with AI models. If there is new data to be trained on, they are going to want it.” In the end he said, “This might lead to the siloing of data, keeping things off grid, so it can’t be scraped or stolen. It protects your IP but goes against what the internet was meant to be. It would be sad to go back to a world where information becomes gated again. But smaller, medium language models might become more prevalent in the future where people create these models based entirely on their own verifiable data in a closed system. I can imagine that being useful for libraries – it could make the data libraries own very powerful.”
Transparency
He is similarly sceptical about the lack of transparency from big tech, especially around its environmental impact: “When it comes to the biggest players in the world, like Open AI or Microsoft, or Apple, they are usually very cagey about the environmental impact of what they do because it’s pretty bad PR.” He said they provide details on their environmental offsetting strategies but less about what it is they’re offsetting: “They talk a lot about mitigation without really going into the damage AI is causing to the environment.”
He said that Microsoft carbon emissions had increased by 30 per cent since 2020 and Google’s by 48 per cent since 2019 and that Goldman Sachs believes data centres will use eight per cent of US total power by 2030, up from three per cent in 2023.
And while Microsoft is one of the biggest purchasers of carbon credits – 83 million tonnes worth – he pointed out there were concerns about whether carbon offsetting worked practically and ethically. However, his main concern was that the other side of the equation – how the type and scale of environmental damage being done is being identified and measured – was missing. IP approached Microsoft for comment but has not received a response. “If we don’t know the harm AI is doing we can’t possibly know if these compa- nies’ efforts are meaningful or simply performative,” he said. “What I would like to see is the full supply chain environ- mental impact of AI – from data centre construction, data scraping and training and staffing.”
He said this also needed to take into account the impact of AI use by end- users – and the by-products of marketing efforts by AI organisations, like the use AI recreationally, which he thinks wouldn’t happen if it wasn’t marketed so heavily, or the impact was understood.
But also whether the carbon footprint of AI should include the mitigation forced on other organisations – like libraries – when
38 INFORMATION PROFESSIONAL
www.scottishai.com
they have to defend their systems and stop them being swamped with AI bots.
Metaphorical communication One challenge is maintaining public atten- tion. “The problem is that for a long time, we’ve spoken about environmental impact in terms of metaphors. We’ve likened it to X number of homes, or cars, or football fields of garbage. But I don’t understand
the significance of a tonne of carbon, or what it means to make a car. Unless I’m well versed in the environmental impact and the atmospheric impact of carbon, I won’t know the difference between a tonne of carbon or a million tonnes of carbon.”
“We do talk about environmental impact in the strategy and it’s a big part of our literacy efforts,” he said, adding “We just finished a project with the Glasgow School of Art about how to visualise the environ- mental impact of AI.”
But having a real feel for the impact of Chat GPT 3 is important because ChatGPT 4 is maybe 10,000 times more complex than Chat GPT 3 – making it potentially many thousands of times more damaging to the environment. Gordon doubts human brains deal well with big one-off shocks like this and that realistic options might be something like a real-time carbon footprint gauge: “A while ago there was a tool that tallied up your carbon usage as you spoke to Chat GPT. I think it was very preliminary research, but it showed that if people could see the car- bon impact they were having, they were less likely to use it for extended periods of time. So that could be a way of mitigat-
Autumn 2025
Page 1 |
Page 2 |
Page 3 |
Page 4 |
Page 5 |
Page 6 |
Page 7 |
Page 8 |
Page 9 |
Page 10 |
Page 11 |
Page 12 |
Page 13 |
Page 14 |
Page 15 |
Page 16 |
Page 17 |
Page 18 |
Page 19 |
Page 20 |
Page 21 |
Page 22 |
Page 23 |
Page 24 |
Page 25 |
Page 26 |
Page 27 |
Page 28 |
Page 29 |
Page 30 |
Page 31 |
Page 32 |
Page 33 |
Page 34 |
Page 35 |
Page 36 |
Page 37 |
Page 38 |
Page 39 |
Page 40 |
Page 41 |
Page 42 |
Page 43 |
Page 44 |
Page 45 |
Page 46 |
Page 47 |
Page 48 |
Page 49 |
Page 50 |
Page 51 |
Page 52 |
Page 53 |
Page 54 |
Page 55 |
Page 56 |
Page 57 |
Page 58 |
Page 59 |
Page 60 |
Page 61 |
Page 62 |
Page 63 |
Page 64 |
Page 65 |
Page 66 |
Page 67 |
Page 68 |
Page 69 |
Page 70
