All around the globe, we’re seeing data servers eat up billions of natural resources and leave behind a carbon footprint that isn’t an ideal scenario when it comes to the growing issues of climate change.
The world has close to 8000 different data centers and these are the cornerstone of today’s online technology. But the scary part is how the stats are not stagnant and they’re expecting to grow exponentially, making it an extremely scary situation when we consider the climate.
According to stats from expert researchers, AI being incorporated into the modern digital world may soon utilize 20% of the world’s electricity generated and that means saying hello to increased carbon emissions by nearly 5.5%.
This is a serious matter of concern, not to mention a mega threat as the startups and firms are coming close to what Silicon Valley is already emitting.
The head of Untether AI whose main goal revolves around producing chips that are more efficient for the atmosphere, says that companies cannot ignore the growing carbon footprint of the digital world.
He called it a pandora box and for that reason, companies have much better productivity but not at the cost of sacrificing the environment. The consequences are real and the impact is severe and before it’s too late, we need to really work hard and find a proper strategy, he adds.
What AI brings forward to the table can serve as consequences faced by the human race for the next decade and that’s why each step related to AI must be calculated. So as you can imagine, the scope that such a mission includes is definitely large.
When it comes down to AI and its respective tools like GPT-4 technology by OpenAI and Google’s Palm 2, we can see two main stages. The first one is linked to training the AI models and the second has to do with the execution phase.
To give you a better understanding of the severity of the situation, researchers from Massachusetts studied how much carbon was emitted during the training of AI models. And it’s shocking to see how the figures were so much that they were the equivalent of using five vehicles.
On the other hand, another latest study by tech giant Google spoke about training GPT-3 and that resulted in close to 552 metric tonnes of pollution in the form of carbon emissions. So you can find that equal to driving a car for two million kilometers.
The latest GPT4 technology received training that’s 570 times more than this. As the systems get more advanced with growing time, we can see how AI is setting the stage for more danger in the future.
Nvidia which is popular for producing AI chips is giving rise to new processors that are necessary for the likes of training models such as GPUs. These happen to be more environmentally friendly which is great news.
The other issue is linked to how generative AI is getting used such as when you have to apply trained models to highlight a certain object or response to a particular prompt. It’s all a matter of inference and that could be an issue with AI tools like ChatGPT that could be utilized by anyone out there and incorporated into one’s everyday life with ease like a common web search.
Yes, chips made by Nvidia are energy efficient, they utilize a lot of power. So the goal is to get AI to be both, energy efficient and consume power and that’s where most companies need to focus on.
Read next: What Countries Have the Highest Digital Quality of Life?
by Dr. Hura Anwar via Digital Information World
No comments:
Post a Comment