Environmental impact of AI

❝ The rise of the robots is a problem not just because of the immediate human cost, but because robots are fantastically energy-hungry and thus accelerate the unsustainable, insupportable damage we are doing to Earth. – Rupert Read

, past spokesperson of Extinction Rebellion

Immediate harms are turning intense.

Corporations feed our friends and family’s data into AI. Corporations refine AI to cheaply replace our long-time colleagues. Corporations sell bots used to manipulate and target us.

It’s easy to miss AI’s most diffuse harm—to our environment.

Burning energy

Training just one AI model can gobble up more electricity than 100 US homes

use in one year – just to learn how to predict the next sentences from our online conversations. To generate those next sentences in replies to users, ChatGPT burns as much electricity as a small city

By 2027, all AI may burn as much energy as a nation

, just to process our data in data centers.

Data centers that suck billions of liters

of drinking water, to cool AI’s overheating hardware. Data centers that emit billions of tons
of carbon dioxide, warming the atmosphere.

And yet, data centers are only one stage of AI’s lifecycle.

Hidden toxicity

Today’s AI is not “alive,” of course.

AI doesn’t look yet like the free-roaming robotics we see in movies. AI is stuck running in data centers, on hardware that wears down and gets replaced every ~4 years.

But producing that hardware takes hundreds

of cancerous chemicals
. Materials needed were melted down and reassembled at blazing hot temperatures. Those materials were in turn extracted from mines, using drills and dissolving acids.

Environmental scientists investigate all these stages. They call it the ‘hardware lifecycle’ of AI.

We don’t see the mines

, refineries
, production plants
, data centers
, and waste dumps
. We don’t see the bulldozed lands, the hellish infernos, the toxic chemicals, the climate gasses.

AI is toxic to our environment. Corporations hide the toxicity, out of our sight.

This boom is different

Fortunately, climate activists woke up to the fact that AI burns an astounding amount of energy. States are forced to keep coal power plants

running to cater to AI data centers. Even OpenAI’s CEO admits: “we still don’t appreciate the energy needs of this technology.”

Climate activists are quicker on their feet this time. Years before, another tech boom began: Cryptocurrency. Hardware slurped 0.5%

of US energy to compute crypto. It got so bad that Texas gave millions in tax dollars
to get a Bitcoin company to stop draining so much electricity.

But here is the crucial difference:

  • Cryptocurrencies go bust. They produce little value, other than as alternative money.
  • AI can be used to automate economic production.

AI is an ideal tool to centralize power. AI to extract, to exploit, to extract.

A vicious cycle

  • Corporations extract profit using untested bots
  • To reinvest in more toxic hardware factories
  • To build more energy-slurping data centers
  • To feed more of our free data into AI
  • To automate more work cheaply
  • To extract more profit

This is a vicious cycle.

After two centuries of competition for scaling resource-intensive machines, society is already nearing collapse

. To protect society, why don’t we ease up on excessive corporate tech?

AI corporations accelerate our climate crisis. They scale machine models to predict human behavior. They gradually automate the work needed to re-produce resource-intensive machines