Post # 81
August 7, 2024
Claire Bodanis
Our AI/reporting research project with Insig AI is well under way, and the insights from companies are fascinating in many ways. However, one subject that seems conspicuously absent from internal corporate debates about adopting generative AI technologies is energy…
I must confess that, because of the holidays, today’s blog isn’t quite as ‘today’ as usual, having been written a week in advance. And the reason I must confess to this (aside from my commitment to truth in all things), is that today’s blog was inspired by my lunchtime walk on Hampstead Heath – something I try to do, rain or shine, whenever I’m working from home.
And today is definitely shine. So much so that pausing at the top of Parliament Hill to see the whole of London spread out before me was not the relaxing, uplifting experience it usually is. Rather, since it’s so exposed up there, I couldn’t wait to get down and into the relative cool of the trees at the bottom of the hill. And of course by next week, the British summer being what it is, we may be looking at digging out our jumpers and brollies again, so this blog may feel a little discordant.
Or will it? Today is pretty boiling, although not as hot as last week when, last Monday, 22nd July, was declared the world’s hottest day on record. But the news had already broken on Sunday 21st July, because, until Monday came around, Sunday was the world’s hottest day on record. And Tuesday 23rd was almost as bad. Then on Wednesday 24th, the FT’s Big Read on ‘The dangerous effects of rising sea temperatures’ reported how shocked the scientific community is about how high global sea surface temperatures have reached, and remained, over the past 15 months. Who knows what next week will bring.
The urgency of cutting emissions has never been so great, and it’s a message not lost on companies. Barely a results statement or annual report goes by without a discussion of climate change and what that company is doing about cutting energy consumption and carbon emissions in pursuit of their science-based, net zero targets. At the same time, barely a results statement or annual report goes by without at least some comment about generative AI, and the transformative power this technology will have for the future benefit of the company in question. And, of course, its stakeholders.
But I am yet to find a report that properly joins the dots between the two, and a company including in its plans for adopting and rolling out generative AI a mitigation strategy for the massive uptick in energy costs and carbon emissions that doing so will involve. Back in February, the respected scientific publication Nature published an article, ‘Generative AI’s environmental costs are soaring – and mostly secret’. Amongst other alarming statistics, the article noted that ‘It’s estimated that a search driven by generative AI uses four to five times the energy of a conventional web search. Within years, large AI systems are likely to need as much energy as entire nations.’
It may be that companies aren’t joining the dots because the energy used to power their gen AI experiments is being borne by the provider (Microsoft, Google and so on). Even if that is so, at some point those costs are likely to be passed on. And regardless of who’s paying, the carbon emissions generated are going to sit somewhere in the company’s footprint, affecting its net zero targets. Because, while I’m hopeful that renewable, low-emissions energy will speedily become the norm, I can’t see it happening soon enough to power the AI transformation. More broadly, who’s responsible for the environmental and social burden of creating all that electricity in the first place – provider, user, or both?
This lack of dot-joining may also be because, as our research so far shows, many companies are in an experimental phase with generative AI, so haven’t done their full analysis and planning yet. Or perhaps the people we’ve been talking to haven’t been involved in the wider capex/impact discussion about its adoption. But if the way companies are discussing generative AI so far in their reports is a truthful reflection of internal discussions – at least in the ones I’ve read to date – then the potential conflict between their gen AI ambitions and carbon emissions targets is simply not front of mind.
This reminds me of the early days of digital reporting, when the PDF started to replace printed annual reports, and shareholders started asking questions at AGMs about the environmental credentials of the print production process. Indeed, are there any printed FTSE reports today that don’t have the Forest Stewardship Council logo on their inside back cover? I haven’t checked, but I doubt it.
It's not my usual practice to suggest yet another burdensome requirement for reporting, but perhaps raising the question about broadening a report’s environmental production credentials to encompass the inevitable use of generative AI might push this subject up the agenda.
After all, with my ‘tech optimist’ hat on (yes I do have one!), I’m rather hoping that the use of gen AI itself within companies to interrogate and understand their own data and information will not only bring the contradiction between gen AI adoption and net zero targets to people’s attention, but perhaps even pose some solutions. And in the meantime, I’ll be adding the subject to our AI reporting research discussions – so that, whenever we’re talking about the use of generative AI in the reporting process, we don’t forget that digital reporting has an environmental footprint too.
August is my (if not everyone’s!) report-free month, so with holidays in sight, our next blog will be the first Wednesday of October. I hope you all have a lovely summer break.