cost of integrating AI into search, like gen AI into search, could be reduced by moving data centers onto cleaner energy sources and redesigning neural networks within them to become more efficient. So right now, it is about product.

productivity. These companies are not really taking environmental factors into account when creating these. They're just trying to build them as quickly as possible and make them as fast as possible in order to beat out their competitors. But I would argue that if a company focused on becoming green, they would be more popular among a lot of people.

Then again, we could also just not have generative AI available to the general public until cleaner options have been implemented. But that would also mean that these companies would lose money.

Also, that ship has sailed. It's out. It's available. People are using it. Generative AI does not need to be used by a middle school student writing an English paper or creating random images online for laughs. It just doesn't. That's not why it was created.

Stacia (15:31.47)
The massive amount of carbon emissions does not seem worth it for minor gains in search accuracy on these websites because that is what they're competing for is search accuracy, which is great. That's what we want when we're searching something into Google, but the miniscule accuracy improvement on these searches isn't worth all of this carbon footprint. Like it's really bad. Many researchers and environmentalists

agree with me. So not just me on this. Hopefully we can implement, we can quickly implement ways to make AI cleaner because it's unfortunately not going anywhere. And other than the environmental impacts of it, there are a lot of unethical things about gen AI that have come about in the last year.

Stacia (16:34.606)
Loki, Loki out.

Stacia (17:16.526)
So I also want to discuss the ethics of generative AI and some of the problems that have surfaced because of it. Despite being called quote unquote generative, AI does not create new text or images. Instead, it combs the internet for pieces of information that match the search query and they piece it together in a new way based on what has been trained to do.

As a result, a lot of artists, writers, musicians, they're all having their material stolen if it exists on the internet. There are many examples of an artist's signature still being in a generated piece of quote unquote art that was produced by AI. And usually it's like warped in some way, but still clearly somebody's signature that created

some part of that new piece. I will include a photo on my Instagram and website of an example of this and it's really messed up. And I know some people who have had their artwork stolen by generative AI. There are also a lot of people claiming to be AI artists, which is not, should not be a thing. They're not creating anything. They're plugging keywords into

Gen .ai and the computer is stealing bits of other people's artwork to create a picture. And some people are calling themselves artists because of that. And I have a real issue with that.

In addition to stealing other people's original works, AI is stealing jobs. Companies focused primarily on profits over quality of products are beginning to use generative AI to produce written articles, images for promotions and other creative pieces that would normally be outsourced to an artist, writer, photographer or other creative position. And I know throughout history, people, you know, when new technology comes around, people are going to...

Stacia (19:29.71)
saying, you know, this is stealing jobs. It's different than in the past. These jobs are best done by people. These artistic jobs should not be done by computers. And with a keen eye, you can quickly spot fake images and articles that were written by AI because they just seem a bit off.

because they are. They were pieced together by a computer rather than by a human being. So, yeah, so this is the first time, and this is something that John Oliver talked about on his AI, generative AI episode, which I will link in the sources. I encourage you guys to go watch it. It's on YouTube. This is really the first time that

a new technology instead of stealing blue collar jobs and like improving blue collar work is stealing more white collar work. So it's taking the place of artists, but also it is expected that AI may replace newscasters and lawyers and diagnostic doctors. So,

Those are the things that we may be seeing in the future.

Not only this, but digital art companies such as Adobe have implemented AI and are stealing directly from their consumers who use their online product. Adobe has stated that they have the right to share your work because you use their product. And many artists have come out and stated that they have seen their work stolen in AI generated pieces that only really existed in the Adobe, like

Stacia (21:29.454)
online. I don't remember exactly what it's called, but it's like a Photoshop online program that digital artists use. And a lot of people are leaving Adobe or choosing to use something else because their work is being stolen.

Many teachers and professors are reporting increased levels of cheating in school as students use AI to write their papers or complete other assignments for them. So I use AI, I will admit, not generative AI. I use AI in my keyword research for my articles. The AI program that I use combs Google and finds frequently searched keywords that helps my

articles rank highly on like the front page of Google so that more people see it, more people click on it. That is different than plugging in the like the essay assignment or whatever into generative AI into a platform and then AI spitting out a completed essay, a completed paper and that student did not.

write it. Very different.

Stacia (22:51.982)
Companies are using AI to search through resumes as well. This is also not new, but these AI programs are immediately tossing out the ones that don't fit the algorithm, even if the candidate would be a great fit. So a big problem is that decorative resumes, different types of font, photos on resumes that in the past were used to help your resume stand out.

Like if somebody had a stack of resumes something like that would help your resume stand up stand out now The AI can't read it and it just gets tossed out so That that is a reason why a lot of people aren't getting interviews because a lot of these job application processes are online now and They're combing through resumes

with an AI algorithm and a lot of these resumes are being tossed out for various reasons. John Oliver even talked about how one of these resume programs that he looked into, because these AI programs teach themselves things, which is like crazy to think about. And one of these resume AI programs somehow

taught itself that that people named Jared who played and people who played lacrosse were the best candidates for some random job. And so all of the resumes that had that in it were pushed through and none of the other ones were, which is very strange. And it is a lot easier to teach an AI program something rather than it

to unlearn something. So that can cause a lot of problems.

Stacia (24:54.67)
Celebrities' voices are also being stolen without their consent to create content that they are not being paid for. Scarlett Johansson actually just sued ChatGPT directly for using her voice for an AI personal assistant that they developed, despite telling them not to use her voice. So, that's pretty messed up.

Generative search and image AI has also spread a lot of misinformation to the general public. So the examples I've personally seen and come across are fake but very realistic photos of animals that are wildly inaccurate. For example, one of them that I've seen is a mother bald eagle feeding her babies.

And the babies were just like miniature versions of that bald eagle. So it's, they had the white feathers on the head and the dark feathers on the body. When in reality, baby bald eagles are like, they look totally different. And yeah, baby birds are ugly as hell. And they do not look like adult birds at all until they reach adulthood. So this was obviously a

generated photo by AI and that in itself probably like is not dangerous but think about like the consequences of that as a whole. Like people are looking up what certain animals look like for like research or to teach their kids or whatever and they're like they're completely wrong and it's hard to tell the difference sometimes because these photos are so realistic looking.

Some of the search query answers that pop up at the top of Google now because of AI can be downright dangerous. For example, AI has included fake information that is put into comedy websites such as The Onion, and they have listed them at the top of Google as like fact. So I think one of them that I saw was that like NASA encourages

Stacia (27:05.774)
people to like, or say that it's healthy for people to eat like two or three rocks a day. And it was from like an onion article. And it's being listed as fact. So people could be like doing some really dangerous stuff because Gen. AI on Google told them that it was, it's the right thing to do.

so again, AI combs the internet to find like the best answers to the search query. That doesn't mean that those answers are accurate. We all know that the internet is a cesspool. There's a lot of misinformation, a lot of dangerous information on the internet and the AI is crawling that too. So.

On the darker side of all of this, people have been using generative AI to create and spread pornographic images of others, including children. These are called deep fakes, and it's been a massive problem for teenagers, especially in the last few years. So people are making these deep fakes. They're

making it's also called revenge porn, sex torsion. These are all like internet crimes now and teenagers are using them against each other. People are making them of celebrities. I know a gen AI like pornographic video was recently circulated of Taylor Swift and they all look very realistic.

Many parents are pulling all images of their children from the internet as child predators are beginning to use generative AI to create and spread child pornography using their images. So you can put a completely innocent photo of your kid on your Facebook page. They're just like sitting and playing in the grass. They're fully clothed. They're not doing anything even remotely sexual.

Stacia (29:19.31)
And these predators are taking the image of their face or part of their body and they are creating horrific images that they can then sell to other people.

So, all of this, the deep fakes, revenge points, extortion, and all these other crimes using generative AI are becoming more commonplace and they are very difficult to pinpoint to a creator. So it's, yeah, it's really scary.

Is all AI bad though? Like I listed out kind of the worst of the worst there, but we have been using AI for years and it's not all bad. And it's not all generative. Non -generative AI is called narrow AI and narrow AI has a very specific algorithm, something very specific that they are taught to do.

and they don't continue to teach themselves new things, they don't generate new things, and we've been using this forever. Examples of really great uses of narrow AI include your email spam folder, like automatically recognizing and placing spam emails into that folder so that you don't have to see them in your inbox, frequently searched questions on Google, AI information platforms,

are now being used for easier scheduling. So people or like small businesses are using them for scheduling their employees or whatever, or people are using it as like a personal assistant to schedule out their day. So that's one less thing that they have to do. And people are also using it for faster research, which is what I do with my keyword research. And

Stacia (31:15.054)
Narrow AI is also used for face recognition on smartphones to unlock your phone. And it's also used for recommended content on streaming platforms. So if you're watching Netflix and you finish a movie and it's like recommended on what to watch next, that's Narrow AI.

Stacia (31:35.854)
Not all AI is generative and can be quite useful for people's daily lives and careers.

Generative AI, in my opinion, was released far too quickly without the proper protections in place for both people and the environment. It's causing a lot of damage. There is a distinct difference between everyday AI that we've been using for years and generative AI, and not all AI is bad. It's kind of like a trigger word now. We are a very long way from

general AI, which is a totally different thing. It's like super advanced generative AI. It's like the sci -fi AI robots in like movies like iRobot, Jarvis in Iron Man, and that super freaky lady in Smart House from that Disney movie. That's all general AI, which is like the super generative AI, and a lot of people are genuinely afraid that these

intelligent robots are going to take over the world. And while I'm not going to sit here and say that that's not a possibility, we are a long way off from that kind of AI.

Generative AI can also do a lot of really wonderful things in science and medicine. It's currently being used to recognize early vocal changes in people with early Parkinson's patients. So it's being used to diagnose people with Parkinson's that couldn't be diagnosed yet. And Parkinson's when it's caught earlier is easier to manage. And you can slow the progression a little bit better.

Stacia (33:22.062)
It's also predicting the shape of all proteins known to science, which is absolutely insane and is being used to create a lot of new medications, vaccines, things like that. And generative AI is also being used to better train surgeons before they even step into an OR. So all of these things are fantastic. To me, these scientific advances are worth the extra emissions. But

not teenagers cheating on their papers or spreading fake images of their exes. So until gen .ai is cleaner, maybe it should be restricted. That would cause an uproar, I think, because so many people are just having fun with it right now or using it for their jobs. But for now, it really should be.

only used for these very specific purposes and these purposes are what it was built for.

So here's what you can do about the environmental problems of AI. First, don't use generative AI if you can avoid it. If fewer people are using it, the less money they're going to make and the more inclined these companies are to make these environmental changes. Be very careful about what photos you post of yourself and your children online.

Look up the new recommended template for resumes if you are looking for a new job so that AI will recognize and accept your resume.

Stacia (35:02.606)
Learn to recognize fake text and images and do your best not to spread them unless you are actively pointing out that they are fake.

Double check facts that are written by AI on the top of Google search results and don't just like take them as facts. You want to double check where they got their information from.

Always educate your friends and family. If someone is just playing around with gen AI images, maybe say like, hey, that's actually really bad for the environment and here's why, and maybe back off until these cleaner products are coming out. And finally, vote for politicians who want to implement environmental and personal protections with generative AI. There are a lot of bills and stuff being talked about with politicians.

But because there's a lot of money tied up in gen AI, a lot of these aren't going to be passed because these companies are buying out politicians and that's not a new thing. But we need to be putting pressure on these companies like Google who state that they want to be carbon neutral. They say that they are striving to be

a carbon neutral company within the next like 10 years or so, but that is never going to happen if Gen .ai does not become a cleaner. It's just not going to happen. So if these companies are quickly growing Gen .ai and not attempting to make them better for the environment, they are never going to be carbon neutral and we are never going to be able to fight climate change ever. Like Gen .ai is making it worse, period.

Stacia (36:51.726)
So, yeah, there's only a few small things that we can do really because as always with these pollutants, it's always these massive corporations that are like the ones that have to change things. But we as consumers can put pressure on.

Alright, so that is all I have for you today. I will link that MIT Open Letter in the description. And as always, all my sources will be linked in the description as well. I will have photos for this, maybe some examples of bad photos.

some bad advice that's put on the top of Google, things like that, and warped artist signatures on the bottom of these images. I will have that all on my Instagram. You can follow me on Instagram at StumbleSafari. That Stumble is in fall, Safari is in an African safari. That's also the name of my website. So if you don't have an Instagram, you can follow the link in the description over to my website or just go to stumblesafari .com.

I will have a webpage for this episode with those images on there as well as links to the sources for this episode and transcripts for this episode. So, and yeah, there's also other types of travel content on there as well as webpages for other episodes that I've done. So thank you so much for listening and I will hear you next time.