04:05 Issue 14

04:05 GLOBAL

Analysis” - TIME, 2024 “UN: AI May Become One of the Largest Energy Consumers Globally” - UN Regional Info Centre, 2024 II. Explosive Carbon Footprint Training large AI models emits as much CO₂ as driving over 100 cars for a full year. GPT-3, for instance, required 1,287 MWh of electricity, producing 502 metric tons of CO₂ during its training. Sources: “AI’s Massive Carbon Cost” - State of the Planet, Columbia University “U.S. AI Data Centers Could Soon Consume More Electricity than Poland” - Business Insider, 2024 III. Massive Water Usage Cooling AI data centers consumes enormous amounts of water. A Microsoft facility reportedly used 6% of a district’s water supply in just one month. Projections suggest AI systems may use between 4.2–6.6 billion m³ of water per year by 2027. Even a single 100-word ChatGPT interaction may require 500 ml to 1 liter of water. Sources: “Water Usage in AI Data Centers” - The Guardian, 2024 “By 2027, AI Could Use More Water Than Entire Nations” - Greenly, Splunk, arXiv, 2024

digital dependency syndrome, where humans rely so much on external tools that internal faculties begin to weaken. Risks include: Shortened attention spans Weakened memory recall Reduced problem-solving skills Loss of critical thinking and analytical depth Sources: “Can AI Make You Dumber?” - Scientific American, 2024 “Heavy Reliance on AI May Be Weakening Our Memory and Attention Spans” - ZDNet, 2023 4. The Environmental Cost: Hidden But Massive I. Soaring Energy Consumption Each AI prompt activates complex neural networks across multiple servers. According to the International Energy Agency (IEA), the energy consumption of data centers is expected to double by 2026, primarily due to AI workloads. A single ChatGPT query uses about 0.0029 kWh, around 10 times more than a Google search. That’s roughly 68 grams of CO₂ per query. Sources: “OpenAI GPT Carbon Footprint

68 I 04:05

GLOBAL PAYROLL MAGAZINE ISSUE 14

Made with FlippingBook - Share PDF online