Login
iscover
Share this article on:
Powered by
MOMENTUM
MEDIA
Powered by MOMENTUMMEDIA
The rise of generative AI tools comes with a heavy environmental cost – but it doesn’t have to be that way, one expert has warned.
As of July last year, Australian firms with a revenue of more than $500 million are required to report their greenhouse gas emissions to the ASX, with smaller companies set to fall in line in the coming years.
The timing is, in some ways, quite fortuitous, as many companies at that scale are rushing to embrace large-scale use of generative AI (GenAI) tools and large language models to streamline work processes and boost productivity.
The rise of AI comes with a growing burden on the environment, and the ASX’s new reporting regime will put related environmental impacts in the spotlight – but that’s not necessarily a bad thing.
“To tackle the AI energy problem, we need to admit it does have a real environmental cost. GenAI isn’t just some magical cloud outputting poetry and code, it’s millions of GPUs crunching vectors and consuming serious power, and most people don’t even realise how much,” Simon Wistow, co-founder of cloud computing firm Fastly, recently said in comments regarding World AI Appreciation Day.
“Indeed, AI efficiency can be engineered at many layers, from the model architecture to the infrastructure beneath it. Organisations, for example, can work on research to make it so that large language models are more efficient to run. If they’re based on integers rather than floating point, that should help reduce energy usage.”
Wistow also suggested standardising AI models, or sharing existing ones, to prevent a constant process of “reinventing the wheel”.
“This could involve reducing work between models or having a standard common model funded by a consortium of companies or governments or science institutions,” Wistow said.
“Innovation and sustainability aren’t mutually exclusive, but right now we’re very much leaning towards camp innovation and experimentation, no matter the cost. But there will come a point of diminishing returns. Training these models is incredibly expensive, so the industry will naturally shift towards optimisation, and companies focus on smaller, more efficient models instead of giant, general-purpose ones.”
According to Wistow, the one thing that is of vital importance is transparency. If we know the true environmental costs of AI, organisations can work to offset those costs and come up with solutions to make the technology more efficient.
“The aim is not to duplicate work but to share effort and work on research so that LLMs are more efficient to run and we can reuse work between models,” Wistow said.
“We all need to start thinking about AI and the internet as something physical – because it is. It uses real resources, generates emissions, and has real-world consequences. It’s time for companies, developers and policymakers to take responsibility.
“That means optimising infrastructure, pushing for stronger regulation, and holding ourselves accountable. Because let’s face it: keeping this planet healthy is a lot easier than trying to terraform a new one.”
David Hollingworth has been writing about technology for over 20 years, and has worked for a range of print and online titles in his career. He is enjoying getting to grips with cyber security, especially when it lets him talk about Lego.
Be the first to hear the latest developments in the cyber industry.
