The hidden (eco) cost of AI

“Here I am, vacuum-packed in plastic as a Barbie / Action Man / Marvel hero*” (* delete as appropriate)

Yes, yet another of your social media business connections has jumped on the AI (artificial intelligence) bandwagon and recreated their image as a plastic figure, complete with packaging and accessories.  Just a bit of harmless fun, surely?  Is it a case of #FOMO (fear of missing out)?  No actual plastic was used in the process, so at least it did not cause harm to the environment.

Unfortunately, it turns out that there is an ecological impact, both in the AI training process (machine learning) and in the implementation; AI is incredibly resource hungry.  It is said that the AI industry uses roughly the same amount of power globally as air travel.  Fortunately for end-users, the energy consumption is not adding to their personal electricity bills, but this does mean that the impact is largely invisible.

What costs are involved, then?

Learning the ropes

AI doesn’t just spring into life, complete with the knowledge to assist in all your tasks.  It has to put work in to learn everything it needs.  This artificial brainpower expends a great deal of energy.

The human software developers who are responsible for creating the code behind AI will deploy powerful computers which will also place a burden on power infrastructures.

Undertaking a task

Whenever you ask AI to create something for you, you are unleashing a chain of commands which need to be processed.  These cloud-based transactions will take place in data centres – which are fuelled by electricity.

The data centres

Enormous computing capacity is required and large energy-intensive data centres are popping up all around the world.  Professor Gina Neff of Queen Mary University London was quoted as saying on the BBC News website that data centres used to power ChatGPT consume more electricity in a year than 117 countries.

The creation of the data centres

The data centres are costly in terms of energy, not just in their day-to-day operations but also in their construction.  Transmission infrastructure needs to be put in place and, once built, there is an environmental impact due to their energy requirements and the significant amount of water used for cooling.  Alternative cooling methods are being developed, thankfully.  Efficient uses of the waste heat are also being considered.

The computers within the data centres rely on powerful computer chips.  The manufacturing of these chips depends on the use of lots of power and water.  The good news is that innovations in chip development could result in reduced power demand.

Some researchers estimate that the industry as a whole could consume as much as currently consumed by the Netherlands.

Can use of AI ever be justified then?

AI is here to stay.  Businesses will have to implement AI strategies to keep up with the competition.  It can take over the routine tasks, freeing up managers to focus on the business.  It can make staff more productive.  In the realm of medicine, its attention to detail can identify tumours that are not always obvious to the human eye.

The futuristic world where robots do all the work, freeing up people to enjoy a life of leisure, is not a reality … yet.  Some mundane job roles can definitely be replaced, though.

Businesses can certainly benefit from deploying AI.  The environment will thank you, however, if you refrain from more frivolous wastes of energy.  Maybe go easy on the replica Barbie creations!