

is using AI worse than driving a car?
Jun 4, 2025
Shaolei Ren, an Associate Professor of Electrical and Computer Engineering at UC Riverside, examines the hidden costs of AI. They discuss the environmental footprint of AI compared to conventional transport, highlighting issues like energy consumption and air pollution. Ren emphasizes the overlooked water usage in data centers, particularly in water-scarce areas. The conversation calls for transparency among tech companies and better public awareness of AI's sustainability challenges, urging a reevaluation of its role in our lives.
AI Snips
Chapters
Transcript
Episode notes
Dexter's Car-Free AI Experiment
- Dexter Thomas has lived without a car in LA for a decade, relying on buses, bikes, and Ubers despite inconvenience.
- He used AI to compare his transit impact to car ownership, uncovering unexpected environmental considerations.
AI Query Energy Use Explained
- Each interaction with a large language model like ChatGPT consumes about 3 watt-hours, more than a standard Google search.
- Even small per-use energy costs add up massively at the scale of billions of daily interactions.
AI Training's High Pollution Cost
- Training one large AI model like Meta's LLaMA 3.1 generates air pollution equivalent to over 10,000 round trips by car from LA to NYC.
- The extensive one-time environmental hit from training dwarfs the energy use of typical model queries.