Emma Strubell
AI Researcher
Carnegie Mellon computer scientist whose 2019 study on the carbon cost of training language models forced the AI field to confront its energy use — finding one large model could emit roughly five times a car’s lifetime emissions.
In her words
These accuracy improvements depend on... exceptionally large computational resources that necessitate similarly substantial energy consumption.
Biography
Emma Strubell is an assistant professor in the Language Technologies Institute at Carnegie Mellon University and a natural-language-processing researcher. Unusually, one of her most influential contributions was a warning about her own field’s footprint rather than a new model.
Counting the carbon
Her 2019 paper “Energy and Policy Considerations for Deep Learning in NLP,” with Ananya Ganesh and Andrew McCallum, was the first to put concrete carbon numbers on large-model training. The headline finding — that training a single large model with an extensive architecture search could emit over 626,000 pounds of CO2, roughly five times the lifetime emissions of an average American car including its manufacture — travelled far beyond academia and reframed “bigger is better” as an environmental as well as a technical question.
Strubell has continued to press the field to measure and disclose its energy and hardware costs, from the embodied carbon of chips to the growing e-waste of rapidly obsolescent accelerators. Like Alex de Vries and Kate Crawford, she insists the compute-hungry trajectory of modern AI carries a physical price the industry prefers to leave off the balance sheet.
Where they stand in the war
Who backs them up
Sources & further reading
Canonical record: https://battlelines.ai/topic/emma-strubell