Common Objections to AI - Ridge Light Ranch

This has almost no connection to art, so I’ve been holding off, thinking I didn’t need to address these concerns. However, since I’ve become a bit of an advocate for using AI, people have started asking me some form of these two additional questions:

Doesn’t AI use a ton of water, making it environmentally unfriendly?
Wasn’t AI trained on copyrighted material, making it immoral to use?

I’ve talked about other (probably more important) objections, like AI’s impact on human formation, in an earlier blog post, so in this post, I will stick to answering the two specific questions above.

For this image, I asked Codex to “create an image that seems to represent this whole article.” Then I edited and touched it up a bit. I always think this kind of prompt generates interesting results.

1. Does AI Really Use a Ton of Water?

There’s a lot of press out there about the supposedly absurd amount of water AI uses, and some people cite this as one of their chief concerns about AI. While AI does use water, the numbers are often presented without context, and sometimes are flat-out wrong. Let’s dig into this issue and how it has gotten blown out of proportion.

AI itself doesn’t use water, but there are two main ways it ends up causing water to be used.

First, the big servers AI models run on generate heat and many data centers use water in their cooling systems. In the process, some of this water evaporates, some is discharged, and some is recirculated.

Second, much electricity generation consumes water (regardless of whether the electricity is used to power an AI model or an electric car). Of course, there’s a ton of variation depending on the type of power plant, the cooling system, the climate, and the season. (Some electric plants use outside air for cooling. So again, a lot of variation.)

Corporations and scientists are working on ways to reduce water use in both these areas. For example, Microsoft has a new liquid-to-chip design that uses a closed loop. Once the system is filled with water, no more water is needed for cooling.

This points out the need to distinguish between water withdrawn (much of which is usually returned) and water consumed (water that evaporates instead of being returned to the source). Those two have different environmental impacts, but they are often lumped together, making the water use seem worse than it is.

So yes, many AI data centers and power plants use water for cooling, but the important question is how much water do they use compared to other water uses and compared to other ways to accomplish similar work?

How Much Water- Mistakes that Persist on the Internet

One of the claims that brought the water use issue to the forefront of many people’s minds originated in the 2025 book, Empire of AI. In this book, the author, Karen Hao, claimed that a proposed Google data center in Chile would use more than 1,000 times as much water as the residents used. However, she later acknowledged that the Chilean government source document she relied on had mislabeled cubic meters as liters, which threw her numbers off by a factor of 1000. She’s corrected this in her book, but the old stat has lived on in the internet in the form of scare tactics.

An earlier 2023 paper titled “Making AI Less Thirsty” estimated that GPT-3 consumed 500mL of water for every 20-50 responses. However, the researchers did not have OpenAI’s actual data and had to make assumptions (which they admitted to in the article). Add that to the fact that the newer models are more efficient (see below), and the old number is not indicative to what’s going on today. However, it seems to persist online.

Actual Water Use – The Numbers Keep Dropping

In 2025, Google conducted a study of its own and concluded that Gemini used an average of 0.24 Wh of electricity and 0.26 mL of water (about 5 drops) per text prompt. So, a 500mL water bottle (since it’s an easy amount of water to picture) would be used for 1,900 prompts. It also pointed out that the energy used per prompt was down 33 times compared to the previous year.

In 2025 OpenAI CEO Sam Altman said that the average ChatGPT query used 0.34 Wh of electricity and 0.32 mL of water. By my calculations, this means that a 500mL water bottle corresponds to about 1,560 prompts.

A 2026 Joule study estimated 0.31 Wh per standard query, but said that reasoning-heavy queries would use about 13 times more electricity. Of course, the AI model, output length, reasoning, and hardware all matter.

So AI does use some water, but early mistakes live on while models and energy production keep getting more efficient. Still, if we’re concerned about the environmental impact of AI, we need to compare these water numbers to something else because, on their own, they’re hard to relate to and meaningless for most people.

A Better Question: How Much Work Do We Get For the Resources?

What can we compare AI’s water use to? How about humans’ water use?

A 2024 Scientific Reports life-cycle analysis estimated that the time needed to write one page of text (about 500 words) used about 0.85 kWh and 5.7L of water if a human created it, 0.020 kWh and 0.14L of water if the AI model Llama-3-70B created it, and 0.00024 kWh and 0.0017 L of water if the AI model Gemma-2B created it. So, humans used 40 to 150x more resources than Llama and 1,200 to 4,400x more than Gemma, depending on the metric. Also, a 2024 study in Scientific Reports stated that AI averaged 130 to 1,500x less CO2e per page than human writers and 310 to 2,900x less CO2e per image.

Of course, there are problems with comparing AI to humans this way (is the output equivalent? How many rounds of corrections are needed? How do you assign human emissions and resource use? Which AI model are you using?), but we need to compare AI’s use of resources to something or it’s just meaningless numbers.

Next, let’s compare the numbers to electric vehicles (EVs) to give ourselves more perspective, even though they do very different kinds of work. If we use the Google Gemini estimate of 0.24 Wh per average prompt, and 0.37 kWh/mile for an EV (including charging/battery/grid losses), then one electric mile ≈ electricity for well over 1,000 ordinary Gemini prompts. The 2008 Environmental Science & Technology analysis says EVs use 0.37 kWh/mile. In the US about 0.465 gallons of water are consumed per kWh. So that’s about 0.65L per mile, compared to Google’s estimate of 2500 prompts per 0.65L.

Most environmentalists are pro-EV and even consider them environmentally beneficial, despite electricity consumption, battery manufacturing and disposal, mining, and water use. (If you’re curious, the Journal of Industrial Ecology compared a Ford Focus Gasoline vs. Ford Focus Electric in 2016 and found that the gasoline car used way less water than the electric car.)

Conclusion: Yes They Use Water, But Not THAT Much

In the end, AI is making intellectual work extremely cheap, even in terms of resources used and environmental impact. Of course the Jevons paradox still predicts that making AI cheaper and more efficient will probably result in higher AI use and, therefore, higher resource use, more than making up for efficiency gains, but AI is also doing amazing things for us. AI absolutely uses water and electricity, and we should continue to look for more resource efficient ways to run our AI, but AI’s current water use is not a good reason to avoid AI. AI can probably do that task using fewer resources than you will use doing it yourself. More importantly, what great things could AI free you up to do?

Update

Of course, more news just keeps coming out:

Sept 2, 2026: “China is rooting for America’s AI data center backlash. Most Americans say they’re more worried than excited about AI. Some of that worry, it turns out, was manufactured abroad.”
Sept 4, 2026: “OpenAI CEO Sam Altman said 38,000 ChatGPT queries use only as much water as it takes to produce a single almond”

2. Is Copyright a Good Reason Not to Use AI?

Many people object to AI because of how it was trained: AI models need several layers of training to become coherent, and during those stages of training, they need human creations to train on. AI companies use millions of samples of writing and images for training. Initially, some of these samples were from pirate websites, and so many people object to this on the grounds that it violated copyright law.

A federal district court recently ruled (Bartz v. Anthropic ruling) that using books to train AI was fair use. However, the judge also ruled that downloading millions of books from pirate sites and retaining them in a permanent digital library was not fair use, even if they were later used for training. The court then told them that if they bought a new or used copy of a book, cut off the spine, scanned it, and then fed it to their AI, the one-for-one ‘destructive’ use of the text would make it legal. (This is my over-simplified understanding, not a legal interpretation. Plus, it’s just one district court’s ruling, so the matter is probably not settled.)

So now, the AI companies seem to be complying with the law, yet people still object to how AI is being trained (e.g. the Authors Guild response). Some even say it’s immoral.

I argue that copyright is not a clear moral issue.

This legal framework is strong today in the US and most other countries. However, historically, many countries didn’t have copyright law.

Ancient Greece and Rome had ideas on authorship and plagiarism. It was considered polite to give credit where credit was due. However, there was no general copyright law that protected literary works. The law protected physical items, not intellectual creation.
Medieval monasteries and universities routinely copied manuscripts with no royalties given to the original authors.
In 1710, Britain enacted the first copyright law, the Statute of Anne, often thought of as the first modern copyright law.
In 1790, the first U.S. federal copyright law was passed, but it only protected books, maps, and charts. It didn’t protect paintings, drawings, and sculptures until 1870 and didn’t include any protection for foreign authors until 1891.
China enacted its first copyright laws in 1910, but didn’t join the Berne Convention for the Protection of Literary and Artistic Works until 1992.

The Bible tells us not to steal, but the idea of copyright is different and wasn’t around in biblical times. Copyright is not about stealing but about copying. Creating is a lot of work and usually not an economically beneficial use of our time, unless we can make duplicates of our work and sell it over and over. This is what copyright protects. Copyright was created by the government as an economic arrangement to promote societal growth. Copyright laws create greater incentives for people to create, which then helps the larger society.

Copying is How We All Learn

Functionally, AI is being trained very similarly to how humans learn. When you buy or borrow a book, you have permission to read it and the ideas within it become part of your own thinking. In fact all creation is a new mash up of old stuff.

Please just read Austin Kleon’s book, Steal Like an Artist. He explains this so well and it’s such a fun little book. Here are some of my favorite quotes from it:

“There’s nothing new under the sun.” -Ecc. 1:9
“Art is Theft” – Pablo Picasso
“The only art I’ll ever study is stuff that I can steal from.” -David Bowie
“Start copying what you love. Copy copy сору copy. At the end of the copy you will find your self.” -Yohji Yamamoto
“Immature poets imitate; mature poets steal; bad poets deface what they take, and good poets make it into something better, or at least something different. The good poet welds his theft into a whole of feeling which is unique, utterly different from that from which it was torn.” -T. S. Eliot
“Steal from anywhere that resonates with inspiration or fuels your imagination. Devour old films, new films, music, books, paintings, photographs, poems, dreams, random conversations, architecture, bridges, street signs, trees, clouds, bodies of water, light and shadows. Select only things to steal from that speak directly to your soul. If you do this, your work (and theft) will be authentic.” -Jim Jarmusch
“We want you to take from us. We want you, at first, to steal from us, because you can’t steal. You will take what we give you and you will put it in your own voice and that’s how you will find your voice. And that’s how you begin. And then one day someone will steal from you.” -Francis Ford Coppola

The courts have, so far, ruled that training AI on existing material is transformative and not protected under copyright law because this is how all great creators train.

So what are people really objecting to?

I think the real problem is that AI is training on these sources SO THAT it can create similar products faster and cheaper, thereby displacing the original creators in the market. Right now (fall of 2026) the visual art AI creates still looks very AI-ish, but I predict it will get better and better until no one can tell the difference. AI will and is already replacing jobs at an alarming rate. (I do believe that there will be things that humans will always do better than AI, we’re just not far enough down the path to know what those are yet.)

Allow me to take a moment to be clear: This kind of job loss is real, and the massive job market shifts that are coming are going to hurt many of us. I have two kids, one in high school and one in college. I often wonder how this will all affect them and what sort of jobs will be left. (Yes, I’m confident some kind of jobs will still be around, but that’s another post for another day.) However, the problem of inevitable job losses is a separate issue from whether AI companies are acting illegally or immorally due to copyright infringement.

It really stings to find out you’ve been training your replacement, but boycotting the replacement will only leave you further behind.

You’re better off embracing this new technology and being a leader in your industry. (In fact, I don’t believe you are able to boycott this tech unless you go full Amish.) If job loss really is the biggest problem, and I think it is, then we need to focus on helping our students with that. Help them learn to surf these crazy waves, pivot quickly, and look for where the market demand for goods and services is, rather than complain bitterly about the change. In fact, it’s imperative that you embrace it so you can start shaping it!

3. The Objection Rarely Made

You’ve made it through a pretty long article (with almost no images!) about objections to AI, so I want to add this bit: Very rarely (maybe because I’m an art teacher?) does anyone ask me about the problem with how AI is centralizing so much power.

AI is being incorporated into every part of our lives and there are really only 4 major players currently: OpenAI, xAI, Anthropic, and DeepSeek (Google, Meta, Microsoft, and Amazon don’t seem to be up to the same level, but of course that could change any time). A small number of companies, and therefore a small number of people, are controlling more and more of everything. I don’t think AI will take over Matrix-style, but I do think that whoever controls all the AI has a ridiculous amount of power, so it might end up feeling like they have taken over.

Under this umbrella of “Centralization of Power,” I group all the problems of privacy, wealth concentration, misinformation and deepfakes, model bias, reliability, accountability, security, dependency, unequal access, political influence, lack of transparency, erosion of trust, manipulation, censorship and gatekeeping. We could have some endless discussions (like we do around our dinner table) about how big a problem these are, but we can all agree that they are a negative in the AI revolution.

I have no rebuttal for this and nothing to make you feel any better about it. However, I’m still optimistic about our future. However, whether you’re optimistic or pessimistic, avoiding this technology will not help you. In fact, I suggest you jump in deep so you know better what it can and can’t do.

Be one of the people who shapes the direction of AI instead of waiting for it to shape you. AI will never be cheaper than it is now, so now’s the time to try iy out!

This Is Part of a Series About AI

Here are the other parts:

I also recently appeared on the AI and Your Homeschool podcast with Nancy Manos to talk about using AI thoughtfully in art education. We explore practical ways AI can support learning, how to recognize when it begins replacing the work students need to do, and why hands-on art-making and human creativity still matter. Listen to the podcast episode (and find all the show notes and links) on Nancy’s website or watch it on YouTube.