AI is on every corner of the internet, whether we like it or not. Website chatbots are now equipped with ChatGPT; AI-generated images are everywhere; our emails and search results are filled with suggestions. ChatGPT has evolved from its primitive first version prone to manipulation and lying to “4o”, which can analyze and create images, write code and search the internet. NVIDIA, the company that makes computer parts essential to training and maintaining AI, has become one the richest companies in the world as their stock prices nearly doubled this year alone. We’re living on a wave of revolutionary discoveries in AI, and it won’t be going away anytime soon.
Technological progress never comes without a catch, though. The blazing fast innovation in AI takes a mind-boggling amount of energy to train and run the programs and cool the computers. The question has become whether this catch is worth it. Althea Delwiche, professor of communication, put into perspective just how much of a power sink these machines are.
“In terms of training and getting the larger models, the predictions are that the next generation of models, the next few years, could take like five gigawatts of power, and that is equivalent to five nuclear power plants, or all of the energy consumed by Manhattan,” Delwiche said.
This sudden spike in energy usage by these computers worldwide could not have come at a worse time. Climate change could accelerate at a much faster rate than we expected as carbon sinks — things that take in lots of carbon dioxide like forests or plankton — fail due to higher temperatures. The past 10 years will be the hottest years on record, and we’ve only got four and a half years to limit warming to 1.5° Celsius before the worst-case scenario becomes real and permanent.
To clarify, it’s not that I’m anti-AI. AI would be a tool for good if it didn’t contribute to the climate disaster. I interviewed Yu Zhang, professor and chair of Trinity’s computer science department, who uses AI for bioinformatics. AI can help doctors save lives and could also help save the climate from itself.
“Many of the AI experts are working on what is called the efficient algorithm. … If you have a more effective algorithm, the training may be just shortened to one day,” Zhang said.
Delwiche also brought up an avenue scientists are exploring to make AI completely energy-free.
“People are looking at alternatives to silicon computer chips using brain material and using actual neurons and flooding them with dopamine … neural networks that actually a number of university researchers are connecting to and using for very, very primitive AI operations,” Delwiche said.
If it sounds terrifying to you, you’re not alone. There are definitely ethical implications to creating an artificial human brain and pumping it with dopamine. The upside is that it doesn’t require nearly any energy, none compared to the vast amounts the supercomputers that run AI take. However, it’s a technology that won’t be up to the standard of computer-trained AI for a while, if ever, and climate change requires solutions today.
An action that meets the urgency of the moment would be government regulation of AI energy use. This could be in the form of taxes for companies that pass a threshold of AI energy usage, or throttling the amount of energy that goes to an AI computer in the first place. There are downsides to this solution, primarily that it would massively slow down the innovation process. This kind of government action will not be happening anytime within the next four years with the election of Donald Trump either.
“I would say with the recent election, we did not elect a candidate known for regulating business and one of his strongest backers, Elon Musk, is also not a proponent of government regulating the technology industry,” Delwiche said.
Innovation is always around the corner, and I believe one day this problem will be put behind us. In the meantime, it is up to us and our institutions to step up where our technology and government cannot. Trinity University should consider banning AI outright across all courses, and we should all think twice about using ChatGPT for daily chores. Climate change needs solutions now, and if it means a few less people using AI in their daily lives, it makes a difference.
