As the world grapples with the unrelenting pace of Artificial Intelligence, software developers seem adamant on packing an ever-increasing AI feature set into the apps, search engines and operating systems we use each day.
These tools and features offer to help us with our daily work, yet for all their wonderful benefits, they also carry a hidden cost.
Information is everything
Now, there’s a thesis that says each generation takes the previous advances of technology for granted, yet I believe that is not entirely accurate. Sure, a few years back everyone seemed happy enough to trade their privacy and personal data to social media platforms (and Big Tech more generally) for free access to helpful tools, services and entertainment. But, that appears to no longer be the case for your average consumer.
Something happened. Or rather, a few things converged into a happening, one that has had a growing impact on not only the current population, but ensured future generations are less “digitally naive” netizens, and more akin to savvy digital natives. The factors that led to this empowerment started with people speaking up about harmful effects, governments listening and legislation eventually coming into play that forced transparency.
These days, owners of websites and online services are compelled to tell you if they intend to track you, your data or access aspects of the device you are using. Notifications about notifications, cookie popups and hardware access requests are now frequent occurrences for everyone that uses a device to engage with the internet. It’s behaviour that follows the same playbook as food and drug standards. That industry also has to adhere to similar rules, obliged by the government to advise consumers what’s in their consumables, both good and bad, and precisely how much of it is in every portion.
In that way, consumers can make an informed choice.
Why don’t we know, or seemingly care, about what goes into powering AI queries?
That said, there are many times when you may reach for a product without much regard for its composition, and that’s your choice. But at least you have that choice. For some, ingredients are important criteria for health goals or even just recipe success. And for others, it’s a matter of life and death. I’m talking about those with severe allergies, where even the slightest occurrence of a peanut could trigger fatal anaphylaxis. Add to that the regulation of alcohol, sugar, caffeine and salt, all of which can have devastating results if not properly monitored.
Again, information is key.
So why then do we not know, or seemingly care about what goes into powering the AI generated content and results that developers seem so keen for us to consume?
AI is not free
I promise that this piece won’t divulge into a commentary into digital surveillance, the unauthorised trading of personal data or how AI is being trained on the hard work of others (including yourself) without any form of compensation to the original authors of said data. Whilst important, I’d like to pivot to something that impacts every one of us, even those that choose to eschew AI in their daily lives.
No, not vast sums of money, or huge amounts of data.
I’m talking, of course, about our shared home – the environment.
Most people have a fair idea that the “magic” output of AI is actually the result of some seriously powerful computation. Yet, that’s probably as far as one dwells on it, deciding to focus on the impressive end result but not really caring about what it took to get there. See, AI, or at least, the big-time, generative AI that is in the headlines these days, is comprised of a complex structure of large language models (LLM), algorithms, neural networks and a host of other elements that form the “mind” of the operation.
Thought I was going to say brains there? Well, AI does indeed have brains, but in this analogy, they take the form of servers, CPUs and GPUs. Just how many depends upon the size and scope of the AI itself, but you can bet there are hundreds if not thousands of servers in data centers all around the world that are home to these super-charged digital assistants. And more are coming online each day, especially as the Big Tech players have offered their AI’s to be white labeled, while businesses and consumers continue to drive demand.
All that circuitry and processing power requires immense electricity to run. Even more so than your average computations in fact. Because AI isn’t just returning a simple answer in the form of binary zero’s and one’s that are translated into human-friendly words. It’s performing complicated calculations, cross-refencing data across multiple axes, generating content, predicting and transforming your query into different modalities like speech, text or pictures. Then there is the training component to consider, whereby the AI gets theoretically smarter from the lessons it learns off each query, in addition to all the pre-training it conducted by scraping huge amounts of data from the internet, making semantic connections, and being guided by human reinforcement methods.
But how much juice does all that artificial thinking take?
Powerful systems require power at scale
Earlier this year, Microsoft unveiled it’s desire to get into the nuclear energy game in order to power its massive AI ambitions. The company envisions a future where a network of small, modular reactors and microreactors would power the server farms and data centers that house its AI technology.
Then there’s the revelation that Google’s AI offerings accounted for between 10 and 15 percent of its electricity bill back in 2021. Obviously, that was before the release of their Bard chatbot in 2023, which has likely further increased energy demands at the house of G.
That said, AI (and technology more broadly), should not be discarded in favour of environmental preservation. One could argue that without advances in technology, I would be unable to author this content and call for deeper thought and consideration around how we utilise these systems and assess their impact.
AI doesn’t have to become a repeat of what happened with cryptocurrency. In fact, there’s a major point of divergence at play here. Unlike Bitcoin, which extracts a heavy toll on the environment in the pursuit of “disputable” wealth generation, AI has the potential to actually help us solve some of our climate concerns before they become catastrophes.
These systems are more than just souped-up predictive text generators. They have ingested vast quantities of data, which, putting aside the issues of copyright and human obsolescence, could aid us in researching ways to overcome our energy dilemma. Let’s unleash them on humankind’s bigger problems and stop appealing to human-capital adverse CEOs and people who are seeking to outsource their cognition.
Hopefully the answer our artificial friends come back with doesn’t involve an outcome resembling Skynet, but instead, a guiding light for researchers, climate scientists and energy engineers to transform entire sectors and finally remove our reliance on fossil fuels.
In the meantime, it’s up to Big Tech and governments to ensure AI is provided to consumers and businesses in a fair and transparent manner. Perhaps they should take a tip from the food industry and slap a label on each AI search box, detailing precisely how much energy and user data your query will cost.
Because at the end of the day, whether kilojoules or kilowatts, it all comes down to energy.