1. Electricity Consumption
If GenAI was a country, it would be fifth place on the global list of energy consumption, between Japan and Russia.
This level of energy consumption cannot be met with green energy.
A single model, like OpenAI's GPT-3 requires ~1,287 MWh of electricity to train.
2. Water Consumption
A single prompt requires around 0.5 - 1 liter chilled water for cooling of data centers.
The global AI demand in 2027 is projected to account for 30-47 million people's yearly water consumption, just to cool the data centers.
3. Toxic Chemicals for GPU
Thousands of GPUs are needed to run the AI models.
The production of these processors is incredibly harmful for the environment, especially due to the sourcing of materials.
4. Digital Sweatshops
Think AI knows how to categorize information instinctually? Think again.
Human workers are needed to sort through the training material. They label, tag, transcribe and process data.
Due to the high demand, this work is increasingly being outsourced to workers in the Global South.
These humans often work in what is known as "digital sweatshops". Here, they consistently report poor working conditions, exploitation, and forms of psychological distress.
A working day in these locations can often last up to 20 hours, during which the workers will shift through approximately 1000 cases.
Understanding these data sets requires human, contextual understanding. This means humans sometimes interact with toxic, graphic and hateful content under already distressful and exploitative working conditions.
5. Making the Rich Richer and the Poor Poorer
The godfather of AI, Geoffrey Hinton, stated that "AI will make a few people much richer, but most people poorer" [6]
The impact of AI on the global income distribution is highly uncertain. However, most predictions suggest it will increase inequality, both within-countries and between-countries.
It is predicted that AI will benefit highly skilled workers, and replace low-skilled jobs and repetitive tasks. Wealth will be concentrated among those who control the technology.
Catch-up growth of developing countries is threatened by AI. Because developing countries may not be able harness AI's benefits or mitigate its disruptions, they may again fall far behind.
sources