����Vlog

Generative artificial intelligence use is now “near universal” among undergraduate students, studies show, with universities increasingly embedding these digital tools across their curricula and assessments.

But what could this mean for the well-publicised promises made by thousands of universities worldwide to reduce their carbon emissions and reach net zero?

While the concerns over GenAI’s impact on academic integrity, learning, assessment and employment have been loudly debated across the sector, few questions have been asked about the environmental cost of spiralling GenAI use.

The training of GPT-4, for instance, used the same amount of energy as the domestic electricity consumption of more than 460,000 people in sub-Saharan Africa, according to . This means that offsetting GPT-4’s training carbon footprint of 25,000 tonnes of CO₂e would require 420,000 tree seedlings to be grown for 10 years. Then there is the energy cost associated with its ongoing use and the huge data centres required to run these models.

On this week’s News Talks podcast Miranda Prynne speaks to THE reporter Juliette Rowsell, who has been investigating whether universities’ embrace of GenAI is undermining their net-zero targets, to discuss how universities’ relationship with GenAI has evolved in the four years since its mass release, what this could mean for institutional carbon emissions and the challenges associated with trying to accurately measure GenAI impact.

They attempt to tackle the question of what universities can or should be doing to improve reporting of emissions associated with GenAI and, ultimately, drive more sustainable use of these tools.