Artificial Intelligence (AI) technology is booming, but it’s still not clear what its net impacts will be in energy/carbon terms- views differ. However, it is clear that there is a growing drive to use it more and more- with big profits to be made by whoever powers the growth. Not surprising then that all the usual suspects have turned up for a share in the proceeds – most obviously power plants powered by fossil gas, but also nuclear and renewable energy.
AI data centres are not exactly popular, since, ideally, they need to be near centres of population, where data processing demand is large. Although AC power can transmitted over long distances (a few hundred miles) without too much energy loss, digital data info is more sensitive - there are latency (delay) effects over long distance. And rather than the use local mains grid power, it’s common for AI centres to set up their own local dedicated power units.
Such is the scale of the AI boom that all sorts of players are trying to get in to the act - for example new gas turbine plants, local solar farms (in the US especially), but also, more speculatively, mini nuclear units - Small Modular Reactors (SMRs). Though none of those exist yet anywhere. All sorts of locations too are trying their luck with AI, especially poor areas, maybe hoping that AI will aid local economic recovery. Coal-fired AI may be too much of a stretch, though fluidised bed coal fired gasification plants have been mooted, but if AI /data centres are combined with new renewables, or even, some say, SMRs when and if they are proven, that might revitalise areas blighted by fossil extraction- e.g. in the USA in old coal fields in Kentucky and elsewhere, or areas with oil or gas wells as in California.
However, it may be a hard fight. Data centres are very unpopular in much of the USA. And interestingly, as Grist has noted, ‘in an era when political polarization is near record highs, data centre backlash represents a rare area of consensus on both sides of the political aisle. Some 70 percent of Americans oppose local construction of AI data centers - 75 percent of Democrats and 63 percent of Republicans, according to polling from Gallup’. Moreover, it seems there are ‘more conservative Republicans (53 percent) who oppose data centers in their local area than moderate Republicans (44 percent) - meaning that staunch conservatives are actually nearer to Democrats in their opposition.’ However, the emphasis of opposition may be different. Grist notes that ‘Republican officials often raise concerns about tax incentives and energy grid strain, while Democrats tend to focus on environmental impacts and resource consumption,’ according to a Data Center Watch report.
Whatever the reasons, local hostility and local constraints, may lead to other locations being considered, including putting the AI processing technology offshore, and possibly under water since that also avoids the need for energy for cooling. Floating PV arrays could also be used to provide power. Or, much more wildly, there have been suggestion for putting them in low earth orbit or even on or in orbit around the moon using solar energy collected locally! Latency delays are obviously an issue there, but not for all types of data storage, so hybrid systems may emerge.
That sort of thing is for the future. For the present, the emphasis is on land based projects around the world, with some with differing approaches emerging as to how to deal with the energy demand/emissions issues. For example, the UK is looking to prioritize flexible AI, whereas, in Denmark, AI of any sort will not get priority access to the grid. Clearly though it is still often a battle, given the social and environmental impacts that AI can have in some locations.
Not surprisingly, then, there is lot of debate and coverage of AI in the media and in academia, as I noted in an earlier post, with, for example, MIT recently adding their input. And the paper that Terry Cook and wrote is now fully published in a special AI issue of the Oxford University Press journal Clean Energy. In it we say that ‘AI and green energy are developing rapidly, with a synergy possibly emerging. AI needs green energy, if climate issues are to be avoided, and, arguably, green energy needs AI to aid its efficient development,’ and we look at what this interaction may means for the future, focussing on employment and skill related issues. See our summary chart below.
However, we don’t ignore the potential down sides, with, for example, some seeing AI as further accelerating unsustainable growth. There certainly are big issues. And we call for careful technology assessment and the full adoption of effective regulator controls. Is that possible? Will it be enough? The debate continues, with, for example, worrying new data emerging on net energy impacts and some quite wild speculation about the global AI future and the need for a temporary moratorium on AI. This one will run and run…
Meanwhile though, after this short post, I am going back off line on holiday for while again now
Comments
Post a Comment