Government policy and the data centre industry are failing to differentiate between “good AI” and “bad AI”, the Oireachtas committee on artificial intelligence has heard.
“The problem is that people treat AI as one homogeneous thing,” Andrew Parnell, professor of data science for weather and climate at University College Dublin told the committee.
“Some (AI models) benefit society. They should be encouraged and kept separate,” he said.
One of them was in use by Met Éireann and other forecasters and enabled huge improvements in weather forecasting, Parnell said.
READ MORE
“Perhaps the most remarkable statistic about this new AI method is that it is estimated to be 1,000 times more efficient at producing a weather forecast than our previous approaches.
“Weather forecasts can be obtained in a fraction of a second even on a standard laptop that you or I might have on our desks.”
He said this allowed forecasting to be carried out with less energy and cost than previously, meaning critical information was not a luxury only rich countries could afford.
However, he stressed the AI used for this purpose drew from data sets and used computing capacity that were much smaller than the large language models (LLM) being developed and used by AI to replicate human language and activity.
[ Data centre electricity demand could exceed entire power usage of Irish RepublicOpens in new window ]
“Even though we take the data from the whole world every hour, it’s tiny compared to LLM and the computing system is tiny too,” he said.
Parnell was speaking as the committee discussed the environmental impacts of AI and the growth in data centres needed to facilitate its rapid expansion.
Data centres already consume more than a fifth of Ireland’s electricity, due to rise to almost a third in the next few years, and a recently agreed Government policy would facilitate growth that would match all of the country’s electricity use by 2040.
Much of that extra power would need to be provided by fossil fuels, increasing greenhouse gas emissions and leaving the country further from reaching already very challenging climate targets.
Fianna Fáil TD Malcolm Byrne, a committee member, said data centres were vital and questioned if there was any alternative.
Hannah Daly, professor of sustainable energy at University College Cork, said while AI and digital infrastructure were part of everyday life, it did not mean they should not face restrictions.
[ Government departments are using AI, but one clear red line has been laid downOpens in new window ]
“AI is a tool. It’s not universally good or bad. We need a transport system but that doesn’t mean there should be unregulated growth of motorways or private jets.”
Jennie Stephens, professor of climate justice at Maynooth University, said there was an industry narrative that had “captured governments”.
“It’s the inevitability narrative – that the train has left the station – or that if you’re against it (AI), you’re against progress,” she said.
