The Comical Conundrum: The Journey Towards Sustainable Artificial Intelligence
“What It Will Take to Make AI Sustainable”
“Artificial intelligence has the potential to change the world, but it comes with an energy cost that’s unsustainable under today’s methods. OpenAI’s GPT-3, for example, consumes the energy equivalent of flying across the country 30 times to train once.”
Turning a blind eye to AI’s energy guzzling habits has become society’s default reaction. Picture Mary Poppins; AI is her charming magic carpet bag that produces one fantastical item after another. But, in the interest of full disclosure, isn’t it time the conversation shifted to the alarming amount of energy this bag devours with a single training session?
Lamenting over the energy-hungry attitude of these algorithms doesn’t change the fact that GPT-3 – OpenAI’s golden child – could essentially consume a flight’s worth of carbon every time it attempts to master a new trick.
If AI is to claim its rightful place as the world-changing technology it aspires to be, it has to sever ties with unsustainable energy practices. Easier said than done, right? How on earth do we convince these energy gluttons to follow a more moderate diet?
See, blindly tossing more computational power in an AI’s direction, imagining it will miraculously stumble upon the solution, is not strategic. The expression “work smart, not hard” has never been more required, like a seasoned personal trainer coaxing their clients into more efficient workout routines.
Won’t it be wonderful if the AI of the future could be effectively trained on less data? Imagine that, mastering a new language or a complicated video game, all while making less of a dent on the planet’s energy resources than a cross-country flight. Not a dream, but a high IQ necessity.
And then there’s the issue of energy sourcing. Like nutrition-conscious humans swapping candy bars for kale smoothies, AI training can and must make the switch from fossil fuels to renewable energy sources.
While governments and other powerful entities scrap over AI domination, they conveniently gloss over the elephant in the room – our Earth is becoming the unintentional casualty of AI’s carbon footprint. Dressing the AI in a green cape might not be a tantalizing prospect for countries racing to the finish line, but undoubtedly, it’s a detour we must take to ensure that the final destination is sustainable and reachable.
Bunny hopping over the energy issue serves no one. AI’s potential glory will inevitably be overshadowed by its preposterous energy consumption if nothing is done. The time has come to prep AI for a sustainable future because, let’s face it, without a sustainable planet, what good will any technological advancement serve?
A sustainable AI isn’t merely a “nice-to-have” footnote on a tech company’s corporate social responsibility agenda. It’s a pressing requirement, an urgent call to action. With as much energy as flying across the country 30 times, training these AI models once succinctly makes the case for sustainable AI. It’s time to rethink, revamp, and perhaps reshape the future of AI, with energy efficiency front and center.