Abracadabra! AI is not magic.

It's Water, Watts, and a whole lot of Worry.

Let me paint you a picture. You sit at your desk. You ask ChatGPT what the capital of Bolivia is. Easy. Fast. Magic.

Except it isn't.

That question you just asked likely consumed three to five times the energy of a traditional web search. Some estimates put the water usage for training a single large AI model at over 700,000 litres; the equivalent of 370,000 bottles of Evian. One query alone can use up to 500ml of water to cool the servers. Ask two questions and you’ve basically bathed in AI without even unwrapping your bath bomb.

We tend to think of AI as floating in the cloud, weightless and invisible. But the cloud has a footprint, and it stomps hard. Data centres draw power from grids still run largely on fossil fuels. And the bigger the model, the bigger the appetite. The scale is dizzying, and we’re only just getting started.

From productivity to paranoia

Now let’s turn to your office. Claudinea from Finance just used ChatGPT to help draft a report. Helpful? Definitely. But she also pasted in the company’s unreleased financials to get a better summary. That data? It might now be part of the AI's training set; or at the very least, stored somewhere far beyond your firewall.

This isn't theory; it has already happened. Samsung engineers last year accidentally leaked confidential code via ChatGPT. Once it’s out, it’s out.

It’s not that we are careless. It’s that we’re human. And AI is very easy to use badly if you’re not aware of the wider context of how it works and what it consumes.

The ethics of automation

Then there’s the deeper stuff. If AI writes your copy, decides who sees your ads, and summarises your reports, who’s actually communicating? If algorithms filter job applicants or provide legal advice, whose bias is baked into that black box?

We’re entering a world where the most important business decisions might be made not by people, but by a machine that was trained on Reddit posts from 2012. Don’t get me wrong, I like Reddit, but if I’d crowd-sourced my business strategy from there I very much doubt I’d be where I am today.

So what do we do?

We lean in; but we do it with our eyes open.

AI is an incredible tool. It’s saving time, boosting productivity, and opening creative doors we didn’t even know existed. But it’s not consequence-free.

That’s why at Dodgems and Floss, we’ve developed Neural Web Optimisation; a process that ensures your website is accessible, indexable, and understandable to AI tools. It’s how businesses will be found in the age of AI Search. But we do it responsibly.

We don’t just chase clicks or appease algorithms. We remember the humans; your customers, your team, your future. AI might start filling in more and more of the pages, but it’ll still be us, people, who will define the next chapter.

Let’s build with purpose, clarity and conscience. And maybe with a little less water?

___

🌍 Environmental Impact of AI

• Energy Consumption per Query: A single ChatGPT query consumes approximately 0.0029 kWh of energy, nearly 10 times more than a standard Google search, which uses about 0.0003 kWh.engineeringprompts.substack.com+3kanoppi.co+3reddit.com+3
• Water Usage for Model Training: Training OpenAI's GPT-3 model required about 700,000 liters of freshwater, equivalent to the annual water consumption of several households.en.wikipedia.org+2deepgram.com+2news.ucr.edu+2
• Water Usage per Query: Each interaction with ChatGPT can consume up to 500 milliliters of water, primarily due to the cooling needs of data centers. interestingengineering.com

🔐 Data Security and Ethical Concerns

• Samsung Data Leak Incident: In 2023, Samsung engineers inadvertently leaked sensitive internal source code by inputting it into ChatGPT, leading the company to ban the use of such AI tools among employees. forbes.com
• Ethical Implications: The integration of AI into decision-making processes raises concerns about bias, transparency, and accountability, especially when algorithms influence hiring, legal advice, or content moderation.