The Other Side of the Coin

I've been thinking about the ethics of AI. A lot.?

As a linguist, I’d like to start this article by looking at the matter from a language point of view.?

There is no hotter topic on this very platform - or anywhere really - today.?

And there is no better adjective than “hot” to describe it.? “Hot” it is indeed, in more ways than one.?

“Hot” as in, a subject widely discussed in the media of all types and genres. (“A hot topic” is the perfect idiom here.)

“Hot” as in, the most anticipated and promoted technology of today. (Idiom: “To be hot on something”)

“Hot” as in, every industry feels the need/pressure to implement AI for fear of being left behind (Idiom: “Hot and bothered”)

“Hot” as in, the fastest-growing technology sector based on the venture capital being poured into large AI tech companies and the number of startups wanting to join the party. (Idiom: “Strike the iron while it's hot”)

“Hot” as in, the most promising type of stock for AI tech founders, investors/shareholders and Wall Street in general. (Idiom: “Get rich quick”. Ah, I couldn't help it. Although adding “hot” might enhance the effect.)

For all the buzz, those meanings are figurative.?

It's the literal one that we don't hear much about.?

AI is literally ?? because it requires tremendous amounts of clean, potable water to cool the servers.?

It is ?? because it uses tremendous amounts of energy to train and to use.?

Let me take off my language hat for a minute and give you some numbers:?

• ChatGPT's daily power usage is nearly equal to 180,000 U.S. households, each using about 29 kWh per day. (https://t.ly/pjURy)
• Estimates of the number of cloud data centers worldwide range from around 9,000 to nearly 11,000. More are under construction. The International Energy Agency (IEA) projects that data centers’ electricity consumption in 2026 will be double that of 2022 — 1,000 terawatts, roughly equivalent to Japan’s current total consumption. (https://e360.yale.edu/features/artificial-intelligence-climate-energy-emissions)

• Training GPT-3 in Microsoft's state-of-the-art U.S. data centers can directly evaporate 700,000 liters of clean freshwater. (https://arxiv.org/abs/2304.03271)

• The global AI demand may require 4.2 – 6.6 billion cubic meters of water withdrawal in 2027, which is more than the total annual water withdrawal of 4 – 6 Denmark or half of the United Kingdom (https://oecd.ai/en/wonk/how-much-water-does-ai-consume)

• Training LLMs like GPT-3 consumed an amount of electricity equivalent to 500 metric tons of carbon, which amounts to 1.1 million pounds. A typical coal-fueled power plant working continuously for 24 hours burns about 2.7 million pounds of coal. Training LLMs is therefore equivalent to burning coal for 10 straight hours, or nearly half a day. Importantly, these models are not trained only once: LLMs need to be systematically trained on the most up-to-date data not only to remain relevant and accurate but also to deliver improvements over time. (https://arxiv.org/abs/2104.10350)

Let’s bring these numbers closer to home. Our routine usage of AI per session (10-50 prompts) is worth:?

• 500 ml of potable water = one bottle
• 0.01 kWh of energy = 5 ceiling fans running for an hour
• 45 - 2,025 g of CO2 = up to half a liter of gasoline = driving an average of 2.3 miles

Besides energy and water consumption and CO2 emissions, there are other implications of the AI proliferation:

• Mining for cobalt and lithium has significant health and safety ramifications for the local communities, as well as severe environmental consequences in Africa and South America, where much of the extraction takes place.(https://earth.org/cobalt-mining-in-congo/)
• Ecosystem and habitat destruction as well as deforestation and displacement of local populations due to lithium mining and construction of new data centers. (https://t.ly/VIoTM), (https://t.ly/hghKF)?

Sounds like Armageddon, right? But I don't want to only appear hot under the collar about it. As this article title suggests, there are two sides to this coin.?

So I'll give you some more numbers:

• AI can help mitigate 35% of aviation’s climate impact by analyzing weather, satellite and flight data to predict flight paths resulting in fewer emissions.? (https://t.ly/3VAAk)
• By scaling currently proven applications and technology, AI has the potential to unlock insights that could help mitigate 5% to 10% of global greenhouse emissions by 2030. (https://t.ly/BWiHs)?????

• Deepmind AI, a machine learning platform owned by Google, can predict wind patterns up to 36 hours in advance so that wind farm energy production can be optimized. (https://earthly.org/en-US/how-ai-and-big-data-can-help-reduce-the-impact-of-climate-change)
• A 2019 study of AI-based solutions for agriculture showed a 53.6% reduction in pesticide use, and a pest identification rate of 90%. (https://link.springer.com/article/10.1007/s10311-023-01617-y/tables/3)
• AI can be applied to various aspects of waste management, such as waste-to-energy, waste sorting, waste generation models, plastic pyrolysis, logistics, disposal, and resource recovery. By implementing artificial intelligence in waste logistics, transportation distance can be reduced by up to 36.8%, cost savings by up to 13.35%, and time savings by up to 28.22%. Artificial intelligence can accurately identify and sort waste with 72.8–99.95% accuracy (Fang et al. 2023) (https://link.springer.com/article/10.1007/s10311-023-01604-3)

These numbers tell us what tremendous power and impact lie in this technology, and that a responsible approach is not optional, it is mandatory.?

Just as we do not allow our kids to drive without first completing a driver’s course and obtaining a driver’s license, it would be foolish to unleash this force without first implementing safety guardrails.??

Around the world, efforts are in place to regulate this hot mess. At least 37 countries have proposed legal frameworks for AI safety and governance.

In the US, a bill “Artificial Intelligence Environmental Impacts Act of 2024” has been proposed to establish a voluntary reporting system for environmental impacts of AI. (https://www.congress.gov/bill/118th-congress/senate-bill/3732/text#:~:text=To%20require%20the%20Administrator%20of,the%20Director%20to%20develop%20a)

In March 2024 The European Parliament approved the Artificial Intelligence Act that “ensures safety and compliance with fundamental rights, while boosting innovation.” (https://www.europarl.europa.eu/news/en/press-room/20240308IPR19015/artificial-intelligence-act-meps-adopt-landmark-law)

ISO (the International Organization for Standardization) has published another voluntary set of requirements for “establishing, implementing, maintaining and continually improving an AI management system within the context of an organization.” (https://www.iso.org/standard/81230.html)

ISO is also developing a standard to inform the public of the environmental sustainability aspects of AI. (https://www.iso.org/standard/86177.html)

These are all great initiatives, no doubt.

But let's not leave the task of regulation and mitigation to governments alone, lest we end up feeling controlled and powerless.?

Providing information to the public and asking organizations to voluntarily adhere to ethical norms is a great step forward, but let's not yet again - remember the push for recycling? - put the onus of responsibility on the user alone, lest we end up feeling controlled and powerless.?

Unless everyone - users, legislators, investors, all the stakeholders of all organizations and companies - make it a personal responsibility, there is a risk for us of ending up running on the same familiar track of extracting the most benefit at the least effort and cost to self. That is the way our brains are wired. But we are not slaves to our brains and our habits. To bring about not only a brave, but a safe new world, we must act conscientiously and responsibly. We must form new habits. And why not start now, at this pivotal point in our human existence?

Individual users would do well to take into account the heads and tails of this coin.

Next time you automatically reach for ChatGPT/Claude/Teal, etc. to help you write that post or resume or profile update, try to use your natural intelligence first, knowing that you’re reducing your environmental impact and training your own neural pathways.?

Corporations and businesses of all sizes would do well to take into account the heads and tails of this coin.?

Next time you promise a reduction in costs to your clients thanks to the use of AI, please remember the hidden cost to the planet: to its vulnerable populations, ecosystems of plant life and wildlife, our air quality and the finiteness of fresh water.

And think about how you may direct at least a portion of those savings and profits towards offsetting the hidden damage: from educating yourself and your clients about the full picture, to participating in environmental initiatives, supporting the impacted communities, (including those whose jobs and livelihoods have been affected by the proliferation of AI), choosing to orient your AI business towards the positive impact on climate and global well-being.?

Or even simply helping to spread the word that the other side of the coin?also exists. Don't make a hot potato out of it.?

If, in our pursuit of a pretty penny, we forget to flip it to see both sides, WE will become the vulnerable population, landing ourselves in hot water, and all the pennies in the world will not buy us back our lost planet. Or humanity.?